Human Development Network Health, Nutrition, and Population Series Health Economics in Development Philip Musgrove, Editor THE WORLD BANK Selected titles from the World Bank's Health, Nutrition, and Population Series: Better Health Systems for India's Poor: Findings, Analysis, and Options. 2002 (D. H. Peters, A. S. Yazbeck, R. R. Sharma, G. N. V. Ramana, L. H. Pritchett, and A. Wagstaff) The Burden of Disease among the Global Poor: Current Situation, Future Trends, and Implications for Strategy. 1999 (D. R. Gwatkin and M. Guillot) Combating Malnutrition: Time to Act. 2003 (S. Gillespie, M. McLachlan, and R. Shrimpton, editors) Health Expenditures, Services, and Outcomes in Africa: Basic Data and Cross-National Comparisons, 1990-1996. 1999 (D. H. Peters, K. Kandola, A. E. Elmendorf, and G. Chellaraj) Health, Nutrition, and Population Indicators: A Statistical Handbook. 1998 (E. Bos, V. Hon, A. Maeda, G. Chellaraj, and A. S. Preker) Health Policy Research in South Asia: Building Capacity for Reform. 2003 (A. S. Yazbeck and D. Peters, editors) HIV/AIDS in Latin American Countries: The Challenges Ahead. 2003 (Anabela Garcia Abreu, Isabel Noguer, and Karen Cowgill) Improving Women's Health in Pakistan. 1998 (A. G. Tinker) Innovations in Health Service Delivery: The Corporatization of Public Hospitals. 2003 (A. S. Preker, A. Harding, editors) Investing in Maternal Health: Learning from Malaysia and Sri Lanka. 2003 (I. Pathmanathan, J. Liljestrand, J. M. Martins, L. C. Rajapaksa, C. Lissner, A. de Silva, S. Selvaraju, P. J. Singh) Measuring Country Performance on Health: Selected Indicators for 115 Countries. 1999 (J. Wang, D. T. Jamison, E. Bos, A. S. Preker, and J. Peabody) Private Participation in Health Services. 2003 (A. Harding and A. S. Preker, editors) Prospects for Improving Nutrition in Eastern Europe and Central Asia. 2002 (C. Rokx, R. Galloway, and L. Brown) Reducing Maternal Mortality: Learning from Bolivia, China, Egypt, Honduras, Indonesia, Jamaica, and Zimbabwe. 2003 (M. A. Koblinsky, editor) Reproductive Health in the Middle East and North Africa: Well-Being for All. 2001 (A. Aoyama) Social Re-Insurance: A New Approach to Sustainable Community Health Financing. 2002 (D. Dror and A. S. Preker, editors) Toward a Virtuous Circle: A Nutrition Review of the Middle East and North Africa. 1999 (A. Aoyama) Health Economics in Development Human Development Network Health, Nutrition, and Population Series Health Economics in Development Philip Musgrove, Editor Washington, D.C. © 2004 The International Bank for Reconstruction and Development/The World Bank 1818 H Street, NW Washington, DC 20433 Telephone 202-473-1000 Internet www.worldbank.org E-mail feedback@worldbank.org All rights reserved. 1 2 3 4 07 06 05 04 The findings, interpretations, and conclusions expressed herein are those of the author(s) and do not necessarily reflect the views of the Board of Executive Directors of the World Bank or the governments they represent. The World Bank does not guarantee the accuracy of the data included in this work. The boundaries, colors, denominations, and other information shown on any map in this work do not imply any judgment on the part of the World Bank concern- ing the legal status of any territory or the endorsement or acceptance of such boundaries. Rights and Permissions The material in this work is copyrighted. Copying and/or transmitting portions or all of this work without permission may be a violation of applicable law. The World Bank encourages dissemination of its work and will normally grant permission promptly. For permission to photocopy or reprint any part of this work, please send a request with complete information to the Copyright Clearance Center, Inc., 222 Rosewood Drive, Danvers, MA 01923, USA, telephone 978-750-8400, fax 978- 750-4470, www.copyright.com. All other queries on rights and licenses, including subsidiary rights, should be addressed to the Office of the Publisher, World Bank, 1818 H Street NW, Washington, DC 20433, USA, fax 202-522-2422, e-mail pubrights@worldbank.org. Library of Congress Cataloging-in-Publication Data Health economics in development / edited by Philip Musgrove. p. cm. -- (Health, nutrition, and population series) Includes bibliographical references and index. ISBN 0-8213-5570-8 1. Medical economics--Developing countries. I. Musgrove, Philip. II. Series. RA410.5.H4133 2003 338.4 33621 091724--dc21 2003057155 Contents Foreword xiii Acknowledgments xvii Introduction 1 1 What Is the Minimum a Doctor Should Know about Health Economics? 23 2 Public and Private Roles in Health 35 3 The Rationale for Government Intervention in the Tobacco Market 77 4 Measurement of Equity in Health 105 5 What Should Consumers in Poor Countries Pay for Publicly-Provided Health Services? 121 6 Compensatory Finance in Health: Geographic Equity in a Federal System 133 7 An Ounce of Prevention Is Worth How Much Cure? 143 8 DALYs and Cost-Effectiveness Analysis 153 9 Criteria for Public Spending on Health Care 167 10 Cost-Effectiveness and the Socialization of Health Care 187 11 Is the Eradication of Polio in the Western Hemisphere Economically Justified? 203 12 Cost-Benefit Analysis of a Regional Vaccination System 223 13 Cost-Effective Malaria Control in Brazil 247 14 Do the Poor in Brazil Pay More for Food? 277 15 Do Brazilian Nutrition Programs Make a Difference? 297 v vi · Contents 16 Economic Aspects of Food Fortification 325 17 Malnutrition and Dietary Protein 341 18 Family Health Care Spending in Latin America 361 19 Basic Patterns in National Health Expenditure 375 20 Economic Crisis and Health Policy Response 401 Tables I.1 Principal Differences between Health Insurance and Insurance for Nonhuman Assets 4 I.2 Principal Differences and Similarities between Education and Health 6 4.1 Measures Related to Equity in Health Care, Peru, 1982: Health Care Resources and Sanitation Services 109 4.2 Measures Related to Equity in Health Care in Peru, 1984: Morbidity and Medical Attention 110 4.3 Measures Related to Equity in Health Care in Peru, 1984: Consultations, Hospitalizations and Expenditures 111 4.4 Measures Related to Equity In Health Care in Peru, 1984: Vaccination Coverage (%) of Children Under 5 113 11.1 Costs and Benefits Associated with Polio Eradication During a Successful Five-Year Campaign and an Ensuing Ten-Year Maintenance Period, Assuming all Polio Victims are Treated 209 11.2 Costs and Benefits Associated with Polio Eradication During a Successful Five-Year Campaign and Ten-Year Maintenance Period, Assuming only a Fraction of all Polio Victims are Treated 214 11.3 A Comparison of Ten-Year Costs and Results of Two Hypothetical Polio Eradication Campaigns, One Immediate and One Delayed, Discounted at 12% per Year 218 12.1 Cost of SIREVA, by Year, in Constant US$ 226 12.2 The Projected Numbers of Individuals to be Vaccinated, by Target Disease and Year 229 12.3 Total Discounted Numbers of Disease Cases Prevented, Deaths Prevented with Treatment of all Cases, and Deaths Prevented without Treatment of any Cases, over 20-Year and 30-Year Horizons 229 12.4 The Estimated Cost of Treatment (in Constant US$), Probability of Preventing One Case, and Implied Maximum Cost of Vaccination, by Disease, Independent of the Number of Vaccinations 234 Contents · vii 12.5 The Average Implied Maximum Cost of Vaccination (in Constant US$) Derived from the Number of Individuals Vaccinated, Treatment Costs, and the Cost of SIREVA, for 20-Year and 30-Year Horizons 238 12.6 Implied Minimum Benefit per Case Prevented and per Death Prevented as a Function of the Cost of Vaccination, without Considering Patient Treatment (20-Year Horizon) 240 12.7 The Effects of Adjusting for the Delay between Vaccination and Disease Onset upon the Implied Maximum Cost of Vaccination and upon the Implied Minimum Benefits of Preventing a Disease Case, in Constant US$ 242 13.1 Population (Millions) and Number of Municipalities by Risk of Malaria Transmission, Based on API, 1988­1995 250 13.2 Estimated Cases, Lives and DALYs Saved by Preventing and Treating Malaria, 1989­1996. Cases Prevented and Cases Reported and Savings in Lives and Morbidity 258 13.3 Estimated Cases, Lives and DALYs Saved by Preventing and Treating Malaria, 1989­1996. Conversion of Savings in Lives and Cases to Disability-Adjusted Life Years 259 13.4 Costs of the Malaria Control Program, 1989­1996 (Thousands of 1996 US$, Discounted at 3% for 1989­1995) and Cases Treated (Thousands of Hospital or Ambulatory Patients) 262 13.5 Cost-Effectiveness of Saving Lives from Falciparum, 1989­1992, 1993­1996 (Discounted Present Values). Most Plausible Assumptions 265 13.6 Cost-Effectiveness of Saving Lives from Falciparum, 1989­1992, 1993­1996 and 1989­1996 (Discounted Present Values). Sensitivity Analysis: Cost-Effectiveness of Prevention Related to Projected Incidence (No Change in Severity or Lethality) 266 13.7 Cost-Effectiveness of Saving Lives from Falciparum, 1989­1996 (Discounted Present Values). Sensitivity Analysis: Untreated Falciparum Case Fatality Rate 25 or 5% 268 13.8 Effects of Reduced Incidence, Severity and Lethality on Mortality Reduction 270 viii · Contents 14.1 Sample Size (Number of Retailers) by City Size, Type of Neighborhood, and Type of Establishment 281 14.2 Mean Prices, in June 1985 Cruzeiros, of Selected Basic Foods by Municipality for Small Retailers Not Affiliated with Subsidy Program 283 14.3 Mean Prices, in June 1985 Cruzeiros, of Selected Basic Foods by City Size for Small Retailers Not Affiliated with Subsidy Program 285 14.4 Mean Prices, in June 1985 Cruzeiros, of Selected Basic Foods by Type of Establishment: Three Metropolitan Areas 286 14.5 Mean Prices, in June 1985 Cruzeiros, of Selected Basic Foods by Type of Establishment: Six Medium-Size Cities 286 14.6 Prices as Functions of City, Neighborhood, and Type of Establishment: Regression Coefficients 287 14.7 Quadratic Regression of Price versus Quantity for Fractional Purchases of Eight Basic Foods, in all Establishments and Small Retail Outlets only 290 15.1 Principal Characteristics of Four Nutrition Programs, Brazil, 1974­86 299 15.2 Aggregate Estimates for Physical Distribution of Foods, Total Cost, and Total Public Expenditure at Constant June 1984 Prices, for the PNS, PROAB, and PCA, from the Beginning of Each Program through the End of 1986 301 15.3 PINS: Number of Families Entitled to Purchase Subsidized Food ("Participating") and Number of Families that Actually Bought, by Subsidy Model, 1978­80 303 15.4 Changes Observed in Weight for Age in Children Participating in the PNS in Two Municipalities in São Paulo, March 1980­April 1982, According to Frequency of Visits to Receive Food; Expressed as Number of Children 305 15.5 PNS, PCA, and PINS: Cost Structure in Percentage Terms (Data for One or More Years between 1978 and 1980) 307 15.6 Recife, Pernambuco: Subsidy Level (Percentage) and PROAB Prices (Percentage Difference from Supermarket Price, with and without Subsidy) for Six Products, June 1980­July 1986 308 15.7 Actual Transfer of the PROAB Subsidy to the Consumer, in Relation to Minimarket Prices, Recife, March-August 1984 309 Contents · ix 15.8 Changes Observed in Children Participating in the PCA, PNS, or PINS, According to Anthropometric Criterion, Duration of Participation, and Initial Nutritional Status, 1976­80, Expressed as Number of Children 312 15.9 Changes in Weight for Age in Initially Underweight Children Participating in the PNS or PINS, by Program, Initial Age, and Duration of Participation, Expressed as Number of Children 314 15.10 Number of Children Initially Aged Under Six Participating in the PINS, According to Initial and Final Nutritional Status, by Anthropometric Criterion and Duration of Participation, 1978­80 316 15.11 Changes in Weight for Age Observed in Children Initially Aged Under 12 Months, Participating in the PINS, According to Duration of Participation, 1978­80, Expressed as Number of Children 318 15.12 Changes Observed in Children Participating in the PNS, Initially Aged between 6 and 12 Months Old, According to Anthropometric Criterion and Duration of Participation, 1976­80, Expressed as Number of Children 319 15.13 Weights of Live Births to Mothers who were Beneficiaries of the PNS and PCA, Compared with Non-Beneficiaries 320 17.1 Variables for Urban China in 1979 345 17.2 Determinants of the Height and Weight of Young Adults (Age 18­25 Years) in Urban China in 1979 346 17.3 Variables for Rural Chinese Counties in 1983 348 17.4 Determinants of Adult Height and Weight in Rural Chinese Counties in 1983 348 17.5 Variables for International Comparisons 350 17.6 Determinants of Adult Male Height (41 Populations in 40 Countries) and Adult Female Height (33 Populations in 32 Countries), ca. 1960 351 18.1 Elasticities with Respect to Total Expenditure 364 18.2 Estimated Total Family Expenditure per Person and Private Health Care Expenditure, by Quartile 1 (low) to 4 (high) of Total Spending per Person, in Ten South American Cities (1968 Dollars per Year) 365 18.3 Mean Values of Total Family Expenditure (PCE) and Private Health Care Spending (HEA), within Each of Nine Classes of Total Family Expenditure, x · Contents by Urban and Rural Areas, in Four Regions of Brazil (Cruzeiros of August 1974 per Year). 367 18.4 Brazil, 1974: Private Family Health Care Spending as a Function of Total Family Expenditure, Region and Metropolitan/Other Urban/Rural Differences 369 18.5 Mean Values of Private Family Expenditure on Drugs and Medicines and on Hospitalization and Surgery, within Each of Nine Classes of Total Family Expenditure, by Urban and Rural Areas, in Four Regions of Brazil (Cruzeiros of August 1974 per Year). 371 19.1 National Health Accounts Estimates for 191 WHO Member States for 1997, Revised Data as of 31 May 2001 377 19.2 Countries Grouped by WHO Region, Mortality Stratum, and GDP per Capita 384 19.3 Regression Statistics for Health Expenditure as a Percentage of Gross Domestic Product 388 19.4 Regression Statistics for Out of Pocket Payments as a Percentage of Total Health Expenditure 390 19.5 Regression Statistics for Public Health Expenditure as a Percentage of Total Health Expenditure 392 19.6 Regression Statistics for Public Health Expenditure as a Percentage of Total Public Expenditure 394 19.7 Regression Statistics for Out-of-Pocket Payments per Capita as a Function of Income per Capita 395 19.8 Regression Statistics for Total Health Expenditure per Capita as a Function of Income per Capita 396 19.9 Regression Statistics for Total Public Expenditure per Capita as a Function of Income per Capita 396 20.1 Indices of Public Expenditures on Health (Central Government or Total Public Sector) in Thirteen Latin American and Caribbean Countries, 1980­86 (1982 100) 410 Figures I.1 Decision Flowchart, Sources of Health Financing 17 2.1 Three Domains of Health Care 40 2.2 Need, Demand and Supply for Health Care 55 2.3A Appropriate Use of the Instruments of Public Intervention in Health Care 66 Contents · xi 2.3B Typically Inappropriate Use of the Instruments of Public Intervention in Health Care 67 4.1 Schematic Stages of Illness or Accident, Treatment and Outcome 106 4.2 Lorenz Curves of Inequality of the Distribution of Physicians Rrelative to Population (by Department) and of Ministry of Health Patient-related Expenditure Relative to Population with Symptoms (by Health Region), Peru, 1982 115 5.1 Demand for Medical Care as a Function of Income and Prices 128 7.1 Marginal Costs of a Curative and a Preventive Intervention, when Prevention is Always Cheaper and Everyone Receives One Intervention or the Other 144 7.2 Marginal Costs of a Curative and a Preventive Intervention, When Only Part of the Population Requires the Curative Intervention 146 7.3 Cost-minimizing Distribution of Effort between a Curative and a Preventive Intervention 148 9.1 Nine Criteria for Public Spending on Health Care 169 9.2 Cost-effectiveness versus Horizontal Equity: Comparison of Two Interventions Where All Beneficiaries are Identical Except for Costs 175 9.3 Cost-effectiveness versus Vertical Equity: Comparison of Four Interventions Differing in Disease Severity, Intervention Effectiveness and Cost 178 9.4 Decision Tree for Public Resource Allocation in Health Care 182 11.1 Costs and Benefits of Polio Eradication, Assuming Treatment of Only Some Victims or Reduced Numbers of Cases 215 13.1 Cases of Malaria and of Falciparum and Total Budget for Malaria, 1975­1996 252 18.1 Family Health Spending versus Total Family Spending, by Income Quartile, in 10 South American Cities 366 18.2 Brazil: Family Health Spending versus Total Family Spending by Income and City or Region 368 18.3 Brazil: Family Spending on Drugs and Medications versus Total Family Spending by City or Region 372 19.1 Total Health Expenditure as % of Gross Domestic Product (GDP) versus GDP per Capita (191 Countries) 386 xii · Contents 19.2 Out-of-Pocket Expenditure as Proportion of Total Health Expenditure (THE) versus Gross Domestic Product (GDP) per Capita (191 Countries) 386 19.3 Public Expenditure as Proportion of Total Health Expenditure (THE) versus Gross Domestic Product (GDP), per Capita (191 Countries) 387 19.4 Tax-funded and Other Expenditure as Proportion of Public Expenditure on Health (PEH) versus Gross Domestic Product (GDP), per Capita (191 Countries) 387 19.5 Hypothesized Needs and Actual Spending for an Essential Package of Health Services versus Gross Domestic Product (GDP) per Capita 398 20.1 Effects of Economic Crisis on Health 403 Boxes 12.1 Discount Factors, by Year, for a Rate (r) of 10% per Year 227 12.2 Glossary of Symbols, Variables, and their Relationships--in their Approximate Order of Appearance in the Text 232 Foreword My pleasure at being asked to write this foreword for Philip Musgrove's collected papers is both personal and institutional. The personal pleasure derives from my association with him and the respect I have developed for his thinking and work over the past 20 years. It was Philip who critically reviewed for me my first major address on the subject of health and devel- opment, and I remember with pleasure his providing me with a list of required reading on the subject and insisting that I read and reread Selma Mushkin's seminal work on health as investment. The institutional pleasure arises because it was Philip who in my view introduced the appropriate line of thinking about health economics into the Pan American Health Organization in the decade of the eighties. I had been convinced that the majority of health economists were concerned mainly with the costs of health services and presenting to the developing countries various recipes for reduction of costs, including the employment of user fees. Philip Musgrove's early work that I recall was a critical analy- sis of the impact of the economic crisis of the eighties on the health serv- ices and health outcomes in Latin America and the Caribbean. It was partly from this work that he drew the conclusion, which is set out ele- gantly here, that public financing of health should be countercyclical. This very sensible view has yet to gain widespread acceptance. I believe that one of the great virtues of this book, and Philip's papers in general, is that it is presented in eminently readable fashion and avoids the turgid writing that obfuscates rather than illuminates. This should be read by all health workers who are interested in the practical treatment of many of the problems that are usually dealt with by economists in the jar- gon of their guild. xiii xiv · Preface It is gratifying to see the really difficult issues, such as the role of the state or the market, treated so clearly, avoiding the all-or-nothing thinking that unfortunately can be common among health workers. These two major actors in the organization and financing of health services are nei- ther lionized nor demonized here. It is clear from Philip's papers that they both have their place even, or especially, in societies that are committed to equity in health outcomes as well as in the determinants of those out- comes, with emphasis on the role of the health services. Almost all of his conclusions need to be repeated to health workers and policy makers in all sectors, but the one that is likely to strike the loudest chord is his assertion that, except for the most inexpensive care, out-of­pocket spending is prob- ably the worst way to pay for health services. I hope that those who read this book Health Economics in Development will be convinced to abjure forever the loose thinking that goes into the expression "social sector." I appreciate that this will be difficult, but no one who reads his arguments can fail to be convinced of the naivety if not absurdity of the expression. This is not a matter of semantic jousting; experience shows that the public budget for health suffers in more ways than one when it is included in a fixed social sector budget, and govern- ments studiously avoid discussing the merits of health spending in the dis- tribution of public resources. In his discussion on the cost benefit analysis of programs, Philip refers to the effort to establish the economic benefit from the program to eradi- cate poliomyelitis in the Americas. The outcome is now history, and for the past eleven years there has been no transmission of polio in the Western Hemisphere. There was never any doubt about the technical feasibility or the social desirability of the program, but when Philip demonstrated that it was also economically beneficial, the stage was set for the success of what has to be one of the most remarkable achievements of public health in recent times. That success led to the decision to mount a worldwide effort, which is now approaching its final goal of universal eradication of polio. I hope this book will be well received and widely promoted, not only by financial institutions like the World Bank, where much of this material was produced, but also by institutions such as the Pan American Health Orga- nization, where Philip first began to work on health. There is a need for critical and dispassionate thinking about all the issues he addresses. For those of us in the public health profession, there will be special interest in his treatment of the concept of equity and the possibility of equity in the Preface · xv provision of health care in countries where health services are segmented-- to the detriment of the majority of the population and especially of the poor. I wish the book well and recommend it to you. Sir George Alleyne Director Emeritus Pan American Health Organization World Health Organization Washington, D.C. Acknowledgments Numerous colleagues and friends, particularly at the World Bank but also at the Pan American Health Organization, the Inter-American Develop- ment Bank, and the World Health Organization, encouraged me in the only two beliefs that might justify this venture. One is that at least some of the papers included deserve a longer life and a wider audience than the original publications alone would provide; the other is that because of the development of certain ideas over the years and the connections between one piece and another, putting the papers together might create a whole greater than the sum of its parts. This collection owes its existence to Alexander Preker, who insisted on its potential when I doubted it, and to James Christopher Lovelace, who allowed me to devote to it the necessary time during my last months at the World Bank. I have also accumulated debts, beginning more than twenty years ago, to all those who provided the environment and the experiences from which these ideas grew. These include a large number of government offi- cials and scholars, especially in Argentina, Brazil, Chile, and Colombia, with whom I had the pleasure of working on empirical studies and health sector reform projects, and the many PAHO and World Bank colleagues who collaborated on those efforts. I trust they will forgive the failure to name them all individually. The acknowledgments for several chapters thank those whose help was particularly important in stimulating or improving a paper or in providing information and direction. Turning a collection of papers into a book would have been impossible without Nicki Marrian's repeated guidance and encouragement. Daniel Cotlear, María Luisa Escobar, and Adam Wagstaff commented extensively on the first full draft, and contributed greatly to making it both leaner and more thoughtful. The final organization of the volume, and especially that xvii xviii · Acknowledgments of the Introduction, are due to their suggestions. Wagstaff also suggested the felicitous title. Olufumilayo Orisadipe undertook the tedious task of retyping papers written so long ago that there were no electronic files from which to work. Yvette Atkins and Sithie Naz Mowlana kept track of successive revisions. Rick Ludwick supervised the transformation of a manuscript into a book, with unfailing patience and efficiency. A Note on Authorship Five of these papers were written with co-authors, who in three cases deserve the lion's share of the credit. Chapter 13 derives from a much larger study by Dariush Akhavan for the World Bank. At the invitation of Alexandre Abrantes, who was the task manager for the Bank malaria con- trol project, I condensed the study and developed the graphic and tabular presentations to emphasize how a change in strategy improved the cost- effectiveness of control efforts. The fourth co-author, Renato d'A. Gusmão, of the Pan American Health Organization, was responsible for the introduction of the new malaria control strategy into the World Bank project. Chapter 14 also derives from a larger study, by Osmil Galindo, undertaken for the Pan American Health Organization under my direc- tion at the Fundação Joaquim Nabuco in Recife, Brazil. Besides oversee- ing Galindo's work, my contribution was to condense the study and put it into English (the original was in Portuguese). Dean T. Jamison and Joanne Leslie wrote the first version of Chapter 17, and later invited my help with reconstructing and interpreting the calculations and with the final writing. Chapters 3 and 19 are more conventional cases of co-authorship; I am most grateful to all the colleagues who shared in these efforts. I had the satisfaction of contributing to two major publications that are not excerpted here. One is the World Bank's World Development Report 1993: Investing in Health, for which I wrote much of Chapter 3 and part of Chapter 2. The other report, for which I wrote Chapters 1 and 2 and also edited the rest of the text, is the World Health Organization's World Health Report 2000­Health Systems: Improving Performance. Both these reports were such joint products with several co-authors, from whom I learned a great deal, that it does not seem proper to appropriate any part of either one as my own work. Many of the key ideas in these two documents, which deal with the role of the state in health, with the use of cost-effectiveness analysis, and with how health is (or should be) financed, are either foreshadowed by, or subsequently developed in, several of the papers included here. Introduction The papers in this collection span 21 years of thinking and writing about health economics, first at the Pan American Health Organization (1982­1990) and then at the World Bank (1990­2002, including two years, 1999­2001, on secondment to the World Health Organization). They are divided into six general topics, which together touch on several of the major issues in this field. Chapters 1 through 3 concern the connection between health, particularly public health, and economics--a connection that has occupied much of my professional effort, in part because I started to work on the subject in an organization dominated by public health professionals, and only later moved to an organization dominated by other economists. Chap- ters 4 through 6 treat several different aspects of equity, while chapters 7 through 17 deal with effectiveness and efficiency, first in general terms and then with specific attention to communicable diseases and to malnutrition. Equity and efficiency are among the main issues in any branch of econom- ics, and--as several chapters illustrate--they often cannot be sharply separated. Chapters 18 through 20 concern how health is, and how it should be, paid for--questions that involve both equity and efficiency. Calling the collection Health Economics in Development conveys two mean- ings: that the science of health economics itself has been developing during the period covered, and continues to do so; and that the application of health economics contributes to development, broadly defined, both by improving health and by reducing the waste of resources devoted to health care. As to the first, health economics is a relatively young subdiscipline, roughly 40 years old and expanding rapidly. In 1982, the Pan American Health Organi- zation (PAHO) sought to hire an economist with 10 years' experience in health issues in Latin America. That was a novel departure for PAHO, and it was resisted by a number of staff members trained in public health and deeply suspicious of economists and their ways of thinking. Moreover, there was probably no one alive who could meet the job requirements then. By the end of the century, the situation was quite different: PAHO not only 1 2 · Health Economics in Development employed several economists but promoted and published studies of the relations between health and development. This change of attitude reflects the growing realization that economic thinking is not inimical to the ethical concerns of health professionals, and that the expansion of health econom- ics as a field of inquiry has made it steadily more relevant and more useful to health organizations. As to the second, there is increasing evidence that improvements in health, far from being a pure consumption good or even a luxury, often represent valuable investments in people's capacities to learn and to work, and are sometimes essential to rescue people from poverty or prevent their impoverishment (Ruger, Jamison, and Bloom 2001). As is natural for papers written over a long interval on a variety of topics, these pieces originated in several different ways. The majority are responses to specific requests from supervisors or colleagues, or (in two cases) friends. This is the case for chapters 1 through 3 and 11 through 17, as well as chapter 19. Only chapters 4, 5, 7, 9, 16, 18, and 20 were completely unpro- voked either by such requests or by criticism of previous work. Considering the time devoted at the World Bank to health projects, chiefly in Argentina, Brazil, Chile, and Colombia, it may seem odd that few papers resulted from those efforts. Only chapters 6 and 13 derived directly from Bank projects, in the former case as part of the design and in the latter from ex post evalua- tion. The kind of analysis and the style of writing required to develop proj- ects do not often lend themselves to publication for a wider audience. In fact, the influence probably ran more often the other way, from the ideas in a paper to the way a project was conceived or conducted. I am grateful to the publishers of the journals or other sources in which these papers first appeared for permission to reprint them here. Each pub- lisher's approach to punctuation and the numbering of references and notes has been retained. What is Peculiar About Health There are two strongly connected issues in chapters 1 through 3: first, what it is that distinguishes health from other sectors of the economy, what makes it peculiar in economic terms; and second, what the consequences of those peculiarities are for the appropriate roles of the state and the market. That is, how far should governments interfere in the markets for health care and other determinants of health, to what ends, and with what instruments of intervention? This continues to be a contentious topic, both because the economics of health is complicated and because there is sometimes too Introduction · 3 much ideological rigidity, oversimplified theory, and mutual misunder- standing on both sides of the discussion. Health is emphatically a place where "market fundamentalism" is misguided, but what one might call "public health fundamentalism" can be equally mistaken. Chapter 1 argues that it is valuable for health professionals to understand better how economists think in general, and how they approach issues in health in particular. A certain minimum understanding is desirable, but that understanding does not require any specific knowledge of many areas of eco- nomics. (Some public health specialist should perhaps write the mirror- image paper on the minimum that economists should know about health and medicine. Hsiao (2000) has provided an economist's view on what econo- mists, particularly macroeconomists, not familiar with health issues should know about the subject.) Economists readily invade other professions' domains, in part because most microeconomics is about something particu- lar and not just about "the economy," and also because economists travel light and carry relatively little baggage in the form of data or models specific to their subject (Hirschleifer 1985). Medical professionals, in contrast, must know a great deal of extremely detailed information, which makes them less prone to incursions into others' territory and often leads to a more defensive posture. These differences in the kind and amount of knowledge they need for their work help explain why the two professions often have trouble com- municating with each other, to the detriment of how health systems function. Chapter 2 is an excerpt from a much longer World Bank Discussion Paper, specifically about public and private roles in health. The empirical material in that piece, on health expenditure and relation to health outcomes, has been omitted, because chapter 19 includes more recent and reliable infor- mation. The theoretical or conceptual discussion has also been abbreviated, while trying to preserve the central ideas. Four of these, which can--I hope-- be widely accepted, are worth mentioning here. The first is that all health interventions can be classified into just three groups, depending on how far they are public or private goods (in the economist's sense) and how much they cost. This classification links the characteristics of interventions to the ques- tion of who is to pay for them, emphasizing the importance of insurance or other forms of prepayment, including public financing. Second, the distinc- tion between catastrophically costly interventions and those that consumers can afford to pay for out of pocket depends on people's incomes. Poverty greatly complicates the public/private distinction in the health sector, and becomes a justification for public spending on grounds of equity. The rela- tion among income, costs of interventions, and public subsidy is treated fur- ther in several chapters, particularly 5 and 9. Third, delivering the right 4 · Health Economics in Development health interventions to the right people requires a coincidence among need, demand, and supply, which is why the task is so difficult. It is also the reason why most health systems are a mixture of the state and the market, each of which is prone to certain failures which the other can at least partially offset. The fourth idea is that what really makes health peculiar is rooted in biol- ogy. The asset that health interventions exist to protect, mens sana in corpore sano, is unlike any nonhuman asset in several crucial ways, starting with the fact that one cannot separate oneself from the asset (Miller 1978). These dif- ferences are what make health insurance unlike any other form of insurance, and make health financing more complicated than the financing to protect any other type of asset. Table I.1 summarizes some of the most important dif- ferences, which carry numerous implications for both equity and efficiency. There are of course other human assets than one's state of health, notably one's knowledge and skills, including those acquired through formal education. All these are forms of human capital, and as such have certain Table I.1 Principal Differences between Health Insurance and Insurance for Nonhuman Assets TYPE OF ASSET CHARACTERISTICS OF THE INSURANCE DWELLING VEHICLE BODY Is the asset itself insured? Yes Yes No Does the asset have a well-defined market value? Yes Yes No Can the asset be replaced? Yes Yes Only some parts Can the asset be alienated? Yes Yes Only some parts Can a substitute be used while the Yes Yes No asset is repaired? Does the insurance cover catastrophic Yes Yes Yes damage? Is the owner responsible for maintaining Yes Yes Only in the asset? part Does the insurance cover ordinary wear No No Yes and tear? Is the insurance cost related to owner No Yes No behavior? Does the insurance pay directly to Yes Sometimes Sometimes the owner? Are there guarantees from the asset Sometimes, Yes, if new No producer? if new Does someone else pay for the insurance? No No Often Introduction · 5 features in common. Nonetheless, the peculiarities of health are such that no other sector is much like it, and in particular, it differs in several fundamen- tal ways from education. These differences are partly intrinsic and partly socially determined; on balance, they make schooling and health care more unlike than similar. If schooling were really very much like health care, one would see attacks of ignorance whose sufferers were rushed to a university for specialized care; parents would seek to buy education insurance against such costly risks; nonemergency students would visit school only occasionally, and only chronic cases would spend much time there. One crucial difference is that there is nothing in education corresponding to referrals in medical care. A sick or injured person can be referred "up" from a health center or physician to a hospital and referred "down" when hospital care is no longer required. There is a natural hierarchy of organizations and treatments in health care, but there is no natural sequence like primary educa- tion followed by secondary schooling followed by university or other higher- level training. If the health system worked the same way as education, no patient could get into a hospital until he or she had spent years at health posts, and then more years attending clinics. In schooling, the worse results are at one level, the harder it is to proceed to the next higher one; in health care the exact opposite is true. This is one of the reasons why health care costs increase more rapidly than educational costs: those who fail primary school are not sent, at great expense, to college. To emphasize these differences is to attack the common and rather imprecise idea that there is something called "the social sector" (Castro and Musgrove 2000), which is relatively homoge- neous and is sometimes quite wrongly distinguished from "the productive sector." Lumping, on this logic, is more dangerous to clear thinking and sound public policy than splitting is. Table I.2 summarizes some of the key differences and similarities between the two sectors, noting that the similari- ties are greatest at the primary level, with increasing differences as emergen- cies and more complex levels of health care are considered. The arguments in chapter 2 are illustrated by reference to several issues in the public/private balance, but no single issue is treated in detail. Chapter 3 applies some of the general ideas to one such issue, that of how far the state is justified in interfering with consumers' smoking habits. This is a much- disputed question, pitting public health views (actions bad for one's health should be suppressed) against those of consumer sovereignty (people should be free to decide for themselves whether to smoke). The only easy point of agreement between advocates of these different views is that people should be informed about the undisputed health risks of smoking. Simple-minded 6 · Health Economics in Development Table I.2 Principal Differences and Similarities between Education and Health HEALTH CARE PRIMARY HEALTH CHARACTERISTIC IN GENERAL CARE SCHOOLING Responds to Sometimes Sometimes, often with Never emergencies referral Predictable demand Only in part More predictable than Largely for health in general Insurance market exists Yes Yes; less needed than No for health in general Nature of Episodic Episodic, sometimes Cumulative improvements cumulative Natural "good" state Yes Yes No exists Hierarchy of facilities Yes Yes Yes Referrals among facilities Yes Yes No Tendency to cost Strong Less than for health in Slight escalation general Concentrated early No Yes Yes in life Time-consuming Sometimes Sometimes Always Uniform treatment Sometimes More uniform than for Usually health in general Measurement of quality Very difficult Difficult Relatively easy Universal coverage All services Package of "basic" Up to some level services Gains from universal Nondecreasing Nondecreasing Decreasing coverage Public budgeting Difficult Less difficult than for Easy health in general Public finance Quite variable Quite variable Always high Relation to technical Technophilic Less technophilic than for Technophobic change health in general Externalities Communicable Communicable General and diseases diseases diffuse Concern for equity Yes Yes Yes Share of spending Low Higher than for health Very low on the poor in general Powerful providers Yes Less than for health Yes in general economic thinking would stop there, but chapter 3 develops several economic arguments why simply providing information, and relying on consumer rationality for the rest, is an inadequate response to the dangers of tobacco use. One of these arguments concerns the young age at which people typi- cally start smoking, and illustrates the general idea from chapter 2 that chil- dren do not fit the model of homo economicus, able to take risks into account and make sensible decisions about them. The addictive nature of tobacco also undermines an overly market-oriented view of the issue. The attempt to Introduction · 7 reconcile smoking with the notion of a rational consumer leads to the "rational addiction" model, which is an oxymoron and is, in practice, hardly different from models of habit formation that do not include addiction. Economic science is not yet very good at incorporating aspects that under- mine its own standard assumptions about how human beings behave. Judging and Promoting Equity Chapter 4 proposes a scheme for thinking about equity at each of several stages of an idealized or simplified episode of illness or accident. One may or may not become sick or hurt; receive treatment or not, when needed; be cured or at least benefit from treatment, or not; and recover without treatment, ver- sus dying or continuing to suffer poor health. For a given illness episode, equity can be very different from one stage to another, and the amount of inequity (and the inequality from which it derives) depends on how the pop- ulation is classified--by age, gender, income, location, or other characteris- tics. The conceptual scheme is filled in, so far as household and health system data allowed, with information from Peru. Where the health system is con- cerned, equity can also look very different depending on which resource-- physicians, nurses, hospital beds, or money--or which activities--ambulatory consultations, hospitalizations, or immunizations--is studied. Given all these distinctions, it does not make sense to try to summarize equity in a single measure. Matters only become more complicated when one considers the equity of how health is paid for, since equity in finance can be interpreted in several different ways. Moreover, equity in finance is no assurance of equitable treatment, nor does financial inequity necessarily prevent equitable care. Chapter 5 takes up one financing issue related to equity: it asks what prices consumers should pay for health care when payments do not have to cover costs because the services are publicly subsidized (and, often, publicly pro- vided as well). The key assumption, which attempts to model how a govern- ment in a poor country might think, is that the Ministry of Health values both the quantity of services demanded and delivered and the revenue obtained from patients. The first element implies fees as low as possible, given that demand falls as prices rise; the second may justify substantially higher fees, if demand for health care is relatively inelastic. The model can also be complicated by allowing for price discrimination among consumers according to their incomes, and by trying to distinguish between necessary and frivolous demand. The accumulated empirical evidence concerning these issues strongly suggests that it is difficult to administer a user fee system so as 8 · Health Economics in Development to bring in significant revenue and at the same time to protect those least able to pay (Creese 1991; Newbrander, Collins, and Gilson 2000). It also seems clear that consumers' reactions to fees for health care do not follow a med- ically sound distinction between care that is more needed or justified and interventions that can be considered more frivolous or of low priority (Lohr et al. 1986). Prices or fees can certainly be used to ration health care, as with any other good or service, but not necessarily to rationalize its use, as is some- times carelessly assumed. If the object is to promote needed care and dis- courage what is less needed, other means are required, and the burden of making that distinction should not fall primarily on patients. Chapter 6 takes up a very different aspect of equity, the geographic distri- bution of resources to finance health care--specifically, how to allocate funds from a national government to state or provincial governments in a federal sys- tem in such a way as to compensate for differences in income, health needs, or capacity to raise revenue. Analytically, this question fits between a larger and a smaller issue. The larger issue is how all intergovernmental transfers are deter- mined, including those that are not earmarked or that are designated for par- ticular uses other than health. The net fiscal impact on a state or other subna- tional unit may be quite inequitable even if the health-specific transfer is equity-enhancing, and vice versa. Moreover, the net impact on the health of the state's population may differ from what the health transfers aim to achieve if other transfers facilitate or hinder spending by the state out of its own resources. The smaller issue is how the transfers specifically intended for health are actually used. A national government may preferentially distribute funds to a state or province because on average, its inhabitants are poorer or sicker than those of other states; but if the state then uses those funds primarily to benefit the richer or healthier part of its population, equity is hardly served. The issue treated in chapter 6 is therefore far from a full exploration of how intergovernmental transfers promote equity in health or fail to do so. It is nonetheless ethically important to design those transfers to be equitable in their own right, and politically they can be of great importance in federal sys- tems. The United States, Canada, and Brazil furnish three quite different examples of how federally organized countries have determined such alloca- tions. All three schemes have two significant features in common. First, the formula for allocation contains only one arbitrary parameter, which has the advantage of concentrating political attention on a single decision. Second, the variables that go into the formula offer little scope for either level of gov- ernment to take advantage of the other or to manipulate the outcome. These seem like valuable principles to follow in many different circumstances where Introduction · 9 resources have to be shared, so that what is judged technically to be equitable can also be readily understood and therefore accepted politically. Costs and Outcomes: Effects and Efficiency Chapters 7 through 10 are conceptual explorations of different aspects of effectiveness and efficiency. Chapter 7 asks the question of how to balance preventive and curative interventions when both are available to combat a given disease. Prevention is always preferable so far as pain, suffering, dis- ability, and anxiety are concerned--"an ounce of prevention is worth a pound of cure" in those terms--but it does not follow that it is always the better choice once costs are taken into account. Assuming that the same final health outcome can be attained by prevention or by treatment, the marginal costs of the two alternatives determine the cost-minimizing mix- ture of activities. Under these circumstances--although not in general--the solution that minimizes costs also maximizes cost-effectiveness. Some very simple microeconomics leads to the conclusion that there is no general superiority to either prevention or treatment so far as efficiency is con- cerned. This idea is illustrated in detail in chapter 13, about combating malaria in the Amazon Basin of Brazil. A change in strategy developed by the World Health Organization, to focus preventive efforts more sharply and to give more emphasis to case treatment, greatly reduced the average cost of preventing a death. Preventive efforts, it turns out, can be extremely wasteful. Treatment at least has the advantage of not being applied to peo- ple who do not need it because they are not sick or hurt. Chapter 11 considers a somewhat similar question--whether it would be economically justified to spend the considerable resources needed to eradicate polio in the Western Hemisphere, as the Pan American Health Organization concluded in 1985 was technically feasible. In this case the choice was not between prevention and treatment, but between a combination of immuniza- tion and treatment, with both continuing into the foreseeable future, and elimination of the disease so that treatment could stop. The answer from the empirical analysis was that eradication would be not only ethically preferable but actually cheaper than continuing to vaccinate some but not all the sus- ceptible children and to treat those who contracted the disease. At that time, the choice was easy, and polio actually was eliminated in the Americas. The same logic--that elimination is cheaper as well as better than continuing to permit cases of the disease--has been followed in the near-elimination of 10 · Health Economics in Development polio in the rest of the world. However, it is no longer clear that eradication of the disease is possible, and there are now seen to be substantial risks of ceas- ing to immunize once the disease is eliminated everywhere (World Health Organization 2002). To choose the best course of action--how much to vac- cinate, and with which of the two available vaccines--would now require a more thorough analysis than was necessary in the 1980s, and would have to take account of the uncertainties concerning the various risks of continued circulation of poliovirus and of future outbreaks in unprotected populations. Both chapter 11 and chapter 13 illustrate ideal cases, in which it is possi- ble both to improve health and to save money by a proper choice of strategy. Economic analysis can be decisive in such cases, without the need for ques- tionable assumptions. Both chapters also deal with cases in which costs, or costs relative to effectiveness, can be analyzed without making comparisons across diseases. More often, the question is whether an additional health gain is worth the additional cost involved, relative to other ways of improving health or to nonhealth uses of the resources. In such cases, economic analy- sis is still helpful but depends more on assumptions and value judgments. For polio, an effective vaccine existed and costs could be estimated rather well. When the vaccines needed to control diseases do not yet exist and the resources required to create and apply them cannot be known with any cer- tainty, the question becomes, how large would the benefits need to be to justify the likely costs? Alternatively, how cheap would vaccination have to be for a program still to achieve benefits in excess of costs? This is an exam- ple of cost-benefit analysis, as distinct from the simpler cost-effectiveness analysis used in the study of malaria control. The benefits in this case could include the reduction in pain, suffering, disability, and death, rather than just the reduction in treatment costs that by itself justified the eradication of polio. Chapter 12 deals with these questions for a hypothetical immu- nization program considered by the Pan American Health Organization for the control of pneumonia, meningitis, and typhoid fever, including the development of vaccines. The analysis did not actually estimate the bene- fits: in fact, none of these papers presents a full cost-benefit analysis, because simpler analyses sufficed for the particular questions considered. This avoided the need to make assumptions about unobservable variables-- the values or utilities different people attribute to different kinds of benefits. The use of cost-effectiveness. Cost-effectiveness analysis is the common thread in chapters 8, 9, and 10, as well as chapter 13. Both chapter 8 and chapter 10 derive from the World Bank's World Development Report 1993: Investing in Introduction · 11 Health (World Bank 1993), which used that analysis as a basis for recom- mending "essential" packages of health interventions to be given priority in low- and middle-income countries. More particularly, both these chapters derive from criticisms of the approach taken in that report. In the case of chapter 8, the criticism concerned how the "burden of disease" was calcu- lated and used to help determine priorities (Paalman et al. 1998). In the case of chapter 10, the criticism, voiced by many World Bank economists during preparation of the Report, was that cost-effectiveness is a wrong, or even an irrelevant, criterion because it does not correspond to maximization of util- ity or welfare. Arguments continue over both these issues, especially the lat- ter. Economists are trained to look for ways to maximize or at least improve welfare, which depends on many things besides health, while health spe- cialists focus on health gains, often to the exclusion of any other considera- tion. Economists also differ among themselves on this issue. Some develop and employ cost-effectiveness analysis as a good approximation to cost- benefit analysis based on total welfare (Drummond et al. 1997; Gold et al. 1996), while others reject that approach or are uncomfortable with it. As chapter 8 shows, some of the criticism of how burden of disease esti- mates were created and used results from simple misunderstanding. It remains true, however, that such estimates incorporate a number of unveri- fiable and subjective parameters, so that there is no right answer. Empirical applications of the method in different countries are often not comparable because of different choices about those parameters (Bobadilla 1998). In this respect such calculations violate the recommendation of chapter 6 that the number of arbitrary or subjective parameters should be held to a minimum, preferably to just one. It does not seem possible, however, to follow that rule once the health effects of interest take account of nonfatal disabilities and are therefore more complicated than just additional years of life. Progress in cre- ating broader measures of health status or of the gains from health interven- tions therefore brings with it the necessity for assumptions that, ideally, reflect consensus about the damage from morbidity and disability. The argument in chapter 10 was initially developed to justify the cost- effectiveness analysis in the World Development Report by starting from the economic theory of an individual consumer and moving toward a societal perspective. It begins by accepting that the health intervention an individ- ual would choose to buy for him- or herself, in order to maximize utility, would not necessarily be the most cost-effective intervention against a par- ticular disease or condition that he or she actually had or was exposed to. Still less would an individual care about the relative cost-effectiveness of 12 · Health Economics in Development interventions against other diseases or conditions he or she did not suffer or face a risk from. Nonetheless, when decisions about health interventions are socialized, whether through insurance or, even more clearly, through pub- lic finance, cost-effectiveness becomes relevant, as the total potential health gain for a group of people is considered relative to the total resources for purchasing the corresponding interventions. Rather than a right-or-wrong dichotomy, it appears that cost-effectiveness ranges from being largely irrel- evant at the individual level--although both costs and effects matter, choices are not necessarily based on the ratio of the two--to being increas- ingly useful as a criterion for setting priorities as decisions are socialized and concentrated at a collective level. This argument is more acceptable as one ignores distributional consid- erations, so that it does not matter which individuals gain additional or bet- ter years of life, and only totals or averages are considered. It is also more persuasive, as the resources necessary to achieve those gains can be trans- ferred costlessly from one individual to another who can gain more from their use. It is therefore more acceptable, as there is a single agency making the allocation decisions and responsible to all the potential beneficiaries--a situation approximated by the health systems of some rich countries but not at all typical in poor countries. Somewhat paradoxically, then, cost- effectiveness is on firmest ground theoretically in high-income countries where most health care is financed collectively, while the potential gains from emphasizing cost-effective interventions are greatest in poor countries where the bulk of care is paid for out of pocket. This line of argument, even if fully accepted, certainly does not mean that the cost-effectiveness of different interventions is the only criterion needed for choosing which of them to deliver or finance, as some public health pro- fessionals, eager to have the apparent support of economists, sometimes assume. Cost-effectiveness is one issue among several, even when the deci- sion concerns only the use of public funds and not what individuals choose to spend out of pocket or on voluntary purchases of health insurance. Chap- ter 9 specifies no fewer than eight other criteria for public health spending, and shows how several of them are related to cost-effectiveness, sometimes being compatible with it and sometimes at least potentially in conflict. In particular, cost-effectiveness can easily be incompatible with both horizontal and vertical equity. For some interventions, especially those that are public goods, cost-effectiveness is an adequate criterion, whereas for those services that are private and not catastrophically expensive, cost-effectiveness may not affect priorities at all. The connections among such concepts as public Introduction · 13 goods, externalities, equity, poverty, catastrophic cost, and the proper role of insurance elaborate on some of the arguments in chapter 2. The logic of chapter 9 has been accepted and used in several other instances, including an analysis of the appropriate public sector role in mental health interventions (Beeharry et al. 2002; World Health Organization 2000). Nonetheless, the paper has also fueled continuing controversy about the legitimacy of cost- effectiveness analysis and in particular about whether cost-benefit analysis should always be used instead (Jack 2000; Musgrove 2000). Beliefs and evidence about malnutrition. There was a widespread suspicion in Brazil in the 1970s and early 1980s that poor consumers were paying more for basic foodstuffs than better-off purchasers, and therefore needed a subsidy to prevent their being even worse off than they would be just because of low incomes. While this may have been true formerly, by the time of the study reported in chapter 14, food markets in the northeast of the country, where poverty is concentrated, were sufficiently competitive that the poor did not systematically pay more for their food. It also turned out that a government food subsidy program, based in part on the assumption of prices being higher for the poor, was not very successful at transferring the full amount of the subsidy to the intended beneficiaries. Policies and programs that are based on incorrect beliefs are less effective than they were intended to be. Chapter 15 goes beyond the analysis of costs and prices and into cost- effectiveness, to judge whether four different food and nutrition programs in Brazil made any difference to the growth of young children who were the intended beneficiaries. The analysis is based on a number of detailed evalu- ations of one or more of the programs, conducted by Brazilian researchers. Comparisons among the programs emphasize the variability of outcomes-- some children benefited and some did not, and many variables affected the result--and make it clear that malnutrition is due not only to poverty but also to illness and ignorance, so that programs need to address those causes as well. As in the case of tobacco, overly simple assumptions about con- sumer rationality do not hold up well; it is evident that the people who are supposed to benefit do not always behave as program designers expect. Arguments over which foods people would or should eat, and over whether they needed any assistance other than lower prices to improve their diets and their health, were sometimes quite ideological and based on little evi- dence (Castro 1985; Musgrove 1986). Despite these limitations, there were many Brazilian studies that actually measured outcomes. This situation is, alas, quite unusual: In 18 other Latin 14 · Health Economics in Development American countries, there were many food and nutrition programs in the 1980s but almost no evaluations of results (Musgrove 1993). The available data counted heads (of beneficiaries, or of malnourished children), money spent, and calories or protein distributed, but not who consumed the food or whether it made a difference to physical growth or other aspects of health. The amount being spent on the programs might have sufficed to eliminate most malnutrition, if it were better used. However, almost noth- ing can be concluded about which programs worked best, or how much malnutrition was being prevented or cured Chapter 16 draws on cost-effectiveness analysis to justify fortification of basic foodstuffs with the three best-studied micronutrients: iron, iodine, and vitamin A. Relatively straightforward economic analysis does not pro- vide any argument against the use of fortification to reach all of the popu- lation that suffers, or is at risk of, micronutrient deficiencies and that consumes purchased foods. (Households or communities that raise most of their own food are of course harder to reach in this way.) In particular, it is hard to see any reason to consider fortification less natural or less sustain- able than other approaches to increasing micronutrient intake, such as pro- moting household gardens or otherwise changing what people eat. Taken together, the chapters on actual and potential food policies and programs tell a frustrating story--what is actually accomplished in improving child health seems to fall far short of the potential that has been well established by research (Allen and Gillespie 2001). The research reported in chapters 14 through 17 mostly occurred in the late 1970s and early 1980s, which may give the impression that these matters are of no more than historical interest. However, recent increases in knowl- edge and changes in consumer behavior contribute to keeping some of these issues very much alive. For one thing, it appears increasingly that much con- ventional wisdom about diet, often codified into official advice, is seriously wrong (Willett and Stampfer 2003). Fat is not necessarily so bad for one's health, nor carbohydrates always so beneficial, as has been assumed; and it even appears that alcohol, while quite dangerous in excess, is so protective against heart disease that many adults would be better off drinking moder- ately than not at all (Klatsky 2003). And so far as growth is concerned, scarcity of protein is apparently more of a limitation than inadequacy of calories. Chapter 17 uses information from two surveys in China and a comparison among 41 populations for males and 33 populations for females, to show that availability of protein in the diet is strongly associated with adult heights, and sometimes with adult weights, while caloric availability is much less closely related to height or even to weight. Since growth in height is usually Introduction · 15 completed in adolescence, and seems to be strongly related to experience in the first few years of life; and since the data are aggregates rather than refer- ring to individuals, it is "astounding" (Wray 2003) to find a clear association with current average food intake. This association is always found for males but not always for females; women do not, in many cultures, share equally in what protein is available, and when more protein is consumed on average, it appears that men account for most of the extra consumption. It is a matter of continuing controversy whether the result is due to protein alone or whether various micronutrients, found particularly in foods of animal origin that pro- vide the highest quality protein also affect attained height. In any case, the composition of the diet, and not simply its energy content, appears crucial. What is not in doubt is that underweight, short stature and micronutri- ent deficiencies together account for a large share, probably at least one- third, of the burden of disease in poor countries, a burden concentrated among young children but with effects on physical and cognitive develop- ment that can last a lifetime (Mason, Musgrove and Habicht 2003). At the same time, the definition of "malnutrition" is broadening beyond these tra- ditional measures of deficiency, to include anthropometric states and health conditions associated with excess. In rich countries such as the United States, and even in middle-income Brazil, overweight children now outnumber those who are underweight, and obesity among adults is epi- demic. Meanwhile, programs aimed at feeding poor children continue to operate on the assumption that they eat too little rather than too much (Besharov 2002). It is also clear now that markets can change people's eat- ing habits thoroughly and rapidly, in contrast to the assumption in chapters 15 and 16 that such changes would be sluggish. Governments may need to play a major role in changing habits created or accelerated by the market, and not just traditional eating patterns, to ensure good nutrition. Paying for Health Paying for or "financing" health care can be thought of as consisting of three subfunctions: funding, or collecting revenues; pooling them so as to share risks among individuals and households; and using them for purchasing health- related goods and services (World Health Organization 2000, chapter 5). This distinction collapses in the case of out-of-pocket payments, because they are not pooled across households. The act of purchasing coincides with and determines, ex post, the corresponding amount of funding. All financing that is not out of pocket constitutes some form of prepayment, which includes 16 · Health Economics in Development all three subfunctions. Chapters 18 through 20 deal with only the first two subfunctions: funding and pooling. Purchasing involves decisions on what goods or services to buy, whom to buy them from, and how to pay the providers. Some of the issues involved are treated in chapter 9. Funding, pooling, and purchasing are often described as if they followed a temporal as well as a logical sequence--funds are first collected, then allotted to one or more pools, and finally spent from those pools under various pur- chasing arrangements. This is a useful way to think of the flow of funds, which have to be collected somehow in order to go into a pool, but it does not always describe the decisions about those funds. For general revenue or "tax- based" financing, which can be considered implicit insurance because risks are shared among beneficiaries but there is no defined premium and there may be no defined benefits, a government first collects taxes and then determines what share of revenues to allocate to health. For any kind of explicit insurance, whether private or quasi-public (such as social security contributions), the causal relation is actually the other way around. The definition of the pool determines the nature of the funding and the way it is collected. Funds are collected--voluntarily or involuntarily--only because the pool exists, and the volume of funds is the product of the premium or mandatory contribution and the number of contributing members (or the sum of such products, if dif- ferent insured members pay different premiums or contributions). Decisions that governments take affecting explicit insurance are in the first instance decisions about pooling, with funding as the consequence. The argument for government to be involved in health financing turns on two facts. First, health needs are often unpredictable and can imply cat- astrophically high costs for individuals and households: hence the need for insurance rather than relying solely on payment out of pocket at the moment of need. Second, health insurance does not work like insurance for nonhuman assets, because the need for financial protection bears no rela- tion to anyone's capacity to pay. Hence the need not only for subsidies or transfers from the healthy to the sick, parallel to transfers from the lucky to the unlucky in other forms of insurance, but also from the better-off to the worse-off. Pooling of funds is crucial, since it allows for sharing risks and therefore for the first kind of subsidy. Whether it also promotes the second kind depends on who belongs to which pool(s), how much revenue goes into the pool(s), and how the contributions to that revenue are distributed among the insured. Government decisions about health financing therefore typically are joint decisions affecting the sources of revenue and the num- ber, size, and composition of pools and the relations among them. Introduction · 17 Figure I.1 Decision Flowchart, Sources of Health Financing Prepayment Out of Pocket Implicit Explicit (via taxes) (via Insurance) General Health Quasi-Public Private Revenue Specific (via Social Insurance) External Assistance Other "Sin Taxes" Job-Related Voluntary Allocation of Allocation of Public Budgets Household Budgets Figure I.1 illustrates some of the major choices about funding, starting with the fundamental distinction between prepayment and out-of-pocket spending, and continuing through different kinds of taxes and insurance arrangements to consider how these choices affect both government budgets and those of households (Musgrove 2001). At each stage in the figure, there are choices to be made about using particular taxes or social security schemes and promoting or inhibiting one or more forms of pri- vate insurance. Several of the issues related to these choices are not spe- cific to health, since they concern the structure of a tax system and its relation to private sources of finance. What looks at first glance like a good way to raise resources for health may be quite ineffectual, or inequitable, when the entire financing system is considered. There is considerable naïveté on this point in the typical Ministry of Health, so it is worthwhile to create a better understanding of financing issues generally, rather than concentrating only on the size of the public budget or on particular taxes or other funds that are dedicated to health. Here, economic analysis can make a contribution that is independent of the specific epidemiological or medical issues considered in most other chapters. Partly for that reason, it may be easier to reach understanding between economists, whose specialty this is, and public health professionals. 18 · Health Economics in Development Who pays for health care. Chapter 18 examines the pattern of household spending on health in Latin America three decades ago, using purely eco- nomic data that said nothing about health conditions or needs. The funda- mental characteristic of the health systems generating the observed expendi- tures is that public and private health care facilities competed mostly on price, not on quality. As a result, consumers bought private care, which was more expensive but--at least in their opinion--better than subsidized public care as soon as they could afford it, leading to a very high elasticity of expenditure with respect to income. This effect was far from uniform, however; spending on drugs tended to rise quickly and then level off, whereas private hospital costs continued rising with income. This situation reflects the prevalence of self-medication among the poor and the fact that people often have to pay for drugs even if consultations are free or covered by insurance. Treating one input to medical care differently from the others makes little economic sense and merely exploits people's often desperate willingness to pay for medicines. (If what a patient needs are 15 minutes with a nurse or doctor and a bottle of pills, there is likely to be little gain from 30 minutes and no pills, or from no professional diagnosis or advice and two bottles of pills.) There is now much more information available about families' health spending, from household surveys that also ask about episodes of illness and care-seeking behavior; the analysis in chapter 4 is based partly on an early example of such surveys. However, some of the most undesirable features of health financing in poor countries continue largely unchanged. Except for inexpensive care that presents no serious financial burden to the consumer, out-of-pocket spending is without doubt the worst way to pay for health services. Almost any form of prepayment is preferable. Chap- ter 19 looks at how health was actually financed in the mid- to late 1990s, using national health accounts estimates or less reliable approximations for all 191 member states of the World Health Organization. The strongest and most disturbing conclusion from that analysis is that out-of-pocket spend- ing is relatively highest in precisely the poorest countries where people most need financial protection. Public spending is often only a small share of the total, and private prepayment is nearly nonexistent. Most spending directly by families is out of pocket, and the burden often falls chiefly on the poor and on rural dwellers who have little access to publicly subsidized care. Impoverishment for health reasons--because people cannot afford care and families lose their livelihoods to death, illness, or injury, or because they become poor paying for care--is a serious risk and a major impediment to poor people's acquisition of capital, whether human or material. Introduction · 19 The two great failings of health finance. In static or cross-sectional terms, then, the way health is financed in poor countries is far from ideal. Too much spending is out of pocket, insurance of whatever sort is the privilege of the better off, and public health spending is often financed by dedicated taxes that may be unstable or inequitable and that in any case may not affect total expenditure because other sources are withdrawn in proportion. What is true in comparisons among countries is often true also when comparing how the health care of different population groups is financed. Those who least need prepayment, because they could afford all but catastrophic care, are the best protected, and sometimes also the best covered by public spending. Dynamically, the situation is often even worse. In an economic crisis or even a downturn in incomes, the needs for health care are likely to rise just when people's capacity to pay for it declines because of lost income or employment and their financial protection weakens or disappears as they lose insurance or social security coverage. When that happens, only public expen- diture can fill the breach--but it is usually cut back also, as public revenues fall. There is a strong argument for public spending on health to be counter- cyclical, an argument first made in 1984 (Musgrove 1984). Chapter 20 develops that argument and supports it with Latin American data on health expenditure and economic variables. A counter-cyclical health policy would of course require that planning for health expenditure be contingent on eco- nomic circumstances, with explicit decisions to protect health in downturns and to let private or social security spending meet more of the need in boom times. Such a policy would be compensatory through time, as the allocation policies discussed in chapter 6 are compensatory across space. That in turn would require long-term political agreements, including decisions as to which kinds of public spending to cut back when income and revenue fall. No government seems to have achieved such agreements--the political obstacles are daunting--and the response to economic crisis is often improvised. In these as in many other respects, there is still great room for improvement in health financing policies to promote both equity and efficiency. Concluding Reflections Economic analysis can lead to or support sound decisions to improve health: chapters 11 and 13, on polio eradication and malaria control, are clear illustrations of the potential benefits. Eventually the economic analy- sis of tobacco control in chapter 3 may prove of comparable value by 20 · Health Economics in Development overcoming some of the conceptual barriers to such control. And it may be hoped that the findings and arguments in chapters 14 through 17 will lead to more effective efforts to combat the huge problems of different kinds of malnutrition in poor countries. The contribution of health economics to development is potentially even broader than such instances suggest. Eco- nomic analysis, properly applied, can often help to clarify what the choices are for health policy, how to choose among different criteria, how to decide what to buy and how to pay for it, and how to evaluate the results. The indi- rect effects of good economic thinking, when dealing with such questions as the best use of taxes, insurance, and out-of-pocket payment, or the best way for governments to intervene in health, may affect a population's health and welfare more than decisions about how to combat particular maladies or risk factors. If it is true that "the ideas of economists and political philosophers, both when they are right and when they are wrong, are more powerful than is commonly understood" (Keynes 1965), then it is impor- tant to get those ideas right. This is especially so when the issues are increasingly complex and when ideology and misunderstanding, although on the decline, are still widespread. Getting the ideas right is not easy, for those same reasons. The impor- tance and peculiarities of health insurance, and the reasons why competi- tive private markets for it are likely to produce both inequity and ineffi- ciency (Arrow 1963) are now widely understood. How to deal with these market failures without falling into equally dangerous government failures is a question still only partly settled. And while there is far more under- standing and even rapprochement among economists and public health specialists concerning the state and the market than there was a few decades ago, the right balance of roles and the right choice of instruments continue to be debated, sometimes acrimoniously. Similarly, there is today much more understanding than formerly of how to relate the costs of health interventions to the various kinds of benefits they can yield, includ- ing their effectiveness in improving health. But controversy continues over the legitimacy and applicability of different methods, particularly when comparisons need to be made over different diseases or conditions, people with different health or economic prospects or cultural views, or completely different kinds of beneficial or harmful outcomes. Several of the chapters here try to develop better answers to these questions, or to provide persuasive counter-arguments to some wrong answers and mis- taken views. The reader may judge how well this has been accomplished, and perhaps learn a little more about both the meanings of health eco- nomics in development. Introduction · 21 References Allen, Lindsay, and Stuart Gillespie. 2001. What Works? A Review of the Efficacy and Effectiveness of Nutrition Interventions. Manila: United Nations Administrative Committee on Coordination, Sub-Committee on Nutrition, and Asian Development Bank. Arrow, Kenneth. 1963. "Uncertainty and the Welfare Economics of Medical Care." American Economic Review 53: 941­73. Beeharry, Girindre, Harvey Whiteford, David Chambers, and Florence Baingana. 2002. "Outlining the Scope for Public Involvement in Mental Health." World Health Organization working paper. Geneva: World Health Organization. Besharov, Douglas J. 2002. "We're Feeding the Poor as If They're Starving." The Washington Post, December 8, p. B-1. Bobadilla, José Luis. 1998. Searching for Essential Health Services in Low- and Middle- Income Countries. Policy Background Study No. SOC-106. Washington, D.C.: Inter-American Development Bank. Castro, Cláudio M. 1985. "Fubá, Formulados e Fundamentalistas." In Cláudio M. Castro and Marcos Coimbra, eds., O Problema Alimentar no Brasil. São Paulo: Almed Editora. Castro, Cláudio M., and Philip Musgrove. 2000. "On the Non-Existence of `The Social Sector,' or Why Health and Education Are More Different than Alike." Working Paper. Washington, D.C.: Inter-American Development Bank. Creese, Andrew. 1991. User charges for health care: a review of recent experience. Health Policy and Planning (6)4: 309­319. Drummond, Michael, Bernie O'Brien, Greg L. Stoddart, and George W. Torrance. 1997. Methods for the Economic Evaluation of Health Care Programmes. 2nd edi- tion. Oxford: Oxford University Press. Gold, Marthe R., Joanna E. Siegel, Louise B. Russell, and Milton C. Weinstein, eds. 1996. Cost-Effectiveness in Health and Medicine. New York and Oxford: Oxford University Press. Hirschleifer, Jack. 1985. "The Expanding Domain of Economics." American Economic Review, 75, Supplement: 53­68. Hsiao, William C. 2000. "What Should Macroeconomists Know about Health Care Policy? A Primer." Working Paper. Washington, D.C.: International Monetary Fund. Jack, William. 2000. "Public Spending on Health Care: How Are Different Criteria Related ? A Second Opinion." Health Policy 54: 3. Keynes, John Maynard. 1965. The General Theory of Employment, Interest and Money. First Harbinger Edition. New York: Harcourt, Brace and World, p. 383. Klatsky, Arthur J. 2003. "Drink to Your Health?" Scientific American 288(2), February: 75­81. Lohr, K. N., R. H. Brook, C. J. Kamberg, G. A. Goldberg, A. Leibowitz, J. Keesey, D. Reboussin, and J. P. Newhouse. 1986. "Use of medical care in the RAND 22 · Health Economics in Development health insurance experiment: diagnosis- and service-specific analyses in a ran- domized control trial." Medical Care 24 (supplement), 51­587. Mason, John B., Philip Musgrove, and Jean-Pierre Habicht. 2003. "At Least One- third of Poor Countries' Disease Burden Is Due to Malnutrition." Disease Control Priorities Project Working Paper No. 1. Bethesda, Md.: Fogarty International Center, National Institutes of Health. Miller, Jonathan. 1978. The Body in Question. New York: Random House. Musgrove, Philip. 1984. "Health Care and Economic Hardship." World Health, October: 27­29. ------. 1986. "Ideología, Pesquisa y Realidad de la Situación Alimentaria y Nutricional del Brasil." Cadernos de Estudos Sociais (Recife, Brazil) 1(3), January­June: 329­348. ------. 2000. "Cost-Effectiveness as a Criterion for Public Spending on Health: A Reply to William Jack's `Second Opinion'." Health Policy 54: 229­233. ------. 2001. "Choices in Health Financing." Draft paper for the World Health Organization. Geneva. Newbrander, William, David Collins, and Lucy Gilson. 2000. Ensuring Equal Access to Health Services: User Fee Systems and the Poor. Boston: Management Sciences for Health. Paalman, Maria, Henk Bekedam, Laura Hawken, and David Nyheim. 1998. "A Critical Review of Priority Setting in the Health Sector: The Methodology of the 1993 World Development Report." Health Policy and Planning 13(1): 13­31. Ruger, Jennifer P., Dean T. Jamison, and David E. Bloom. 2001. "Health and the Economy." In Michael H. Merson, Robert E. Black, and Anne J. Mills, eds., International Public Health: Diseases, Programs, Systems, and Policies. Gaithersburg, Md.: Aspen. Willett, Walter J., and Meir J. Stampfer. 2003. "Rebuilding the Food Pyramid." Scientific American 288(1), January: 64­71. World Bank. 1993. World Development Report 1993: Investing in Health. Washington, D.C. World Health Organization. 2002. Informal Consultation on Economics Research on Post-Certification Polio Immunization Policies. October 7­8, Geneva. ------. 2000. World Health Report 2000--Health Systems: Improving Performance. Geneva. Wray, Joe, 2003. "Response to `Malnutrition and dietary protein: Evidence from China and from international comparisons': Commentary." Food and Nutrition Bulletin 24: 291­5. CHAPTER 1 What Is the Minimum a Doctor Should Know about Health Economics? Acknowledgments I must thank Bertoldo Kruse for inviting these reflections. Many colleagues and friends have contributed to these ideas, particularly those associated with the World Bank's Flagship Course in Health Reform. None of them bears any responsibility for what is said here. Why Do Doctors Need to Know Any Health Economics? The answer to this question is not obvious: after all, when a physician is actually practicing medicine there seems to be no room or need for eco- nomic understanding. In fact, it might get in the way, when what the doc- tor wants is to concentrate on the patient before him or her and bring to bear all his or her medical knowledge, which is typically much more detailed--and certainly more important at the moment of diagnosis or treatment--than what an economist typically knows or thinks about. And doctors have been treating patients, well or badly, for centuries without troubling themselves with economic concerns. Economics perhaps has no place in the surgery, the consulting room or the laboratory, but that is not what matters. In each of those settings, resources are being used and a production process is under way, supposedly for the benefit of a consumer--and the use of limited resources to produce Reprinted, with permission, from Revista Brasileira de Saúde Materno-Infantil 1(2), May­August 2001, (Recife, Brazil). 23 24 · Health Economics in Development goods and services for intermediate or ultimate consumers is what econom- ics is primarily about. How those resources are themselves produced, how they are combined, who chooses what to produce with them, who will pay for them, and what all that costs, create the setting in which the physician operates. Almost everything that happens prior to the encounter between the physician and the patient is relevant to the economist, even if the latter is kept outside of the medical practice itself. If there is something the doc- tor ought to know of health economics, it concerns those prior steps, including many of the factors that bring the patient to his or her attention in the first place. There are at least three reasons why a physician might disregard this argument and suppose that economics has nothing useful to offer his or her profession. One is the fact that health economics is a relatively new sub- discipline. The seminal article explaining some of the subtleties that distin- guish health from other sectors, particularly in relation to how it is financed, was published only in 1963 [1]. That opened the whole field of inquiry into risks and information that characterize health economics today and that has become steadily more important as more and more of health care is financed by insurance and the costs of it have risen. The Journal of Health Economics, the first publication devoted entirely to the subject, began to publish only in 1982; by now there is an entire two-volume Handbook of Health Economics [2] and a number of journals that publish on the subject. Economists are quick to "invade" fields they find interesting, and the prac- titioners of those subjects may take time to notice that they have become of economic interest. A second reason is the mistaken supposition that economics is nothing more than accounting, and while accounts must be kept in medical practice as in other professions, the logic of the accounting is no different and the accountant has no special insights to offer. Much of economics does in fact depend on proper accounting: the creation of national accounts of income and product, starting more than a half-century ago, is the precursor of today's effort to create national health accounts [3] to show where the funds spent on health come from and where they go. But the interpretation of those flows does not follow only from their magnitude, but from economic theory about how doctors, patients, and financing agencies behave. A third, even more mistaken reason, is summarized in the attitude that "health is not a business", or should not be one. Some doctors, and public health professionals in particular, often find it hard to accept that health care is financed, produced and delivered in a constellation of markets--as What Is the Minimum a Doctor Should Know · 25 though markets or "business" were intrinsically inimical to human health. This argument usually rests on the claim that health care is a basic right or a basic need, and therefore too important to be left to markets. But food, which is a much more basic necessity than health care, is produced and delivered in markets, and there is nothing wrong with that. The question, in the case of health care, is whether those markets work in socially desir- able ways, or whether they lead to situations in which some people cannot afford needed care, or the wrong kinds of care are produced, or at too high a cost, or something else goes wrong. Economics is, to a large extent, the science of how markets operate, so it is extremely relevant to markets in which failure may be a matter of life and death. What Economics Does a Doctor Not Need to Know? So it might be helpful for medical professionals to understand some eco- nomics, as it applies to medicine and health. Does that mean they need to comprehend all of economics, or would it be safe to ignore large areas of the subject? Fortunately, there is much that a doctor does not need to know, starting with the specific economic issues that arise in sectors very different from health. The frequent (and frequently loose) use of the adjective "social" to describe some sectors of the economy might suggest that med- ical professionals wanting to understand health economics need to know something about the economics of related sectors such as education. Fortu- nately, this is not the case: in economic terms these fields are much more different than they are alike [4], and although similar issues arise in both [5], it is more confusing than helpful to think of a general economics of "the social sector". The peculiarities of health economics mean that a doctor wanting to learn something about it need not try to understand the eco- nomics of any other sector in detail. Currently there is great interest in what might be called "the macroeco- nomics of health", and a Commission on Macroeconomics and Health [6] has been created to study particularly the question of whether better popu- lation health contributes to economic growth, making health even more of a paying investment than it has traditionally been considered from an indi- vidual's perspective [7]. That is an interesting question, but health does not need to be subordinated to income or growth in order to be regarded as vitally important. (It is even dangerous to justify health investments by appealing to their effect on economic outcomes, since such investments 26 · Health Economics in Development may pay off best for young adults and thereby lead to discrimination against the very young and the very old.) And even if there is a strong connection, it does not mean that medical professionals need to know anything about macroeconomic theory in order to learn something useful for their own field. All that a doctor should know is that there are good macroeconomic policies and bad ones, that inflationary populism is a very bad policy and that poor macroeconomic management is bad for a country's health, partic- ularly the health of poor people. Much of the criticism directed at "struc- tural adjustment" and its supposed damaging effects on health really should be directed at the economic irresponsibility that sometimes made such adjustment necessary in the first place [8]. Of course, what economists think they know is often a mixture of what they know and what they only think, including their more ideological posi- tions and beliefs. (The same is true of public health specialists, to be sure.) One reason that doctors are reluctant to learn more economics is that they reject some views as ideological--sometimes with good reason, sometimes mistakenly. For example, the claim by economists that most of the time, markets are an efficient mechanism for allocating resources to production and consumption may sound like ideology, but it is actually a strong empir- ical proposition. The history of efforts to control prices, dictate production or otherwise interfere with the normal working of markets, including par- ticularly the sad history of Soviet-style economic management, offers abun- dant evidence. However, the claim by some economists that all markets are basically alike, and that in particular markets work just as well in health care as anywhere else, is not well supported either theoretically or empirically but includes a large dose of ideology. One needs to understand how markets work, without being taken in by "the mystique of markets" [9]. Doctors who have never talked much with economists--or who have had the misfortune to talk only to mediocre economists--often think that econ- omists care only about efficiency and not at all about equity, equality, rights, or the suffering of the sick and the dispossessed. It is certainly true that in economic theory, it is easier to agree on what constitutes or leads to effi- ciency than to agree about equity; and it is also true that inefficiency means waste, which means less of something desirable for someone. But economic thought also includes a long and deep tradition of thinking about ethical issues, about what constitutes a just society, about rights and entitlements [10] and about the possible conflicts between equity and efficiency and the frequent necessity for choices among societal objectives [11]. Even for such a relatively narrow question as what health interventions to purchase with What Is the Minimum a Doctor Should Know · 27 public money, there are no fewer than nine relevant criteria, of which at least three concern equity rather than efficiency [12]. The conclusion to draw from all this is that a doctor wanting or needing to learn some eco- nomics does not have to abandon his or her ethical principles or political views. What he or she should be prepared to do is to question those princi- ples and views in the light of economics and see how well they hold up. Eco- nomic thinking can help to identify contradictions or poorly formulated opinions. It does not impose a set of ethical or political suppositions or pref- erences. (In fact, the economics of consumer behavior starts with an unquestioned respect for preferences.) Understanding How Economists Think More than knowing any particular conclusion of economics, a doctor needs to understand the way that economists think: incomprehension and conflict arise more from differences in the way the two professions approach questions, than from the specific answers to those questions. An economist does not, contrary to popular superstition, think only or pri- marily about money, even if he or she often tries to find monetary equiv- alents of other measures. Economists think about resources, and particu- larly about whether those resources have prices and if so, whether they are the right prices to assure efficiency or equity. Since resources have costs, whether those are recognized or not, economists want to know if the use of those resources produces effects (non-monetary) or benefits (usually monetized) sufficient to justify how they are used. Much work in econom- ics is devoted to comparisons among these concepts, under the names of cost-effectiveness analysis, cost-utility analysis or cost-benefit analysis [13]. It is important for doctors to understand that while costs are the specialty of economists, the definition and estimation of effects or outcomes is the province of medical professionals: these analyses have to be joint efforts. Given an estimate of an effect (deaths averted, for example), economists often then go on to try to put a monetary value on the result, and such efforts can be questioned and rejected. What a doctor needs to understand is that while any particular kind of effect can be related to costs without monetizing the effect, there is no common currency besides money in which to compare different kinds of effects (health outcomes versus edu- cation, say), and that to avoid monetary valuations is to abstain from all such cross-sectoral comparisons. 28 · Health Economics in Development As mentioned earlier, economists naturally think about markets, ideally without any prior assumptions about how well they work. To reject the idea of markets because some market outcomes are inefficient or inequitable or both, is to miss one of the main ideas that economists always carry with them. But markets are not simply theaters in which two characters called "supply" and "demand" interact, important as those two concepts are. Mar- kets are places where people interact, in many different roles, as payers, investors, providers, patients, consumers and citizens; so economists con- centrate on the behavior that occurs in markets, and in particular on the incentives that people face to behave one way or another. It is true that econ- omists tend to talk mostly about financial or economic incentives, because they understand those best. That does not mean that other incentives--the desire to help others, professional pride, and so on--do not matter, only that economic analysis starts by taking those for granted, and then asks what happens to behavior when prices, means of payment, regulations or other incentives are modified. Particularly in the health sector, the economic incentives are often perverse, acting contrary to the desired outcomes, so it is crucial to analyze them and correct them if possible. In considering the incentives and regulations to behave one way or another, economists have to assume that behavior is not simply a collection of responses to random impulses, but that people have some set of goals or objective function, that they are trying to get the most (or the least) of something out of their actions. It makes a difference, sometimes a great dif- ference, what those objectives are. For example, a producer of a good or service will behave differently, depending on whether he or she aims to maximize profits, to maximize revenue, to assure a particular level of income, to capture a particular share of a market, to minimize risk, or to produce the highest possible quality of output. Since objectives are not always stated, and may not even be clearly known to the agent whose behav- ior is of interest, there is necessarily some speculation involved, and the confrontation of different assumptions with observed behavior. In this respect, economics has much more in common with psychology than with accounting or engineering. Incentives, to be effective, have to work on peo- ple's objectives; misunderstanding what they want or are trying to do can lead to perverse incentives and unwanted outcomes. Finally, economists pay much attention to who has, and who needs, how much and what kind of information. People make all kinds of decisions based on the information they have (or think they have), and entire markets can work badly when information is incomplete (no one knows) or asymmetric What Is the Minimum a Doctor Should Know · 29 (buyers know more than sellers, or vice versa), particularly if revealing information would damage the interests of the person who has it. Ignorance is obviously dangerous in the face of an epidemic, or for a person who faces a risk but is unaware of it or does not respond to information about it. Smoking is a marked example of this danger [14, especially chapters 7 and 8]. Some kinds of information lend themselves to accounting and standard- ized reporting (the basis of national health accounts and of much of epi- demiology), but others do not, because they concern only individual actors or are costly to collect or interpret. Medical professionals also recognize the importance of information for detection, diagnosis, treatment and evalua- tion. What economic thinking adds is the emphasis on how information or the lack of it influences behavior, with economically important conse- quences [15]. Important Specifics of Health Economics First, health is a very peculiar asset because unlike almost anything else, including even some other forms of human capital, it is almost entirely inalienable. One can donate blood or even a kidney to improve someone else's health, but "health" itself cannot be transferred, and one must have some state of health, however poor. Since health is subject to many random shocks of illness or accident, and since health care can be catastrophically costly, one needs insurance against financial risk as well as the protection against physical risks provided by good nutrition, exercise and a range of public health measures such as sanitation and immunization. But the char- acter of health makes it harder to insure than other assets, especially since the value of one's health and the financial risk are not correlated with one's capacity to pay. Thus one of the principal obstacles to making a health sys- tem work properly, is the difficulty of financing it so as to provide a reason- able and affordable degree of protection to everyone, without creating incentives either to do without such protection or to over-use medical care because the cost is borne by others--and while assuring that subsidies flow in the desirable directions. This difficulty is independent of the amount spent on health. The emphasis on financing in discussions of health economics is entirely justified, then; but a doctor also needs to understand that there are three parts to it. It matters not only how health is funded, that is, who pays for it and through what mechanisms (taxes, social security, voluntary insurance, 30 · Health Economics in Development charity, out of pocket payments) but also whether and how those funds are pooled to share risks among population groups, and how they are then used to purchase goods and services [16, chapter 5]. Each of these stages presents its own set of questions and difficulties, often with conflicts between eco- nomic efficiency and equity or fairness. One important source of conflict is that what people want in the way of health care does not necessarily match what doctors think people need; and when needs and demands do not coincide, it is impossible for the supply of services simultaneously to satisfy both of them [17, pp. 23­24]. Several of the reasons why need, demand and supply do not automatically match up, go by the name of "market failure", meaning that while there is a working market for health care, it does not reach the kind of efficient equilibrium that a so-called perfect market would achieve. Doctors need to understand these reasons, which include standard economic concepts such as public goods, externalities, information failures, and non-competitive behavior. They also need to distinguish these problems from other reasons for unsat- isfactory health outcomes which are just as important but which are not "failures" in the economist's sense--such as poverty, and inequality of risks or of income. "Failure" is one word to which economists give a fairly exact meaning that may not match the commonsense notion doctors are likely to have, and it is important, as in any dialog, to develop a clear, shared vocabulary. Argu- ing over whose definition to use, or not recognizing that the same word may be used in two meanings, is wasteful: the best example of this is the differ- ence between the economist's term "public good" and the medical sense of "public health". All public goods in health are part of public health, but the converse is not true, and the difference matters for public policy. Purchasing, the last stage of financing health, involves two complex ques- tions: what to buy, and how to pay the providers--doctors, hospitals, ven- dors of goods and services. The difference between need and want, and the enormous variation in costs of medical procedures, are crucial for the first choice, as is the definition of what one is trying to achieve. Maximal overall population health as an objective will lead to different choices than improv- ing the health of the worst-off, or giving everyone something like the same chance to have his or her health problems resolved. And doctors need to understand that while the size of a health problem--for example, the bur- den of disease attributed to a particular disease or condition--is highly rel- evant to how much it might cost to deal with the problem, decisions about what interventions should have priority do not depend simply on the What Is the Minimum a Doctor Should Know · 31 magnitude of the problem [18, 19]. A full evaluation of a health system draws on many different kinds of information [20]; some ways of using or combining different kinds of data are useful or legitimate, and some are not. Incentives are crucially important to the second question, and deciding on the best way to pay providers is greatly complicated by a feature that is peculiar to health care--the practice of referral from one level or type of facility or professional to another. There does not seem to be one ideal way to pay all the different providers involved in a system, so a doctor needs to understand the virtues and deficiencies of different payment systems (fee- for-service, global budget, per bed-day, for diagnostic-related groups, and so on) and how they interact. Aligning incentives and creating a good insti- tutional environment is most important for hospitals, the most complex organizations in the health system [16, chapter 3]. How to pay, as opposed to how much to pay, is an example of the impor- tance of institutions and regulations in health economics: it is not a matter of costs, in the first instance, but it may have a large impact on both costs and health outcomes. More generally, doctors need to know that much of health economics is concerned with the rules of the game and not simply with the flows of money, goods and services. One particular issue of this sort is that of the right degree of autonomy for individual doctors and for the organizations in which they work and the organizations which purchase their services--which are not the same, when there is a "purchaser-provider split" between funding and purchasing agencies, and the producers of med- ical services. Too little autonomy, too much dictation from above or outside, is practically a guarantee of waste; too much freedom may be an invitation to abuse, low quality or excessive costs. As with many other issues, econom- ics does not provide final answers, but it does offer a way of thinking about them that can facilitate better decisions and ultimately better outcomes. Gains From Better Understanding Suppose a medical professional accepts the need to understand some health economics, perhaps including the specific ideas just discussed. What can he or she hope to gain thereby? What is the likely pay-off for the effort involved in learning some new vocabulary, accepting new and different viewpoints, and possibly having to give up or modify some cherished ideas? The most obvious benefit is that it becomes easier to talk with economists, when one cannot avoid doing so--and as decisions about health care come to depend 32 · Health Economics in Development more and more on economic considerations, it becomes harder to keep econ- omists out of the discussion. Doctors sometimes fear being crowded out of decisions over which historically they had full control. Perhaps the best way to assure that their knowledge and views continue to be respected is to learn something about the knowledge and views of the newcomers to the other side of the table. Reducing the level of incomprehension and antagonism that often characterizes such encounters at first, is worth some trouble. Ideally, better mutual understanding between medical professionals and economists will actually improve the efficiency of health care, and maybe even its equity. By examining their own behavior and responses to incen- tives in the light of economics, doctors may see ways to be more effective or less wasteful of resources; and they should be better prepared to accept, and influence, reforms to how they work and how they are paid. There is nothing guaranteed, or easy, about reform processes in health, but they seem sure to work better when all involved have at least some knowledge of all the relevant factors. For society as a whole, a better working health sys- tem is clearly the greatest potential gain from a fuller understanding between the two professions. Finally, at least for some medical professionals there can be a purely sci- entific or intellectual pleasure in exploring the thinking of another profes- sion and thereby seeing one's own profession differently. Of course, this can lead to frustration, because the new ideas may be hard to put into practice and can lead to friction with one's own colleagues. This is especially likely when techniques of economic evaluation are stretched too far or their results conflict too strongly with perceived political imperatives [20, pp. 12­17]. But such stretching and conflict are often a necessary part of learning, and may ultimately be the basis for different political imperatives and reform opportunities. References [1] Kenneth Arrow, "Uncertainty and the Welfare Economics of Medical Care". American Economic Review 53: 941­73, 1963. [2] Anthony J. Culyer and Joseph P. Newhouse, eds., Handbook of Health Economics. Volume 17 of Handbooks in Economics. Amsterdam: Elsevier, 2000. [3] Jean-Pierre Poullier and Patricia Hernández, "Estimates of National Health Accounts (NHA) for 1997", WHO/EIP Discussion Paper 27. Geneva: World Health Organization, 2000. What Is the Minimum a Doctor Should Know · 33 [4] Cláudio de Moura Castro and Philip Musgrove, "On the Nonexistence of `the Social Sector', or why Education and Health are more different than alike". Washington, DC: Inter-American Development Bank discussion paper, 2000. [5] Emmanuel Jiménez, Pricing Policy in the Social Sectors: Cost Recovery for Education and Health in Developing Countries. Baltimore: Johns Hopkins University Press for the World Bank, 1987. [6] World Health Organization, Evidence and Information for Policy, The Commission on Macroeconomics and Health (CMH): Overview. Geneva: World Health Organization, 2000. [7] William D. Savedoff and T. Paul Schultz, eds., Wealth from Health: Linking Social Investments to Earnings in Latin America. Washington, DC: Inter- American Development Bank, 2000. [8] Michel Garenne and Eneas Gakusi, "Health Effects of Structural Adjustment Programs in sub-Saharan Africa". Paris: French Center for Population and Development Studies (CEPED) working paper, 2000. [9] Sara Bennett, The mystique of markets: public and private health care in developing countries. London: London School of Hygiene and Tropical Medicine, Public Health and Policy Department Publication No. 4, 1991. [10] Amartya Sen, Poverty and Famines: an Essay on Entitlement and Deprivation. Oxford: Oxford University Press, 1982 [11] Arthur M. Okun, Equity and Efficiency: the Big Trade-Off. Washington, DC: the Brookings Institution, 1975. [12] Philip Musgrove, "Public Spending on Health Care: How are Different Criteria Related?" Health Policy 47: 207­223, 1999. [13] Michael Drummond, Bernie O'Brien, Greg L. Stoddart and George W. Torrance, Methods for the Economic Evaluation of Health Care Programmes. 2nd edition. Oxford: Oxford University Press, 1997. [14] Prabhat Jha and Frank Chaloupka, eds., Tobacco Control in Developing Countries. Oxford: Oxford University Press, 2000. [15] Victor R. Fuchs, The Health Economy. Cambridge, MA: Harvard University Press, 1986. [16] World Health Organization, World Health Report 2000 ­ Health Systems: Improving Performance. Geneva: WHO, 2000. [17] Philip Musgrove, Public and Private Roles in Health: Theory and Financing Patterns. Washington, DC: World Bank discussion paper No. 339, 1996. [18] Dean T. Jamison, W. Henry Mosley, Anthony R. Measham and José Luis Bobadilla, Disease Control Priorities in Developing Countries. Oxford: Oxford University Press for the World Bank, 1993. [19] World Bank, World Development Report 1993: Investing in Health. Washington, DC: the World Bank, 1993. [20] George E. Cumper, The Evaluation of National Health Systems. Oxford: Oxford University Press (Oxford Medical Publications), 1991. CHAPTER 2 Public and Private Roles in Health Acknowledgments This paper was written under the guidance of David de Ferranti, who also commented extensively on the first draft and made a valuable suggestion about the overall structure. Jacques van der Gaag, George Schieber and David Dunlop read early versions of the paper and provided immensely helpful information, suggestions and encouragement. Howard Barnum, José-Luis Bobadilla, Xavier Coll and Helen Saxenian also offered useful comments. Linda Kean edited the paper to take all these comments into account. The final version benefited from written review by Nicholas Barr, Ralph Harbison, Elizabeth King, Maureen Lewis, Samuel Lieberman, Paul Shaw and Verdon Staines, and verbal comments by Jeffrey Hammer. Any remaining errors or confusions are the author's sole responsibility. Why the Public Role in Health Care Matters Health care in about 1990 cost at least $1.7 trillion, or about 8 percent of world income [Murray, Govindaraj and Musgrove 1994], making it one of the largest industries in the global economy. On average, 60 percent of this is public spending. If this spending is excessive or otherwise inappropriate, the consequences for the economy and for health outcomes could be sub- stantial. Governments also provide a large share of health services, some- times as large as the share in spending, and often intervene in various ways in the private health care market. Since most health care is a private good, it is surprising that so much of it is provided, financed or regulated by the Excerpted, with permission, from Public and private roles in health: Theory and financing patterns. World Bank Discussion Paper No. 339, 1996. 35 36 · Health Economics in Development state. In contrast to what happens in many other sectors of the economy, this substantial public role is most pronounced in high income countries which are generally very market-oriented; the state usually finances a smaller share of health care in poorer countries. Does this pattern provide models for the future development of the health sector and the public role in it, for low- income countries? And is the variation among countries in the amount and kind of public intervention associated with differences in people's health, in what is spent on health care or in how well health systems function? Besides consuming large resources, many health systems are regarded as inefficient or inequitable or both; they are often described as in "crisis", as needing "reform", or as having "failed". Are the supposed failures of health systems real? If so, they might be caused by misguided public intervention, so they could be corrected by a smaller or different public role and a greater reliance on private markets. Or governments might intervene for sound reasons, to correct or compensate for failings in those markets; that is, out- comes would be even worse if left entirely to the private sector. Is there an appropriate frontier between private and public action, and a best combi- nation of instruments for the state to use when it intervenes? These and related questions must be confronted in any reform effort [Aaron 1994, Cutler 1994, OECD 1994]. It is relatively simple to conclude that govern- ments should do certain things and should leave others to private activity; but often there is a variety of possible solutions and no obviously best approach. Theory does not always provide clear answers, and the empirical evidence is incomplete, extremely varied and difficult to interpret. Choices for State Intervention It matters not only whether governments intervene, but also how they do it: the second essential question is what the public sector should do, given that some problem in the private market appears to warrant some public action. This is particularly important because government failings in the health sector are also common, and often result from intervening in the wrong ways or with the wrong instruments. There are five distinct instruments of public intervention: arranged from the least to the greatest intrusion into private decisions, these are to-- · inform, which may mean to persuade, but does not require anyone to do anything. Governments do this when they publicize the health risks of smoking, or include health and basic hygiene education in public schools. These are examples of information directed at consumers, but Public and Private Roles in Health · 37 governments also inform health care providers and suppliers of health care inputs, as by conducting research and disseminating information on disease patterns and on the effects and risks of medical procedures. · regulate, which determines how a private activity may be undertaken. Governments sometimes regulate the medical profession by setting stan- dards for doctors or accrediting hospitals, although these activities may also be undertaken by private bodies. And government regulation is common in the insurance industry, in the importation of medical equip- ment, drugs and supplies and in the protection of food and water qual- ity. More generally, governments can influence private health care activ- ity in many ways, often combining regulation with some financial incentives to offset the costs [Bennett et al. 1994] but without public financing. Regulation is usually pursuant to a law, and is often deter- mined by an executive or administrative body. · mandate, which obligates someone to do something and (usually, though not always) to pay for it. Compliance with regulations can also imply substantial private costs; but a mandated activity is different in that it must be performed, whereas a private producer can react to regulation by choosing not to undertake the activity. Mandates are usually specified in law, which may subsequently be adumbrated by regulation. The most important mandates, in financial terms, are the requirements that employers provide health services or insurance to their employees, or contribute to social insurance funds for that purpose. Governments can also impose mandates on individuals, as by requiring that children enter- ing school be immunized. · finance health care with public funds. Because mandated insurance is effectively paid for by an earmarked, involuntary contribution which is equivalent to a tax, "public" health expenditure is commonly defined to include such costs along with expenditures from public budgets. The obverse of spending public funds is to tax particular activities or goods, such as alcohol and tobacco, at least partly for health reasons. This issue is not treated here; while taxation which reduces consumption of specific goods may have substantial health effects, it is always limited to very few goods and is not used systematically to promote health. Finally, the state may-- · provide or deliver services, using publicly-owned facilities and civil service staff. This is what Ministries of Health in most poor countries do; so do various governmental bodies in many countries at all income levels. 38 · Health Economics in Development Once a society has decided to finance health services with public funds, the choice arises of whether to provide them through public facilities or to pay private producers to provide them. The appropriate way to con- sider this choice is as a standard "make or buy" decision. The issues for a government are the same as for a private firm, and turn on costs--is it cheaper to produce something than to buy from an outside supplier?-- and on the risks and difficulties of enforcing contracts and avoiding fraud when dealing with such suppliers [Coase 1988]. Because public financing requires public resources, it is perhaps the cru- cial choice about state action. However, all the instruments mentioned have costs; even information is not free. The benefits from any interven- tion have always to be weighed against these costs. In addition, sometimes two instruments overlap, or one requires the use of another: mandates imply regulation, and public provision usually implies at least partial public finance. (User fees or even private insurance payments to public facilities may cover part of the cost and could in principle cover all of it.) And two instruments can be alternatives: for part of the population, governments can either finance health care or mandate financing by employers or other private institutions. Differences in ability to pay make it natural to operate mixed systems of public intervention, such as mandated coverage for the non-poor and public finance for the poor. There are alternatives even within mandated coverage, such as "play or pay" arrangements in which employers can finance health care directly for their employees or pay into a social security scheme. A Conceptual Basis for Public and Private Roles in Health The health sector is sufficiently complicated, and the conditions of coun- tries are sufficiently different, that economic theory by itself is an inade- quate guide to where the frontier should be drawn between the private economy and state action, or to which state interventions should be under- taken, and in what degree. Nonetheless, theory is essential, particularly where two issues are concerned. Both of these refer to market failure, or cir- cumstances in which private markets either cannot be expected to function at all, or can be expected to yield undesirable outcomes which appropriate public intervention might improve on. Some such failures may occur in any sector of the economy; traditional public finance theory [Musgrave 1959] Public and Private Roles in Health · 39 explains these cases and provides guidance for public action. Other possible failures arise where insurance is involved, and for reasons specific to the health sector, present particularly acute problems for health insurance [Arrow 1963, 1985]. Since what the health care sector provides to consumers and beneficiar- ies are specific activities or interventions, it is useful to organize a conceptual basis according to distinctions among these activities. The next section pro- vides this classification; three subsections elaborate on the peculiarities of each area. The three domains of health care While the activities that promote, protect or restore health are very het- erogeneous, they fall into three natural domains, corresponding to public goods, to low-cost private interventions and to catastrophically costly private goods. These domains are constructed by classifying health-related activi- ties along two dimensions, as shown in Figure 2.1: first by the degree to which they are private or public goods, and second, by how much they cost. Both dimensions refer only to characteristics of particular activities or interventions themselves, not to who consumes them or pays for them. As the descriptions of the three domains indicate, private goods are sepa- rated into low-cost and high-cost, whereas all public goods constitute one domain, regardless of their cost. The reason for this is that while the costs of public goods matter for deciding whether they should be produced, the issue of market failure related to such goods is independent of costs. With private goods, in contrast, some problems of market failure occur only with those services costly enough to be financed by insurance; and poverty, or the inability to buy even low-cost services, is a distinct reason for pub- lic intervention. This classification of health-related activities does not reserve a place for "merit goods", interventions which everyone "ought to have". To exclude this category is not to deny the social or political importance of views about what people "have a right to". The difficulty is that there is no good way to define such goods a priori, and societies make different choices about them. Moreover, so far as these goods are supposed to justify public intervention, there are often other grounds for state financing, mandating or regulating of the goods or services regarded as meritorious. For example, it is widely believed that all children have a right to immunization, but public promotion of immunization can also be justified by the market failures involved. 40 · Health Economics in Development Figure 2.1 Three Domains of Health Care Cost of an intervention (no upper limit) 6 3 5 2 1 4 Character of an intervention Public Private Public goods or large externalities Low-cost, mostly private Catastrophically costly private Examples of some specific interventions 1 Immunization 4 Treatment of minor trauma 2 Vector control 5 Normal obstetric care 3 Treatment of tuberculosis 6 Surgery for cancer A particular intervention occupies a small space in Figure 2.1; to indicate how different interventions would be classified into the three domains, some typical health care activities are located approximately. The cost of an intervention can vary depending on many factors, such as how widespread the intervention is; there can also be variation in the public or private good Public and Private Roles in Health · 41 nature of an activity. Because a health system produces a variety of inter- ventions in all three domains, it is spread widely over the space. Figure 2.1 therefore does not serve to compare different systems. Public goods are goods or services such that one person's consumption does not reduce the amount available for others to consume. Typically these are goods from which consumers cannot be excluded: if they are made available to anyone, they are available to all, at least locally or temporarily. Since people can consume such goods without having to pay for them, no one will produce them for sale to individual consumers. Therefore they will be produced only if government (or some other source such as a charitable organization) pays for their production. The notion of a public good is no different in health than in any other sector: wherever such goods or services are to be available, they must be financed by government or some other non-market alternative. Control of disease vectors and protection of food and water safety are examples of (nearly) pure public goods in health. Individual action may be ineffective (if one's neighbor's house harbors rats or mosquitoes), costly (water purification) or virtually impossible (testing for food safety). Most activities in this category are preventive, but some curative actions are also partly public. And not all preventive interventions are public goods. For example, at low levels of coverage immunization confers some public good benefits--because the immunization of part of the population reduces the likelihood that un-immunized people will become infected--but it still pro- duces mostly private benefit for the immunized individuals. However, as immunization coverage approaches 100 percent, the benefit becomes more and more public through the mechanism of "herd immunity": a lone un- immunized individual would be just as well protected as if he had been immunized, and could enjoy this protection without paying for it. When the disease can be eradicated by complete immunization coverage, a pure public good is created. This is an example of how the public or private nature of an intervention may depend on the degree of coverage. When smallpox was endemic, individuals had a strong incentive to be vaccinated, without regard for how many other people were also protected. Now that the disease has been eliminated, everyone benefits. While the distinction between public and private goods is crucial, it does not by itself define the appropriate boundary between private and state action. Moreover, the boundary between public and private goods is not sharply defined, because some interventions provide substantial externalities. (Figure 2.1 treats such interventions as partly public and partly private, rather than locating them in a separate domain.) In these cases, individuals 42 · Health Economics in Development can and do buy an intervention and benefit from it, but they cannot prevent non-consumers from also deriving some benefit. Because the purchasers do not capture all the benefit, they may be unwilling to pay for all of it: in con- sequence, private markets can exist but will produce less of these interven- tions than would be optimal for society as a whole. This problem arises most readily with communicable diseases, because the infected person puts others at risk. Curing one case therefore also prevents others. Tuberculosis control is a clear example: no victim of tuber- culosis is likely to ignore the disease, so there is no problem of people undervaluing the private benefits of treatment. Rather, the cost of treatment--and the fact that they may feel better even though the disease has not been cured--may lead people to abandon treatment prematurely, with bad consequences not only for themselves but for others. The rest of society therefore has an interest in treating those with tuberculosis, and assuming at least part of the cost. Asymptomatic communicable diseases, such as some sexually transmitted infections, also create externalities; but because people may not realize they are infected, the demand for care is too low even when care is free (zero price). There is then an argument not only for subsidizing treatment, but for persuading those infected to seek care. Most health care, however, is a (nearly) pure private good: Figure 2.1 reflects this by showing public goods as only a narrow band at the left-hand side, with private goods occupying most of the space. Largely or exclusively private activities include most curative care--especially for non-communi- cable diseases which pose no threat to others--and all rehabilitative care, and also some preventive or "pre-curative" care (such as well-baby visits, and screening for hypertension, cervical cancer, or glaucoma). They include home treatment, using health-specific purchased inputs, as well as medical or other professional care. This area shades into the myriad activities of daily behavior which also affect health, such as diet, exercise, safety pre- cautions, sexual behavior, and the use of alcohol, tobacco or drugs. Among these activities, child-rearing is of crucial importance, both for the imme- diate effect on health--young children are especially vulnerable to infections and accidents--and for the formation of life-long habits. The health effects of these behaviors are usually small on a daily or episodic basis, but can be very large cumulatively. Figure 2.1 distinguishes among private interventions according as they are cheap or costly. This is not a sharp boundary, because what is affordable for some people is out of reach for others. And activities which are individ- ually not very expensive may have to be repeated often, creating large Public and Private Roles in Health · 43 cumulative expenditures: renal dialysis and physical therapy are examples. Nonetheless it is crucial to distinguish interventions according to whether they can be paid for out-of-pocket, or financed from accumulated savings, or are so expensive as to represent a catastrophic burden. The cost of med- ical care is catastrophic if a family or individual can meet it only by selling assets, or taking on debt, to such an extent as to leave it permanently poor. As indicated above, it is initially assumed that no one is too poor to pay for interventions in the "low-cost" domain. In any health system there is always some private out-of-pocket spend- ing, corresponding to the band along the bottom of Figure 2.1. Before the modern understanding of disease and medicine, all health interventions were of this kind, and were paid for by consumers or by charity. Historically, health expenditure began in the lower right corner of the Figure and has expanded into public goods (to the left) and also into very costly private interventions (toward the upper right) as knowledge, wealth and institu- tional capacity have increased. While this pattern is general, countries have followed different paths in the expansion. Interventions which are needed unpredictably, because disease strikes randomly, and are also too costly for households' ordinary budgets or sav- ings to finance, define the domain of catastrophically expensive care. The only way to deal with the combination of high cost and uncertainty about needs is by risk-sharing, in which people finance health care collectively by contributions which are related to the expected expense in the group but not to any individual's (unknown) likely consumption of care. Although the boundaries are blurred, the domain of risk-sharing is con- ceptually quite distinct from the other two. It normally does not extend into the many routine low-cost, health-related activities, because risk is unim- portant there. (It is true that famine relief and other responses to unfore- seen disasters amount to sharing the risk of inadequate food or other daily needs, but under normal circumstances it is impossible to buy food insur- ance.) And explicit insurance--a contract between the consumer who agrees to contribute, and someone else who agrees to pay for specified care--is simply not feasible for public goods. Of course, when the government or a private charity pays for public goods, it assumes the financial risk. This can be thought of as implicit insurance, and in this sense all publicly-financed health care is a form of insurance, even if there is no explicit contract and no payment of individual claims. Risk-sharing presents the most numerous and complex issues for public policy. The growth in total health care costs is concentrated in this domain 44 · Health Economics in Development because it includes the catastrophically expensive activities. It also shows the greatest variety of institutional arrangements, including substantial partici- pation by private but non-profit providers and financing institutions [Frank and Salkever 1994]. The reason is that while insurance is the natural solu- tion to the risk of needing costly interventions, private insurance markets tend to fail in ways that affect both efficiency and equity, and different insti- tutions represent different partial solutions to those failures. These prob- lems are quite distinct from the inability of markets to deliver public goods or to assure the right level of production of goods with significant external- ities. The question is how and how far governments can and should try to correct for the failings of the insurance market, and whether public finance is necessary or whether other instruments can be substituted for it. Because neither economic theory nor common sense provides as much of an answer as in the domains of public goods and low-cost interventions [Diamond 1992, Zeckhauser 1994], judgments in the domain of risk-sharing are more tentative and depend more on empirical information. Intervention for public goods in health. By definition, public goods cannot be sold in private markets and so create a straightforward justification for col- lective action. However, the good or service also must be worth the required public expenditure: simply being a public good is an insufficient condition for state intervention. To take an extreme example, erecting giant fans to blow away polluted air would provide a public good, but would almost surely cost too much to be justified. The questions to answer then are, which goods are sufficiently public that private markets cannot provide them adequately? and how should they be valued to determine whether it is justified to pay for them? As to the first question, governments usually try to provide such indis- putably public goods as disease surveillance and sanitary inspection. They often err, however, by trying to cover too wide a range of interventions; Ministries of Health sometimes appear to regard all of health as a public good. An alternative explanation is that they regard all health care as a merit good; the belief that everyone has an unlimited or ill-defined right to free care is sometimes enshrined in legislation or national constitutions [Fuenzalida-Puelma and Connor 1989]. When public financing is insuffi- cient to fulfill that promise, and particularly when public provision is poorly managed, the result is likely to be both inefficient and inequitable. Govern- ments may err in another way, by recognizing that it can be efficient for the public sector to supply a service--the alternative being to regulate private provision--but subsidize it when most users could pay for it. Water supply Public and Private Roles in Health · 45 and sanitation services are good examples [World Bank 1992a]; they gener- ate large public health benefits but are nonetheless mostly private goods for which non-poor consumers are willing to pay. Still, the state is not always wrong when it treats a largely private good as if it were entirely public. The most striking example is immunization. Had it been left to private markets during the last few decades, it is incon- ceivable that today some 80 percent of the world's children would be immunized against the six major vaccine-preventable childhood diseases [Geoffard and Philipson 1994]. Treating the Expanded Program of Immu- nization as a public good made possible high coverage even in very poor countries [EPI 1993]--often higher than in the United States, which has relied more on private finance and provision [EPI 1995, Haveman and Wolfe 1993]. This "mistake" doubtless imposed some costs, in the form of public expenditure which was unnecessary because some people would have paid for immunization privately, and in the distortions caused by the taxes to pay for the program. But such costs are negligible in comparison with the health gains. And the public intervention in organizing and largely financ- ing the EPI did not crowd out, but probably stimulated, much private par- ticipation in both the financing and the delivery of vaccinations [van der Gaag 1995]. The second question is how to value a public-good health intervention. This is the natural domain of cost-effectiveness and cost-benefit analyses. If the only benefit from an intervention is improved health, it does not mat- ter whether that is measured in health terms (lives saved, healthy life years gained) or monetized. When there are also significant collateral benefits, different approaches can lead to different rankings of interventions. This is the case for education, water supply and sanitation, and other activities which are valuable for health but also for other reasons--and which may not be justified for the health benefits alone [World Bank 1993b]. Cost- benefit or cost-effectiveness comparisons are also relevant to public inter- vention in the other domains of health care, but they are particularly important for public goods for which no private market prices exist. The issues of which activities to consider public, and whether they deserve to be financed publicly, are difficult. Nonetheless, there are several reasons why this is the simplest domain in which to determine public pol- icy. There is broad agreement on the substantial benefits from a few crucial interventions, which are extremely cost-effective [Jamison, Mosley, Measham and Bobadilla 1993]. Individual poverty is not a major source of problems, as it is with private goods: only society's overall capacity to pay 46 · Health Economics in Development matters. Individual ignorance or absence of demand is also of little impor- tance. Finally, this area does not contribute much to the explosion of health care costs, and its financial importance declines as income rises. The public role in low-cost private interventions. This domain includes so many different activities, which are undertaken repeatedly and usually have little health impact per episode, that continued, universal, direct public interven- tion is simply impossible. Governments cannot be responsible for every- one's daily life, and can probably contribute most by improving households' capacities to look after their own health. Promoting development gener- ally--not only increased incomes but more education and access to all kinds of knowledge, goods and services--seems to be the best way to do this [World Bank 1993b, Chapter 2]. How far interference in people's ordinary behavior is justified, depends on whether the health benefits outweigh the curtailment or modification of individual choices, including non-health benefits. Apart from indefensible extreme positions--for example, that only health matters, and is worth any price; or that only people's private appreciation of their own utility matters, and should be treated as sacrosanct--there is no straightforward answer to this question. Public action cannot be justified simply because of a health improvement; neither can it be rejected just because individual liberty might be limited. Specific public intervention for improved health may be justified under three conditions: ignorance or incomplete knowledge, exter- nalities, and the failure of adults to act as appropriate agents for children. Each of these involves some kind of market failure, or violation of the pri- vate market assumption that rational adults are making informed choices and paying the consequences of their decisions. The first problem is ignorance: people might take better care of their health if they knew how. For example, vitamins are crucial to health but are not observable in food, and people may already believe untruths about diet that help cause vitamin deficiencies [Johns, Booth and Kuhnlein 1992]. In general, ignorance on the part of one or both parties to a transaction is a major source of failure in the health care market. Of course, "perfect decision making is not ever possible, so the real issue is when the government can or ought to intervene in the information mar- ket to improve the market's performance" [Beales, Crasswell and Salop 1981]. Moreover, information is not entirely free, and people do not always act on it. Thus while the cost-effectiveness of efforts to make behavior more efficient can be very high, it is also quite variable, and peo- ple's reactions to information are hard to predict. Public and Private Roles in Health · 47 Correcting ignorance is not simply a matter of telling people something new, but a larger question of changing beliefs and behavior. Where better knowledge alone does not lead to changed behavior, regulation or mandates may also be justified even though they imply more intrusive or coercive intervention. In all such cases, the difficult question is how far it is legitimate to try to change people's views of what they want or what is good for them. Information often complements these other instruments, to reduce opposition to them or improve their effectiveness. The interaction between information and other instruments of behavior change is seen clearly in the successful effort to reduce smoking in the United States [U.S. Dept. of Health and Human Services 1989, 1992]. This situation raises a second problem, of externalities, or interactions among presumably informed adults. Driving while drunk is an example, as is dumping feces or trash in communal water supplies. These activities impose both health damage and financial costs on others, and individual protection may be impossible or very costly. The chief instrument for pub- lic action is regulation, perhaps supported by mandates; these instances do not typically require public finance of health care activities. They may of course also require negative mandates, in the form of laws against certain activities or behaviors. In practice, there is no sharp boundary between this and the first problem, because some of the behavior that imposes costs on others may also arise from ignorance: thus reducing the harm from a par- ticular behavior may require both information and monetary, legal or other incentives. It is more effective to criminalize drunk driving if people are also informed of the dangers, and the health damage can be limited by man- dating the use of seat belts. The third condition is an agent-principal problem [Stiglitz 1989] and, in contrast to the externalities just discussed, is intergenerational. Children are not yet informed, sovereign adults; they are vulnerable not just to accidents and disease but to the indifference and even sadism of their parents. This problem is somewhat similar to the situation of doctors acting as imperfect agents for their patients. However, patients often can choose and contract with the doctors who act as their agents, whereas children have no choice of who acts for them. (Similar problems arise for adults who are mentally retarded or incapacitated by some kinds of disease.) What should the state do when parents are inadequate agents for their children? Requiring that children be immunized is relatively easy, but it is harder to deal with child- beating or exposure of children to secondhand smoke, and still harder to confront parents' beliefs in such matters as sexual education. Where sexual 48 · Health Economics in Development behavior, vehicle use and consumption of alcohol, tobacco and drugs are concerned, these issues continue through adolescence. This is an exceptionally contentious topic, where it is hard to draw the frontier between public and private responsibilities. Different societies have adopted different solutions, and there is often bitter disagreement within societies over the rights and duties of parents and the degree to which the state can or should interfere in family life. There are potentially very large health gains at stake in this debate: eight or nine of the ten worldwide lead- ing causes of illness in young children are substantially correctable at low cost [World Bank 1993, Annex Table B.6], and four of these--diarrheal dis- ease plus three nutritional deficiencies--can be largely controlled by the family, with little public expenditure. These diseases account for about 20 percent of young children's ill health; the total share of child health that depends on parental behavior is of course substantially larger. In all these instances, the principal instruments of state action should probably be information and regulation. Mandates are justified for a few activities such as requiring schoolchildren to be immunized, or that food- stuffs be fortified, and against a few other activities. Substantial public finance, however, is usually justified only because some people are too poor to pay for health-related goods and activities, whether these involve medi- cal care or such necessities as food. Risk-sharing for catastrophically costly private goods. When risks cannot be fully controlled, and the associated costs may be catastrophic, the only solution is to share the risk. None of the features of this domain is unique to health care, but the magnitude and interaction of certain problems are especially important in health care markets. Moreover, the health risk is only partly associated with income or employment, and the financial risk is hardly asso- ciated at all with income or occupation. Health insurance differs sharply from insurance for non-human assets such as homes or vehicles, where the value of the asset, and therefore the cost of insurance, is usually related to income. Another fundamental differ- ence is that medical care allows for preventive maintenance and for repair, but not for complete replacement of the damaged capital, which in this case is a human body. Insurance for non-human assets operates in just the oppo- site way, protecting against the loss of the asset but not paying for its upkeep. Insurance may even cover the cost of a temporary substitute for the lost home or vehicle, which is impossible with health insurance. Risks are also harder to estimate for health insurance, both because of the inherently Public and Private Roles in Health · 49 much greater complexity of the body than of non-human property and because there are often different possible treatments for a given health problem, with different costs, outcomes and risks. Insurance against health risks raises some well-known difficulties [Arrow 1985], leading to various kinds of market failure. One such problem arises because insurance is a con- tract by which someone other than the patient agrees to pay for his or her health care. As with all contracts, there is an incentive for the insured to behave differently because of the insurance; this is called moral hazard [Pauly 1968]. One consequence is that consumers who do not pay the full cost of health care will consume more of it. This is desirable, since the point of insurance is to let people consume health care they could not otherwise afford. It means, however, that the price must cover the increased demand that results from insurance, and not simply the care that people would otherwise want to buy out-of-pocket. In theory, there are two potentially more worrisome problems associated with this moral hazard. The first is that people may not only consume more medical care generally, but care that costs too much relative to its effective- ness, yielding smaller health gains per dollar spent. The second risk is that people may take poorer care of their health via daily activities, because they pay the full cost of those, but only part or none of the cost of the resulting increased curative care. Both problems imply excess resources being dedi- cated to health care. They may also imply worse health, if increased curative care does not fully compensate for reduced prevention and protection. Some degree of moral hazard is intrinsic to all kinds of insurance, but it is more lim- ited in the case of nonhuman assets because the insurance does not cover ordinary wear and tear. And cheating the insurer, by burning down one's house or abandoning one's car and reporting it stolen, is illegal, to prevent the insured person from fraudulently collecting cash. Such compensation is gen- erally not possible under health insurance. (Cash payment for permanent dis- abilities is usually included with life insurance and represents compensation for the loss of part of a life. The only significant moral hazard for such insur- ance appears to be suicide, which is often specifically excluded from the causes of death for which compensation will be paid.) Moral hazard in health insurance is independent of how it is financed, so it does not by itself deter- mine whether insurance should be paid for privately or publicly. A consumer who voluntarily buys private insurance ends up paying for the additional medical care consumed by other purchasers, and judges whether this cost is justified by his or her own greater access to care. When insurance is paid for by taxes or mandatory contributions, however, this choice cannot be made. 50 · Health Economics in Development Moral hazard may then justify controls on what public money is used for, to avoid expenditure on interventions of little health value which consumers would not voluntarily agree to buy for other people [Musgrove 1995a]. There is scant empirical evidence on the importance of these problems, particularly as to whether insurance leads people to be more careless about their health than they would be if uninsured. As to the relation between insurance and less cost-effective medical care, the evidence in the United States is that higher out-of-pocket cost for medical care (higher co-payments or lower deductibles on insurance) does not make consumers choose more cost-effective services, and may even make poor consumers forego highly justified care [Lohr et al. 1986, Newhouse et al. 1993]. Except in the latter case, there is little evidence that making the consumer pay higher costs under insurance leads to wor- sened health. Inefficiency in a competitive insurance market also takes the form of excess purchases of insurance--that is, insurance for interventions which could be more efficiently financed out of pocket [Pauly 1974], or insurance which leads to needless or unjustified use of medical care. This is inefficient to the degree that it leads to excess administrative costs for handling numer- ous small claims, and because of the excess consumption of health care [Feld- stein 1973]. Private insurers can only partly control this tendency through deductibles (which remove small risks from coverage, until out-of-pocket payments reach some limit). This problem arises partly because of ignorance: people tend to overestimate small risks and may buy too much insurance even when they pay its full cost. Moral hazard, however, is a greater problem: peo- ple who do not pay the full cost of insurance will buy too much of it, just as with medical services. Market failure in the form of over-insurance happens primarily through the tax system. Many governments allow private employ- ers to treat insurance cost as an expense, but then--in contrast to salaries-- do not treat the value of insurance as income to workers. Subsidy through the tax system is notorious in the United States [Pauly 1986], both for insurance for workers which is financed by employers and for part of the Medicare insurance for the elderly. It is estimated that employer-financed insurance would decline by one-sixth or more in the absence of this subsidy, and that in consequence the overall demand for medical services would fall by about five percent [Chernick, Holmer and Weienberg 1987]. Alternatively, governments directly subsidize social security health ben- efits (mandated insurance) out of general revenue. General revenues are used to support social security systems throughout Europe and Latin Amer- ica [McGreevey 1990]. In Chile, payroll taxes can also be used to finance private insurance. All these direct and indirect subsidies to insurance are not only inefficient, but highly inequitable when only part of the labor force is Public and Private Roles in Health · 51 covered. The poor typically benefit only when coverage is (close to) uni- versal. When various profit-maximizing insurers compete to sell insurance, there are two further and closely related problems, of adverse selection on the part of consumers and of risk selection on the part of insurers. The former refers to selection of customers which would be adverse to the interests of insurers--fundamentally, it describes the danger of enrolling people who would cost more on average than the insurance could finance. This can hap- pen because the amount of insurance coverage people want and are willing to pay for depends partly on their knowledge of their own health conditions and risks. People who expect to need little health care are unwilling to pay as much as those who expect to need much care, so a policy costly enough to cover high-risk people will lose out in the market to a cheaper policy ade- quate for low-risk people. Universal coverage at the same price for everyone may therefore be impossible to achieve, or may not generate enough revenue to finance all the health care demanded [Summers 1989]. To protect themselves against the combination of low premiums and high potential costs, insurers engage in risk selection or "cream-skimming": they spend more on administration, or create barriers to enrollment, to screen out high-risk individuals (such as the aged) or conditions (such as cancer). Such "underwriting", as it is called, is particularly costly for individual applicants for insurance, and gives rise to large scale economies because when a large group is enrolled, the insurer needs to estimate only the average risk of needing care [Diamond 1992]. This practice is the natural market response to the problem of adverse selection. Inefficiency takes the form of increased administrative costs, and also increased health risks for those excluded from insurance. Particularly when pre-existing conditions are not covered, people with health problems who are insured by their employers cannot readily change jobs without los- ing their insurance; other differences in insurance coverage may also create "job lock" among workers. This labor immobility is another source of inef- ficiency, of unknown magnitude [Congressional Budget Office 1994]. One answer to the problems of adverse and risk selection is price differentiation according to risk, which is theoretically efficient in that it allows everyone to have the insurance he or she is willing to pay for. Such price variation is common to other forms of insurance: for example, rates for automobile insurance often vary by age and by the way a vehicle is used. Unfortunately, there are serious difficulties with letting the market create comparable dif- ferentials in health insurance. One is that some people are willing to pay only a small amount, because they expect to need little medical care. Faced with a price that would cover the cost of care for everyone, they will not 52 · Health Economics in Development purchase insurance. When they drop out of the market, the price of cover- age for those who expect to need more care is driven up because the risk is spread over fewer and higher-cost people. Even if they are willing to pay more than those who anticipate needing very little care, the price of insurance may rise beyond their capacity to pay. They will be unable to buy insurance, despite a willingness to pay more than the average consumer. Such failures do not occur in other insurance markets because risks are more uniform or predictable, or more closely related to income. Of course, people's willingness to buy insurance depends on their expec- tation about future needs for medical care, and they may guess wrong. People who are young and healthy today, and therefore unwilling to spend much on insurance, may when they are older want much more medical care than they now anticipate. But if they become willing to buy substantial cov- erage only late in life, the cost will be higher than if it were spread over a longer period, so they may be unable to pay for insurance once they recog- nize the need for it. The difficulty of predicting health care needs is exacer- bated by the rapidity of technical change in this sector [Weisbrod 1991]. Another problem with differential prices is that despite the importance of many behaviors for specific health problems, rather little of total health risk is under the individual's control. People cannot be held personally responsible for much of their ill health since it is genetic in origin, or due to the actions of others. Often the best that people can do by controlling their own behavior is to postpone problems, which is very valuable but does not necessarily save money over a lifetime [Russell 1986]. Behavioral change may also take a long time to show effect on the burden of disease or the volume of treatment [World Bank 1994a]. Because so little health risk is under people's control, behavior-related prices--whether for health care or for insurance itself--are of only limited value in making markets work. One can charge people for smoking (by taxation) or for not wearing seat belts (by fines), or reward them for careful driving (by lower insurance rates), but prices are not feasible for most health-related behaviors. People have some choice of where to live and whether to drive a car, but no choice about inhabiting the body they were born with [Miller 1978]. Notions of fairness are involved in the choice of how far to allow or control price dif- ferentiation in health insurance, because people often do not think others should be punished financially, in addition to their physical suffering, for bad luck. In addition, the possibility of death or substantial permanent disability sometimes makes treatment urgent. Adverse selection is therefore a problem of equity as well as of efficiency. Public and Private Roles in Health · 53 The problems of moral hazard and adverse selection arise partly from the fact that consumers and insurers possess only incomplete information, which causes market failure in the sense that markets work perfectly only when both buyers and sellers possess full information [Arrow 1985]. Of course, insurance is wanted in the first place because people do not know what will happen to their health, and they agree to share risks when they do not know what will happen to others' health. A further complication is that of information asymmetry: information available to only one side of a market readily leads to market failure. For example, consumers who know their health risks have an incentive to conceal them from insurers so as to avoid higher premiums. They also know how they have modified their behavior, or mean to do so, because of insurance. Insurers, in contrast, generally know more than consumers about average risks and about costs of care; consumer ignorance of these matters can also lead to inefficiency. Unfortunately, it does not follow that the problems of incomplete and asymmetric information could be corrected just by supplying information. Better knowledge on the part of consumers about health risks may lead to more efficient purchase and use of insurance. However, obtaining the infor- mation needed to restore symmetry would be impossible or very costly; too much is still unknown about how much people can control their health through behavioral choices. Even if it were symmetrically available to con- sumers, insurers and providers, more information might make it easier for insurers to practice risk selection and discriminate among customers, and thereby exacerbate inequity. In fact, in an unregulated market, this is the probable consequence of the increasing availability of information linking genetic endowment to the likelihood of developing specific illnesses or health problems [House of Commons 1995]. Information asymmetry also arises between patients and doctors, since the latter typically know much more about medical conditions and treat- ments. Patients may accept, or even demand, treatments they would not buy if fully informed, but which are advantageous, financially or otherwise, to medical professionals. There is however little firm evidence as to how much of this potential "supplier-induced demand" actually occurs [Pauly 1988]. In any case, this is not simply a problem of rich countries, where most people have insurance and therefore do not worry about costs; it has also been doc- umented in poor societies where lack of education and information may make it particularly easy to exploit consumers [Bennett et al. 1994]. In summary, the consequence of these market failures is that in an unreg- ulated, competitive private market in third-party insurance those with 54 · Health Economics in Development chronic conditions or high health risks will be under-insured, administrative costs will be higher than necessary because of insurers' efforts to screen out risks and the costs of processing claims in a market with many insurers and many providers, and procedures of low or questionable value will be performed because neither the provider nor the consumer pays for them. It is in these specific senses that "the market does not work" in health care; these are primarily failures of the insurance market rather than shortcomings of the market for health care itself. Private markets have developed other forms of insurance which reduce, but do not eliminate, these problems, such as health maintenance organiza- tions (HMOs). Under this arrangement, providers also act as insurers and assume the risk. Insofar as this controls costs by shifting the burden to sup- pliers of medical care rather than to consumers or third-party insurers [Ellis and McGuire 1993], it may allow more coverage of the chronically ill, and may reduce the utilization of relatively ineffective procedures. However, without some form of public intervention such arrangements will have lit- tle effect on the problems caused by poverty and adverse selection. The question remains which kinds and degrees of public intervention can best mitigate the problems inherent to private insurance markets without intro- ducing worse inefficiencies or inequities. Market Failure and Health Care Needs Since the unregulated, unsubsidized private market is the extreme alternative to government intervention in health care, much of the debate as to appro- priate public and private roles in the sector turns on whether, how, and how badly markets may fail. Market failure, as an economic notion, refers to pos- sible mismatches or disequilibria between what the market supplies, and what fully-informed, rational consumers of health care would demand. It does not deal with the concept of need for health care, which is theoretically an unsatisfactory concept but is also difficult to do without [Culyer 1995]. People want health care not for any intrinsic utility but because they think they need it, that if care is not provided their health will deteriorate or fail to improve. In contrast, much of the criticism by both health care profession- als and consumers of how health systems operate deals explicitly with needs. Just as demand and supply may be out of balance, there can be imbalance between demand and need or between need and supply of services, as shown in Figure 2.2. Market failures in the narrow economic sense are among the Public and Private Roles in Health · 55 reasons for these imbalances (these are indicated on the Figure by asterisks). Some failures result from barriers to the operation of competitive private markets in bringing supply and demand together. Others distort demand from what it would be if based on complete and symmetric information and if there were no public goods or externalities; this causes imbalance between demands and needs. While competitive private markets are generally the best way to bring demand and supply together, they are much worse suited to make either demand or supply match people's needs. Public intervention in the health market, in contrast, is aimed at satisfying those needs, and runs the corresponding risk of failing to take account of demand. Either a purely private or a purely public health care system is likely to control one of the three potential imbalances, at the cost of failing to control or even worsen- ing one or both of the others [Musgrove 1995b]. This is a major reason why most health care systems are far from being all private or all public [Dun- lop and Martins 1995]. Figure 2.2 Need, Demand and Supply for Health Care NEED Poverty High cost High cost Lack of information (*) Market incentives Supplier induced demand (*) Externalities (*) Lack of information (*) Public goods (*) SUPPLY DEMAND Lack of information (*) Other barriers to competition (*) Non-market incentives (*)Sources of market failure Source: Musgrove (1995b). 56 · Health Economics in Development Dealing with poverty In the discussion thus far, it has been assumed that no one is too poor to buy a variety of health interventions out of pocket. Similarly, it is assumed that no one is too poor to buy insurance against catastrophic health risks; peo- ple differ in income and in their assessment of risks, but everyone can afford some insurance. This means that the distinction between the "low-cost" domain in Figure 2.1 and the domain where insurance is needed is roughly the same for everyone. A minimal state role in the absence of poverty. Under these conditions the state's role in the domain of private, inexpensive health-related activities would be limited mostly to information and regulation; there would be no reason to finance this kind of health care publicly if everyone could afford it out-of-pocket. (Mandates might still be justified to deal with some of the externalities mentioned above.) In the domain of risk-sharing, people who chose not to buy insurance, or bought too little of it, would have to pay for care out of pocket or do without it. Because doing without care would sometimes pose the risk of avoidable death, some kinds of care--emergency services, at least--are typically available even to the uninsured. People's willingness to let others suffer the consequences of imprudence does not usually include letting them die because they bought too little catastrophic protection. This kind of imprudence constitutes moral hazard, and unlike some behavior on the part of insured people, it cannot be dealt with by dif- ferential premiums. Motorcyclists who prefer not to wear helmets can in principle be charged more for their insurance, just as smokers can; the more difficult problem is how much care to provide for the uninsured cyclist whose injuries are worse because of failure to wear a helmet. This situation provides a justification for enough public finance or man- dated insurance to cover the cost of a few crucial services to which everyone would have access and for which everyone would have to pay through taxes [Summers 1989]. Except for these services, there would be no requirement for the government to subsidize insurance for anyone. If there were no poverty, then, the role of the state in the health sector might be relatively lim- ited, and would--except for the minimum insurance requirement just described--concentrate on the adequate provision of public goods and the correction of market failure in the domain of risk-sharing. Whether in order to correct or compensate for that failure the state should mandate or finance insurance beyond that minimum of emergency care, is a question of the social efficiency of doing so, rather than leaving insurance to the private market. Public and Private Roles in Health · 57 The fact that insurance would not need to be subsidized--the removal of an equity justification for interfering in the insurance market--does not mean that governments should do nothing, because the efficiency failings of private voluntary insurance are as important as the inequities to which it gives rise. How poverty complicates public roles. The existence of poverty, of people too poor to buy many "inexpensive" health activities or an "adequate" amount of insurance, complicates the question of what the state should and should not do in several ways. These complications are not limited to the domains of private goods, because problems of public health are often more severe among the poor. They are likely to be at particular risk from contaminated air and water and so to benefit more than the non-poor from public health interventions [World Bank 1992a, 1993b], and they often suffer more seri- ous consequences from common illnesses. In general, imbalance between need and demand may be more important where the poor are concerned, because they have less knowledge on which to base their wants for health care as well as less resources with which to express demand. Medical indigence is in most respects no different from poverty with respect to food and other basic needs, and, as with those needs, the rest of society may agree to subsidize the poor. The difference is that poverty rel- ative to predictable, low-cost needs such as food can be dealt with either by transfers or subsidies in kind, or by supplementing income [Srinivasan 1994]. With health, the risk of needing very costly care generally makes it more efficient to deal with medical indigence by subsidizing insurance than through income transfers. However, there is little experience in most poor countries in subsidizing private providers or insurers to meet the health needs of the poor. This requires government administrative capacity and appropriate pricing mechanisms, to prevent excess provision and even out- right fraud, which has been a major problem, for example, in Brazil [World Bank 1993c; Medici and Czapski 1995]. A public subsidy to private insurance, as an alternative to dealing directly with large numbers of providers, also requires premiums differentiated by age, sex or other conditions, to reduce the scope for risk selection and make it feasible to mandate universal coverage. In consequence, one of the most important effects of poverty is that it makes public provision, with all its typical problems, look attractive or even necessary in poor countries. Aside from the problems of regulation, it is financially difficult to provide the poor with the same level of services enjoyed by those already covered by private or social insurance. Even extending social security coverage to the poor, to replace the more limited services offered by Ministries of Health, 58 · Health Economics in Development would be very costly in many Latin American countries [Mesa-Lago 1992]. Public provision often means poor health care for the poor, but public financing of private services would not easily solve the underlying financial problem even if it led to improved quality of care. Finally, in countries where all public money now flows through government or parastatal facili- ties, shifting to public finance of private providers requires that public hos- pitals and clinics be privatized or at least given sufficient autonomy and capacity to manage themselves and compete for public funds against other providers. Such changes are potentially very valuable, but they are likely to be particularly difficult, since public facilities need to be exposed to some financial risk without the danger of collapse in publicly funded provision. The difficulty of incorporating the poor into the same insurance schemes which cover the non-poor, whether by extending social security coverage or by subsidizing the purchase of private insurance, leads to efforts to create insurance specifically for the poor, typically at the community level. Any such scheme is intrinsically limited by the low incomes of participants, so it cannot finance very costly interventions and can only yield subsidies from the less poor to the more poor. It may nonetheless be appropriate when the insurance is meant to pay for only such health care as can also be provided locally and which therefore is not very costly--although perhaps still catas- trophic for a poor family to finance. Unfortunately, the problems of moral hazard and adverse selection arise even in these circumstances. For example, if insurance is sold for short periods to accommodate families' fluctuations in income, then, as occurred in Burundi, people may buy the "health cards" entitling them to services only when they are already sick or can anticipate a medical need [McPake, Hanson and Mills 1993]. That effectively elimi- nates the difference between an insurance payment and a fee, and reduces the amount of money that can be raised by the scheme. These problems lead to complications such as rewarding people who use less (curative) health care by reducing the cost of their cards for the next period or by charging an additional fee or "fine" to those who pay for insurance only when ill [Chabot, Boal and da Silva 1991]. Such incentives work against moral haz- ard and adverse selection, but if they are large enough to have much effect they may greatly reduce the scheme's revenue or the demand for services. And administrative expenses may absorb a large share of revenue. Poverty also creates or strengthens reasons for the state to intervene in low-cost, health-related activities, whether these are inexpensive medications and services or such non-medical items as food supplements. Some inter- ventions can be accommodated by broadening public health services, for example by including micronutrient supplements or treatment for intestinal Public and Private Roles in Health · 59 parasites. Others can be covered by financing private providers, such as the clinics which operate under "covenants" with the Brazilian social security system [World Bank 1993c]. However these activities are dealt with, poverty pushes governments to finance a wider range of low-cost interventions and to rely less on information and regulation. If this were the only force at work, it would lead to a larger public share of health expenditure in poor than in rich countries, simply because there are more poor people who cannot pay for those interventions. In the sum of health spending, however, this effect is overwhelmed by the tendency (and the capacity) of governments to mandate or finance more insurance for the non-poor, as income rises. A third effect of poverty is to limit the use of prices to curtail demand or control costs. Being poor already greatly constrains demand, and poor peo- ple are necessarily more sensitive to prices for health care than the non- poor [Gertler and van der Gaag 1992]. This means both that user fees can raise relatively little revenue from the poor, and that unless there are offset- ting improvements in quality, utilization may be sharply reduced [Lavy 1994, Litvack and Bodart 1993]. The experience with user fees has been extensively analyzed, notably in Sub-Saharan Africa in connection with the Bamako Initiative [Griffin 1987, Creese 1991, World Bank 1992, Vogel 1993, Makinen and Raney 1994, Nolan and Turbat 1994, Shaw and Griffin 1995]. There is evidence that--as might be expected--utilization declines, sometimes sharply, if fees are raised but nothing else changes. There is rather less information on service characteristics, such as whether user fees improve the availability of drugs. And almost nothing is known about the impact on health outcomes or on system efficiency or cost-effectiveness. Fees are sometimes set arbitrarily or with inconsistent criteria; charges low enough to have no effect on the poor may or may not be worth collecting; and targeting by exempting the destitute from fees does not have to be expensive [Grosh 1992], but there is a risk of high administrative costs and low net revenues. The same problems arise for collecting insurance deductibles and co-payments from the poor. Not only does poverty increase the risk of ill health; sickness and disabil- ity can make or keep people poor. The relation between health and poverty is sometimes regarded as another reason for the state to invest in health, in order to raise productivity. However, the fact that some health care increases incomes is not a separate objective for government action. If health care made people so much more productive that the extra income could pay for the health care, then in perfect markets people could borrow against their future productivity. When capital market failures prevent such borrowing, and those failures cannot be corrected directly, then any public intervention--such 60 · Health Economics in Development as financing the health care or providing loans to consumers--that secures the health gains will also yield the increase in productivity. Justifications and Risks of State Intervention As the foregoing analysis shows, there are three distinct, independent argu- ments for governments to intervene in health care rather than leaving it entirely to private markets. One is to ensure the optimal level of production and consumption of public goods and goods which have a partly public char- acter because of externalities. These can be health care services themselves, activities protective of health, or information that helps people take better care of their health and make better use of services. A second reason is to make insurance work more efficiently and more equitably, for those services which can be produced in private markets but for which risk-sharing is required because of high costs and uncertainty about needs. The third reason is to subsidize those too poor to buy insurance or even, sometimes, those inexpensive activities and services which the non-poor can afford out of pocket. These three reasons derive from the three domains of health care defined by cost and by the public or private nature of services. Market failures underlie two of these reasons, but in different ways. In the case of public goods and externalities, the failure arises from the nature of the good or serv- ice. In contrast, problems in insurance markets arise from the way the good is financed. None of the three reasons is unique to the health sector, but all are more important in this sector than in much of the rest of the economy. The arguments for not leaving health care and health insurance to uncontrolled private markets are all arguments that efficiency or equity can be improved, if the state intervenes appropriately. They are not arguments that anything the public sector does, will improve matters. Just as there is a well-defined set of market failures typical of the health sector, there are con- sistent government failures, ways in which governments act to create worse outcomes than could be reached, and in some respects even worse outcomes than markets would generate. The most common and severe criticism of public action concerns provision [World Bank 1980 and 1987, Birdsall and James 1992]: especially in poor countries, governments offer medical care which is supposed to be free to users, on equity grounds, but which is centrally-controlled, under-financed and of poor quality in both medical and human terms. Because the budgets of public facilities often are unre- lated to service output, and civil service rules make it difficult to fire, trans- fer or discipline unproductive staff, the costs of health gains may be very high even if salaries and other input costs are low. And the pervasive lack of Public and Private Roles in Health · 61 incentives for efficiency means that capital is also bought in excess, not maintained, and under-utilized. The result is that even rather poor people, the supposed beneficiaries of the public system, often pay out of pocket for those private services they can afford. This makes them pay twice for some of their care, exacerbating the inequities arising from the tax system and from difficulties of access due to the geographic location of facilities. Governments typically fail where provision is concerned, by trying to do too much and by competing with private providers only in price terms-- that is, subsidizing provision rather than competing on quality and satisfac- tion. With respect to the other instruments of state action, failures are more varied, and often result from doing too little rather than too much. This is likely to be the case particularly for regulation and for the dissemination of information. Mandates show a very mixed pattern: middle-income countries in particular often mandate insurance for part of the population through social security schemes, but do not effectively mandate either insurance or care for everyone. Richer countries, in contrast, appear much less prone to government failure largely because they rely much more heavily on regula- tion and mandates, and much less on public provision. Where public facili- ties are important, as in some European countries, they operate under greater autonomy than in poor countries, and this is balanced by greater regulation of private providers. The result is to concentrate more on the right roles for public action, and less on dividing the health sector into dis- jointed private and public spheres. The distinction is particularly important because in many countries, the two sectors overlap greatly: the same professionals work part-time in each, private providers often use public facilities to treat private patients, and so on. To provide public goods and to subsidize health-related activities for the poor, two of the three main reasons for state action, both require public finance. In both these areas there is also room for the other instruments of state action; and the problems associated with risk sharing can lead to various combinations of interventions, which may or may not include spending public money. Societies therefore have much latitude in how much, and by what means, the government intervenes in health care markets, just as they have in deciding how much to spend on health in relation to income and to their health problems or needs. The Appropriate Public Role in Health As a first approximation, it is easier to say what governments should not do in health than to specify what they should do. That is, it is clear that certain actions are likely or certain to violate one or more of the objectives of 62 · Health Economics in Development a health care system to an important degree. To apply a topographic metaphor: such actions correspond to falling off a plateau of satisfactory outcomes, all somewhat different but none clearly dominating the rest, and into one of the surrounding chasms of one or another kind of failure. Four "don'ts" are discussed in what follows: they refer to the way the public is taxed, charged or exempted to pay for health care; the providers to which governments transfer public funds; the way providers are paid; and the serv- ices they are paid to provide. What government should not do. The first thing governments should not do is to use the tax system, or any system of fees at public facilities, to make the poor subsidize the health care of the rich. Conceptually, subsidies are justi- fied only for the poor, and broader financing of care through insurance is a question of efficiency rather than of equity. This is not only a matter of whether the rich use more of publicly-financed services than the poor do, although great inequities often arise because the rich have more access to those services. Financial equity also depends on who pays the taxes. More narrowly, governments should not contribute to social security financing from general revenues, unless coverage is universal, because when only part of the population is covered it is usually the poor who are excluded. And governments should not treat private insurance coverage as a cost to employers unless it is also treated as income to beneficiaries. Such practices are not only inequitable; they are also inefficient to the extent that they lead to excessive spending on health care, or reduce labor mobility. Controlling inequitable subsidies does not mean that social security sys- tems should be dismantled or must be made universal. Even when incom- plete, such mandated insurance includes substantial progressive subsidies from high-paid to low-paid workers, and the mere fact that some people in society receive more generous health care is not necessarily a problem, so long as they pay for it. What matters is that governments not make everyone pay for what some are excluded ex ante from receiving. Perverse subsidies not only cause immediate inefficiencies or inequities, they also create interests that oppose subsequent health system reform. This is evi- dent in the United States [Skocpol 1995] and Chile [Musgrove1995d], and it is the reason why the design of subsidies is a crucial part of any reform to extend or improve coverage for the poor. The second thing governments should not do is tie public finance to public provision. The choice of whether to provide care through public or parastatal facilities should be treated as a "make or buy" decision, subordinate to the larger decision about what to pay for. That does not necessarily mean eliminating public Public and Private Roles in Health · 63 provision, which will sometimes be the best solution. It means rather that competition between public and private providers should be based on costs and on quality, and not on price to the consumer, as is commonly the case. To achieve those goals, however, requires other changes in how public institutions operate, changes that probably cannot occur so long as public funds go automatically and exclusively to government facilities. This con- clusion is most pertinent to poor countries where public systems are most likely to be inefficient and to be used only because they are free or nearly so. High-income countries with a large share of public provision, particu- larly in hospitals, suffer less from these problems. A third thing governments should not do is pay for health care by fee-for-service, unless other mechanisms are used to control expenditures. This is seldom a problem with publicly provided services, but makes it hard to control costs when governments finance private providers. Even negotiating or controlling the fees is not enough, as the Canadian experience demon- strates, since providers can respond to fees they consider too low, by increas- ing output [Evans, Barer and Labelle 1988]. This helps explain why Canadian health expenditure has risen faster than that in European countries which rely on other payment mechanisms. European countries which pay for health care by fee-for-service also rely on global budgets, utilization reviews or other instruments of cost containment [OECD 1994, 1995]. Since expenditures equal prices times quantities, and since quantities of services respond to prices and are difficult to control directly, countries which pursue both macroeco- nomic and microeconomic efficiency in health spending usually control both prices and expenditures. In contrast to European and North American expe- rience, Brazil applies this system to block grants for federal expenditure in states with relative financial autonomy; but in states where federal money is paid directly to providers, the federal government also controls one crucial quantity, the number of hospitalizations that federal money will pay for [World Bank 1994b]. In all cases, the government sets prices for services. It also helps, in controlling expenditures, to define "services" as complete treatments for specific conditions rather than as all the individual compo- nents of such treatments, so they can be financed by mechanisms such as the Diagnostic Related Group (DRG) payments used in the United States and the provider can be required to assume part of the financial risk. Changing from a system based on overall budgets and on salaries for providers, to a fee-for-service system without offsetting the resulting incentives to over- provision can lead to an explosion of costs, as in the Czech Republic [Boland 1995]. Finally, if governments mean to pursue some combination of better 64 · Health Economics in Development health and lower costs, they should not--in fact they cannot--simply finance whatever people demand when care is free to consumers. This does not mean governments should not "subsidize demand" rather than "subsi- dizing supply" or providing services. It means only that there must be limi- tations on what will be paid for publicly. Such limitations are a particularly contentious matter, because both good health and cost containment may be opposed to consumer or provider satisfaction, which are also politically important objectives. Nonetheless, two empirical observations are germane to this decision. One is that private insurance always carries some limita- tions, either as to the services covered or as to cost-sharing. Except where poverty is important, so that cost-sharing is more difficult, there is no rea- son for public finance to be systematically more generous in this regard than private risk-sharing arrangements. The second observation is that, as the discussion concerning Figure 2.2 indicates, private markets will tend to sup- ply what people demand, and public intervention typically acts to emphasize needs instead. In other forms of public subsidy, it is common to distinguish between wants and needs, and concentrate spending on the latter. This is the case, for example, with food subsidies, which are also very much health- related. Price subsidies are usually limited to "basic" foodstuffs, and food stamps cannot legally be used to buy alcohol or other non-necessities. This may be considered partly a matter of efficiency, assuring more health gain than would otherwise occur. But it is also a matter of equity: in contrast to actuarial private insurance, where every purchaser buys the expected value of the health services needed, public finance is involuntary. It comes from taxpayers who have a legitimate interest in meeting needs, and thereby get- ting value for their money, but not necessarily in paying for wants. None of these "don'ts" is easy to implement, because some consumer or provider interests can be expected to oppose every one of them, in every country. But they are arguably the most important conclusions about where the public/private frontier should run in the health sector. It is notable that none of these conclusions depends solely on features unique to the health sector. The peculiarities of health matter most for the difficulty of distin- guishing "needs" from wants, and--partly as a consequence--for the dangers of paying providers for whatever they choose to provide, without incentives to control costs. A system which avoided the problems described above would still present complex and difficult questions for the proper role of the state, but it would have more latitude to pursue improvements. And it would not matter so much, exactly what objectives the government pursued or which combination of instruments it applied. Public and Private Roles in Health · 65 What government should do. Beyond the prescriptions for how the state should deal with the problems of each of the three domains of health care, several "dos" appear to be generally valid for governments. If the objective is to min- imize deadweight losses from public intervention and leave as much room as possible for private choices, then the first thing governments should do is to use each less-intrusive instrument to the point where a more intrusive inter- vention is justified, following the sequence of increasingly greater interfer- ence--inform, regulate, mandate, finance and provide services. That is, governments should regulate private activity when merely improving people's information is not enough, deliver services if it is infeasible to finance private providers equitably, and so on. Public finance is inescapable for some actions, but particularly in low- and middle-income countries much can probably be accomplished by better use of information and regulation. Failure to use these other instruments well, can increase the need for public finance. Sometimes the problem is that governments exploit these instruments too little. They do not regulate private insurance when it first begins to expand, which makes subsequent regulation politically more difficult [World Bank 1994b; Musgrove 1995c]; or they do not initially react when health-damaging behaviors such as smoking become more entrenched. Sometimes the prob- lem is inappropriate regulation, which needlessly restricts competition, or enforces inefficiency in the public sector by centralizing nearly all decisions. And poor countries in particular often get the worst of both worlds by pay- ing for activities which should be, but are not, regulated: government subsidy of medical education without adequate control of quality or relation to needs is a common example. In the worst of cases, governments use all the available instruments in exactly the reverse order. They try to provide more health care than they can pay for, with the result that most services are under-financed and of poor quality; they try to finance services, some of which might be mandated and paid for by consumers or employers; they mandate care, as by social security systems, without adequately regulating it; and they do too little to inform the public and providers either of dangers to health or of how the health care system is actually working. Much of the criticism of government failure in the health sector, especially in poor countries, describes the result of getting things backwards. These ideas are represented in Figure 2.3A, which shows the appropriate relation among the instruments by which the state can intervene in health care, and Figure 2.3B, which portrays the kind of inappropriate or imbalanced relation often found in poor nations. Figure 2.3A indicates that whatever is mandated, financed or provided publicly should also be regulated, and much else 66 · Health Economics in Development Figure 2.3A Appropriate Use of the Instruments of Public Intervention in Health Care M P I R F Information to consumers, I providers and insurers Regulation of personal behavior R and firms, providers and insurers Mandates for health care or M insurance coverage Financing of health care (explicit F or implicit public insurance) P Provision of health care Public and Private Roles in Health · 67 Figure 2.3B Typically Inappropriate Use of the Instruments of Public Intervention in Health Care F P MM I R I Information to consumers, providers and insurers R Regulation of personal behavior and firms, providers and insurers M Mandates for health care or insurance coverage F Financing of health care (explicit or implicit public insurance) P Provision of health care 68 · Health Economics in Development besides. The state can finance some care that is not anyone's mandated responsibility, and can also mandate care or insurance coverage which is financed privately. And most if not all public provision should be fully financed by government, but possibly with much more scope for finance than for provision by the state. The largest sphere pertains to information, covering activities in which there is no other public interference with the market. In contrast, Figure 2.3B shows a much smaller effort to inform or regulate, very little use of mandates that are not also publicly financed, and a sphere of provision as large as, or even larger than, that of finance. Of all the instruments of public action, regulation may be the most under-utilized. Brazil and Chile both provide examples of the resulting problems. In Brazil, the state finances three-quarters of medical care but directly provides only about one-quarter of it, so the instruments of finance and provision are used in the appropriate order. But there is very little reg- ulation of the competence of medical professionals, of the quality of care, or of the rapidly-growing private insurance industry. The lack of regulation even interferes with financing the system, since private insurers sometimes send their customers to publicly financed facilities without paying for care, and until recently fraud was widespread [World Bank 1994b; Medici and Czapski 1995]. In Chile, the private insurance industry was created by pub- lic action, with essentially no regulation--but with a mandate allowing peo- ple to spend on private insurance, the tax contributions that formerly were used to finance the public system [World Bank 1994a]. The lack of regula- tion may not have affected medical quality, but it has worsened the financial situation of public facilities, raised administrative costs, and promoted risk selection. Even more serious failures to regulate have arisen in Eastern and Central Europe, as former state monopolies of health finance and provision have given way to competitive private provision. A second thing governments should do is to stimulate competition in the provision of health care. This is largely a matter of promoting public/private competition, for the reasons described above, but it also includes removing any unjustified barriers to competition within the private sector, and between for-profit and non-profit providers such as non-governmental organizations. The lack of competition is usually less of a problem in the domain of low cost, health-related activities than in the domain of costlier activities requir- ing insurance. This recommendation extends a fortiori to non-medical components of health care such as the "hotel" services of hospitals. How far competition should be carried is not always obvious. For example, whether public facilities should make their own purchasing decisions for such inputs Public and Private Roles in Health · 69 as drugs, depends on whether central bulk purchasing reduces costs while maintaining adequate supplies. Even in the latter case, there should of course be competition among suppliers for such purchases. Except for the risk that providers will compete by offering more services rather than by raising quality or reducing costs--a risk that is greatest when payment is by fee-for service--competition appears to be beneficial in health care provision, just as in other industries. Competition is less desirable in health care financing, both because administrative costs are likely to be higher and because it is competition among insurers that leads to risk selection. Experience in OECD countries suggests that good results can be obtained with one or with many insurers, but only if they are closely regulated. Third, governments should put as much of the incentive for cost con- tainment as possible on the supply side of the market, rather than on con- sumers. This is almost a necessity where poor consumers are concerned, since their poverty already sharply limits what they can spend. But the evidence is that even non-poor consumers do not respond to higher prices by using health services more cost-effectively. It also appears that providers have considerable scope for controlling expenditure by limiting volume as well as unit costs. And theory indicates that an optimal payment system should use supply-side measures to control costs; reimbursing providers fully according to costs is never the best solution [Ellis and McGuire 1990, 1993]. As income increases and poverty declines, of course, it becomes eas- ier to pass the burden of cost containment to consumers. However, it does not become medically any more effective or economically any more effi- cient to do so. Moreover, as income increases the capacity for supply-side responses by providers also increases, so it continues to be preferable to keep cost control incentives on the supply side of the market. Finally, it is urgent to deal with the pervasive problem of government fail- ure, and to improve the capacity to do whatever government ends up doing. This is especially important when market failures are so important that for the public sector to withdraw, on the ground that it also is subject to failure, would only make matters worse. Much of the criticism that governments, particularly in poor countries, try to do too much in health arises because of how badly they appear to operate, more than from any evidence that they have exceeded some optimum degree of state intervention. More skill and understanding, and fewer internal barriers to efficiency, make sense whether the state's role shrinks, as by leaving more provision to the private sector, or expands, to finance more coverage for the poor. One of the things govern- ments generally need to do better, particularly in poor countries where much 70 · Health Economics in Development private medical practice may be of low quality, is to use regulation, mandates, training and other interventions to help the private sector function better. This is increasingly recognized as an essential component of almost any health sector reform, and the need for it will only increase as health systems become more expensive and complex. All five instruments of intervention-- information, regulation, mandates, finance and provision--need to be used well, and using less of one instrument and more of another will not, in gen- eral, reduce the need for governments to perform capably. References Aaron, Henry J. 1994. Issues every plan to reform health care financing must con- front. Journal of Economic Perspectives. Vol. 8, No. 3. Abel-Smith, Brian. 1995. Assessing the experience of health financing in the United Kingdom. In Dunlop and Martins 1995. Abel-Smith, Brian. 1984. Cost containment in 12 European countries. World Health Statistics Quarterly. Vol. 37. Arrow, Kenneth J. 1963. Uncertainty and the welfare economics of medical care. American Economic Review. Vol. 53, No. 5. Arrow, Kenneth J. 1985. Theoretical issues in health insurance. In Collected Papers of Kenneth J. Arrow, Vol. 6, Applied Economics. Cambridge: Harvard University Press. Baker, Judy, and Jacques van der Gaag. 1993. Equity in health care and health care financing: evidence from five developing countries. In Van Doorslaer, Wagstaff and Rutten 1993. Barnum, Howard, Joseph Kutzin and Helen Saxenian. 1995. Incentives and provider payment methods. The World Bank. HRO Working Paper 51. Barr, Nicholas. 1992. Economic theory and the welfare state: a survey and inter- pretation. Journal of Economic Literature. Vol. 30, No. 2. Beales, Howard, Richard Craswell, and Steven C. Salop. 1981. The efficient regulation of consumer information. Journal of Law and Economics. Vol. 24, No. 3. Beaton, George H., R. Martorell, K. A. L'Abbe, B. Edmonston, G. McCabe, A. C. Ross and B. Harvey. 1993. Effectiveness of Vitamin A Supplementation in the Control of Young Child Morbidity and Mortality in Developing Countries: Summary Report. Toronto: University of Toronto. Behrman, Jere. 1990. The action of human resources and poverty on one another: what we have yet to learn. Washington, D.C.: World Living Standards Measurement Study, Working Paper No. 74. Bennett, Sara, George Dakpallah, Paul Garner, Lucy Gilson, Sanguan Nitta- yaramphong, Beatriz Zurita and Anthony Zwi. 1994. Carrot and stick: state mechanisms to influence private provider behavior. Health Policy and Planning. Vol. 9, No. 1. Public and Private Roles in Health · 71 Bitran, Ricardo, and D. McInness. 1991. Health care demand in Latin America: les- sons drawn from the Dominican Republic and El Salvador. Washington, D.C.: World Bank, Economic Development Institute (processed). Blendon, Robert J., Mollyann Brodie and John Benson. 1995. What happened to Americans' support of the Clinton Plan? Health Affairs. Summer. Bobadilla, Jose Luis. 1996. Searching for Essential Health Packages in Low- and Middle- Income Countries. Washington, D.C.: World Bank (processed). Boland, Vincent. 1995. Czech health minister sacked in costs crisis. Financial Times. 10 October. Bovbjerg, Randall R., Charles C. Griffin and Caitlin E. Carroll. 1993. U.S. health care coverage and costs: historical development and choices for the 1990s. Journal of Law, Medicine and Ethics. Vol. 21, No. 2. Chabot, Jarl, Manuel Boal and Augusto da Silva. 1991. National community health insurance at village level: the case from Guinea-Bissau. Health Policy and Planning. Vol. 6, No. 1. Chernick, Howard A., Martin R. Holmer and Daniel H. Weinberg. 1987. Tax pol- icy toward health insurance and the demand for medical services. Journal of Health Economics. Vol. 6. Coase, Ronald. 1988. The Firm, the Market, and the Law. Chicago: University of Chicago Press. Cochrane, A. L., A. S. St. Leger and F. Moore. 1978. Health service "input" and mortality "output" in developed countries. Journal of Epidemiology and Community Health. Vol. 32. Congressional Budget Office, U. S. Congress. 1994. An analysis of the Administration's health proposal. Creese, Andrew. 1991. User charges for health care: a review of recent experience. Health Policy and Planning. Vol. 6, No. 4. Culyer, A. J. 1990. Cost containment in Europe. In Organization for Economic Cooperation and Development. 1990. Health Care Systems in Transition: The Search for Efficiency. Social Policy Studies No. 7. Paris: OECD. Culyer, A. J. 1995. Need: the idea won't do - but we still need it. Social Science and Medicine. Vol. 40., No. 6. Cumper, George E. 1991. The Evaluation of National Health Systems. New York: Oxford University Press. Cutler, David M. 1994. A guide to health care reform. Journal of Economic Perspectives. Vol. 8, No. 3. Diamond, Peter. 1992. Organizing the health insurance market. Econometrica. Vol. 660, No. 6. Dunlop, David W., and Jo. M. Martins, eds. 1995. An International Assessment of Health Care Financing: Lessons for Developing Countries. Washington, D.C.: World Bank, Economic Development Institute. Ellis, Randall P., and Thomas G. McGuire. 1990. Optimal payment systems for health services. Journal of Health Economics. Vol. 9. 72 · Health Economics in Development Ellis, Randall P., and Thomas G. McGuire. 1993. Supply-side and demand-side cost sharing in health care. Journal of Economic Perspectives. Vol. 7, No. 4. Elola, Javier, Antonio Daponte and Vicente Navarro. 1995. Health indicators and the organization of health care systems in Western Europe. American Journal of Public Health. Vol. 85, No. 10. Expanded Programme on Immunization (EPI). 1993. Newsletter. Washington, D.C.: Pan American Health Organization. December. Expanded Programme on Immunization (EPI). 1995. Newsletter. Washington, D.C.: Pan American Health Organization. April. Evans, Robert G., Morris L. Barer, and Roberta J. Labelle. 1988. Fee controls and cost control: tales from the frozen north. Milbank Memorial Fund Quarterly. Vol. 66, No. 1. Evans, Robert G., and Maureen M. Law. 1995. The Canadian health care system: where are we and how did we get there? In Dunlop and Martins 1995. Feldstein, Martin S. 1973. The welfare loss of excess health insurance. Journal of Political Economy. Vol. 81. Foldvary, Fred E. 1994. Public Goods and Private Communities. Cornwall, England: Hartnolls Limited. Frank, Richard G., and David S. Salkever. 1994. Nonprofit organizations in the health sector. Journal of Economic Perspectives. Vol. 8, No. 4. Fuenzalida-Puelma, Hernan L., and Susan Scholle Connor. 1989. The Right to Health in the Americas: a Comparative Constitutional Study. Washington, D.C.: Pan American Health Organization. Geoffard, Pierre-Yves, and Tomas Philipson. 1994. Market structure and disease eradication: private versus public vaccination. Chicago: Department of Economics, University of Chicago (processed). Gertler, Paul, and Jacques van der Gaag. 1990. The Willingness to Pay for Medical Care: Evidence from Two Developing Countries. Baltimore, Maryland: The Johns Hopkins University Press for the World Bank. Govindaraj, Ramesh, Christopher J. L. Murray and Gnanaraj Chellaraj. 1994. Health Expenditures in Latin America. Cambridge: Harvard Center for Population and Development. Griffin, Charles C. 1987. User Charges for Health Care in Principle and Practice. Washington, D.C.: World Bank Economic Development Institute. Seminar Paper No. 37. Griffin, Charles C. 1992. Health Care in Asia: A Comparative Study of Cost and Financing. Washington, D.C.: World Bank. Grosh, Margaret. 1992. Administering Targeted Social Programs in Latin America: From Platitudes to Practice. Washington, D.C.: World Bank. Haveman, Robert, and Barbara Wolfe. 1993. Children's prospects and children's policy. Journal of Economic Perspectives. Vol. 7, No. 4. Public and Private Roles in Health · 73 Heysen, Socorro, and Philip Musgrove. 1986. Interdepartmental differences in life expectancy at birth in Peru as it relates to income, household drinking water, and provision of medical consultations. Bulletin of the Pan American Health Organization. Vol. 20. House of Commons. 1995. Science and Technology Committee. Human Genetics: Minutes of Evidence. Wednesday 17 May. London: Her Majesty's Stationery Office. Jamison, Dean, W. Henry Mosley, Anthony R. Measham and Jose-Luis Bobadilla, eds. 1993. Disease Control Priorities in Developing Countries. Oxford: Oxford University Press. Johns, T., S. L. Booth and H. V. Kuhnlein. 1992. Factors influencing vitamin A intake and programmes to improve vitamin A status. Food and Nutrition Bulletin Vol. 14 No. 1. Kim, Kwanrgkee, and Philip M. Moody. 1992. More resources better health? a cross national perspective. Social Science and Medicine. Vol. 34, No. 8. Lachaud, Claire, and Lise Rochaix. 1993. France. In Van Doorslaer, Wagstaff and Rutten 1993. Lavy, Victor. 1994. Household responses to public health services: cost and quality tradeoffs. Conference on social sector investments held at the World Bank, Washington, D.C., 18 May (processed). Lazenby, Helen C., Katharine R. Levit, Daniel R. Waldo, Gerald S. Adler, Suzanne Letsch and Cathy A. Cowan. 1992. National health accounts: les- sons from the U.S. experience. Health Care Financing Review. Vol. 13, No. 4. Litvack, Jennie, and Claude Bodart. 1993. User fees and improved quality of health care equals improved access: results of a field experiment in Cameroon. Social Science and Medicine. Vol. 37, No. 3. Lohr, K. N., R. H. Brook, C. J. Kamberg, G. A. Goldberg, A. Leibowitz, J. Keesey, D. Reboussin and J. P. Newhouse. 1986. Effect of cost-sharing on use of med- ically effective and less effective care. Medical Care. Vol. 24. Makenbach, J. P. 1991. Health care expenditure and mortality from amenable con- ditions in the European Community. Health Policy. Vol. 19. Makinen, Marty, and Laura Raney. 1994. Role and Desirability of User Charges for Health Services. Bethesda, Maryland: Abt Associates, Inc. McGreevey, William. 1990. Social Security in Latin America, Issues and Options for the Bank. Washington, D.C.: World Bank Discussion Papers, 110. McGuire, Alistair, David Parkin, David Hughes and Karen Gerard. 1993. Econometric analyses of national health expenditures: can positive economics help to answer normative questions? Health Economics. Vol. 2. McPake, Barbara, Kara Hanson and Anne Mills. 1993. Experience to date of Implementing the Bamako Initiative: a Review and Five Country Case Studies. London: London School of Hygiene and Tropical Medicine. 74 · Health Economics in Development Médici, André Cézar, and Cláudio André Czapski. 1995. Evolução e Perspectivas dos Gastos Públicos em Saúde no Brasil. Consultants' Report. Washington, D.C.: World Bank. Mesa-Lago, Carmelo. 1992. Health Care for the Poor in Latin America and the Caribbean Washington, D.C.: Pan-American Health Organization and Inter- American Foundation. Milgrom, Paul, and John Roberts. 1992. Economics, Organization and Management. Englewood Cliffs, New Jersey: Prentice Hall. Miller, Jonathan. 1978. The Body in Question. New York: Random House. Murray, Christopher J. L., Ramesh Gavindaraj and Philip Musgrove. 1994. National health expenditures: a global analysis. In World Health Organization 1994. Musgrave, Richard A. 1959. The Theory of Public Finance, A Study in Public Economy. New York: McGraw-Hill. Musgrove, Philip. 1986. Measurement of equity in health. World Health Statistics Quarterly. Vol, 39, No. 4. Musgrove, Philip. 1995a. Cost-effectiveness and the socialization of health care. Health Policy. Vol. 32. Musgrove, Philip. 1995b. Mismatch of need, demand and supply of services: pic- turing different ways health systems can go wrong. Washington, D.C.: World Bank, HCO Working Paper No. 59. Musgrove, Philip. 1995c. Reformas al sector salud en Chile: contexto, lógica y posi- bles caminos. In Giaconi, Juan G., ed., La Salud en el Siglo XXI: Cambios Necesarios. Santiago, Chile: Centro de Estudios Públicos. Navarro, Vicente. 1989. Why some countries have national health insurance, oth- ers have national health services, and the U. S. has neither. Social Science and Medicine. Vol. 28, No. 9. Newhouse, Joseph P., et al. 1993. Free for All? Lessons from the RAND Health Insurance Experiment. Cambridge: Harvard University Press. Nolan, Brian, and Vincent Turbat. 1995. Cost Recovery in Public Health Services in Sub-Saharan Africa. Washington, D.C.: The World Bank, Economic Development Institute. Okun, Arthur M. 1975. Equality and Efficiency: the Big Tradeoff. Washington, D.C.: The Brookings Institution. Organization for Economic Cooperation and Development (OECD). 1992. The Reform of Health Care Systems: a Comparative Analysis of Seven OECD Countries. Paris: OECD. Organization for Economic Cooperation and Development (OECD). 1994. The Reform of Health Care Systems: a Review of Seventeen OECD Countries. Paris: OECD. Organization for Economic Cooperation and Development (OECD). 1995. New Directions in Health Care Policy. Health Policy Studies No. 7. Paris: OECD. Pan American Health Organization (PAHO). 1994. Health Conditions in the Americas. Washington, D.C.: PAHO. Public and Private Roles in Health · 75 Pauly, Mark V. 1968. The economics of moral hazard. American Economic Review. Vol. 58. Pauly, Mark V. 1974. Overinsurance and public provision of insurance: the roles of moral hazard and adverse selection. Quarterly Journal of Economics. Vol. 88. Pauly, Mark V. 1986. Taxation, health insurance, and market failure in the medical economy. Journal of Economic Literature. Vol. 24. Pauly, Mark V. et. al. 1988. Issues related to volume and intensity of physician serv- ices. Report prepared for the U. S. Health Care Financing Administration. Baltimore. Pelletier, David, L. 1994. The relationship between child anthropometry and mor- tality in developing countries: implications for policy, programs and future research. Journal of Nutrition. Vol. 124, No. 10S. Poullier, Jean-Pierre. 1992. Administrative costs in selected industrialized countries. Health Care Financing Review. Vol. 13, No.4. Ricklefs, Robert E., and Caleb E. Finch. 1995. Aging: a Natural History. New York: W. H. Freeman, Scientific American Library. Russell, Louise B. 1986. Is Prevention Better Than Cure? Washington, D.C.: The Brookings Institution. Schieber, George, Jean-Pierre Poullier, and Leslie M. Greenwald. 1992. U. S. health expenditure performance: An international comparison and data update. Health Care Financing Review, International Comparison of Health Systems. Vol. 13, No. 4. Schieber, George, Jean-Pierre Poullier, and Leslie M. Greenwald. 1994. Health System Performance in OECD Countries, 1980-1992. Health Affairs. Fall. Skocpol, Theda. 1995. The rise and resounding demise of the Clinton Plan. Health Affairs. Spring. Shaw, R. Paul, and Charles C. Griffin. 1995. Financing Health Care in Sub-Saharan Africa through User Fees and Insurance. Washington, D.C.: World Bank, Directions in Development. Shea, Dennis G., and R. Patrick Stewart. 1995. Demand for insurance by elderly persons: private purchases and employer provision. Health Economics. Vol. 4. Srinivasan, T. N. 1994. Destitution: A Discourse. Journal of Economic Literature. Vol. 32. Stiglitz, Joseph. 1989. Principal and agent. In Eatwell, John, Murray Milgate and Peter Newman, eds. The New Palgrave: Allocation, Information and Markets. New York: W. W. Norton. Summers, Lawrence H. 1989. Some simple economics of mandated benefits. American Economics Association Papers and Proceedings. Vol. 79, No. 2. Tresseras, R., J. Canela, J. Alvarez, J. Sentis and L. Salleras. 1992. Infant mortality, per capita income, and adult illiteracy: an ecological approach. American Journal of Public Health. Vol. 82, No. 3. U. S. Department of Health and Human Services. 1989 Reducing the Health Consequences of Smoking: 25 years of Progress. A Report of the Surgeon General. 76 · Health Economics in Development Rockville, Maryland: Prepublication Version, DHHS Publication Number (CDC) 89-8411. U. S. Department of Health and Human Services. 1992. Smoking and Health in the Americas. Atlanta, Georgia: DHHS Publication Number (CDC) 92-8419. Van der Gaag, Jacques. 1995. Public and Private Initiatives: Working Together for Health and Education. Washington, D.C.: World Bank, Directions in Development. Van Doorslaer, Eddy, Adam Wagstaff and Frans Rutten. 1993. Equity in the Finance and Delivery of Health Care: an International Perspective. Oxford: Oxford University Press. Vogel, Ronald J. 1993. Financing Health Care in Sub-Saharan Africa: A Policy Study. Silver Spring, Maryland: Basic Health Management, Inc. for the African Development Bank. Weisbrod, Burton A. 1991. The Health Care Quadrilemma: An Essay on Technological Change, Insurance, Quality of Care, and Cost Containment. Journal of Economic Literature. Vol. 29, No. 2. pp. 523-552. Wolfe, Barbara L. 1986. Health status and medical expenditures: is there a link? Social Science and Medicine. Vol. 22, No. 10. World Bank. 1980. Health: Sector Policy Paper. Washington, D.C.: World Bank. World Bank. 1987. Financing Health Services in Developing Countries: An Agenda for Reform. A World Bank Policy Study. Washington, D.C.: World Bank. World Bank. 1991. World Development Report: The Challenge of Development. Washington, D.C.: World Bank. World Bank. 1992a. World Development Report: Development and the Environment. Washington, D.C.: World Bank. World Bank. 1992b. China: Long-Term Issues and Options in the Health Transition. Washington, D.C.: World Bank. World Bank. 1993a. Argentina: Public Finance Review--from Insolvency to Growth. Washington, D.C.: World Bank. Report No. 10827-AR. World Bank. 1993b. World Development Report: Investing in Health. Washington, D.C.: World Bank. World Bank. 1993c. The Organization, Delivery and Financing of Health Care in Brazil. Washington, D.C.: World Bank. Report No. 12655-BR. World Bank. 1994a. Chile: The Adult Health Policy Challenge. Washington, D.C.: World Bank. Report No. 12681-CH. World Bank. 1994b. Private Sector and Social Services in Brazil: Who Delivers, Who Pays, Who Regulates. Washington, D.C.: World Bank. Report No. 13205-BR. World Bank. 1995. The World Bank in the Health Sector. Washington, D.C. World Health Organization. 1994. Global Comparative Assessments in the Health Sector: Disease Burden, Expenditures and Intervention Packages. Geneva: World Health Organization. Zeckhauser, Richard. 1994. Public finance principles and national health care reform. Journal of Economic Perspectives. Vol. 8, No. 3. CHAPTER 3 The Rationale for Government Intervention in the Tobacco Market Acknowledgments I am grateful to all those colleagues at the World Bank who provided the opportunity to work on this subject, including those whose criticisms required sharpening the arguments; among the latter, Michael Walton deserves particular mention. Introduction There is no doubt that prolonged smoking is an important cause of prema- ture mortality and disability worldwide. Strictly on health terms, then, there is a strong reason to intervene to reduce this damage. However, smoking is voluntary and is not illegal for adults, so the exis- tence of an enormous health problem is not, prima facie, sufficient to justify interference with people's choice to smoke. An economic rationale for such intervention requires that failures in tobacco markets are sufficiently large to justify the costs of such interference. Despite the strong consensus that smoking harms health, there is much debate about proper government roles, if any, in reducing smoking (see, for example, The Economist 1997). Co-authored with Prabhat Jha, Frank Chaloupka, and Ayda Yurekli. Excerpted and reprinted, with permission, from "The Rationale for Government Intervention", in Prabhat Jha and Frank Chaloupka, eds., Tobacco Control Policies in Developing Countries. Oxford: Oxford University Press for the World Bank and the World Health Organization, 2000. 77 78 · Health Economics in Development We explore the economic rationale for government intervention in tobacco markets. We first discuss the two key market failures that justify government intervention on efficiency grounds: first, consumers' incom- plete information about the risks of addiction and disease; and, second, external costs. We do not deal with supply side market failures, such as the monopoly power of the tobacco industry. Next, we discuss which interven- tions are available to governments to correct these market failures, noting their specificity and effectiveness and their economic costs. We focus in this section on interventions that would protect children and adult non- smokers, and that would inform adult smokers. Third, we discuss whether government intervention in tobacco markets is appropriate to reduce inequity between rich and poor. This exploration will take account of particular epidemiological features of the tobacco epidemic that are relevant to the economic arguments. The first of these is the early age at which people typically start smoking, which, in high-income countries at least, is during the teen years. The risk of lung can- cer is far higher in individuals who start smoking at age 15 and smoke one pack a day for 40 years than among those who start at age 35 and smoke two packs a day for 20 years (Peto 1986). Therefore, the early age of onset has a direct bearing on individuals' health risks. From the standpoint of economics, the early typical age of onset is also relevant because the standard economic concept of consumer sovereignty, which holds that the consumer knows what is best for him or her, may not apply so forcefully to adolescents as to adults. The second key epidemiological feature of the tobacco epidemic is that fully half of smoking related deaths occur in productive middle age (defined as 35­69 years) (Peto et al. 1994). This is relevant to the economic debate about smoking, since it dispels the notion that smoking kills people mostly in old age, when the economic losses (as well as the health losses) are small. Inefficiencies in the Tobacco Market Smokers clearly receive benefits from smoking; otherwise they would not pay to do it. The perceived benefits include pleasure and satisfaction, stress relief (presumably derived in part from the nicotine content of the smoke), peer acceptance, and a sense of maturity and sophistication (most important for adolescent smokers, and derived from the act of smoking as such). An addi- tional important benefit for the addicted smoker is the avoidance of nicotine withdrawal. There is little that economics can say about the preferences that The Rationale for Government Intervention · 79 determine smoking, except to try to understand how the addictive nature of cigarettes influences subsequent consumption. As with other addictive behav- iors, the decision to start and the "decision" to continue are quite different, and different economic arguments may be relevant to each. The private costs to be weighed against those benefits include money spent on tobacco prod- ucts, damage to health, and nicotine addiction. Defined this way, the perceived benefits evidently outweigh the perceived costs for at least 1.1 billion people who smoke today. Economic theory assumes that the consumer knows best and that privately determined consumption will most efficiently allocate soci- ety's scarce resources. Thus, if smokers know their risks and internalize all their costs and benefits, there is no justification, on the grounds of ineffi- ciency, for governments to interfere (Pekurinen 1991). However, these assumptions may not hold for several reasons, leading to market failures. (Note that even efficient markets do not necessarily achieve equity, and that inequity is not normally classified as market failure. We dis- cuss equity issues later in the chapter.) Below, we analyze three failures in the tobacco market. The first is incomplete information about health risks. The second is incomplete information about addiction, specifically the complex issue of children's tendency to under-estimate the addictive poten- tial of smoking (and therefore the costs of quitting). The third failure con- sists of costs imposed on others. Incomplete information about health consequences Incomplete information about the risks of smoking leads to behavior that smokers would not otherwise choose for themselves. Poorly-informed smokers often underestimate the risks of their action (Weinstein 1998). Since people usually react to known risks by reducing the risky consumption, incomplete information means more smoking than would otherwise occur. There are two principal reasons why smokers tend to be inadequately informed. The first is that the market, far from providing information, has actually hidden or distorted it. The second is the long delay between start- ing to smoke and the onset of obvious disease, which has obscured the link between the two. Each of these are discussed in turn. The tobacco industry, like other industries, has no financial incentive to provide health information that would reduce consumption of its products. On the contrary, the indus- try has consistently hidden product information on the ill effects of smoking or actively misinformed smokers about risks (Sweda and Daynard 1996). Notably, the industry has used advertising and promotion to promote its 80 · Health Economics in Development products as `safe' despite internal evidence that all types of smoking are harmful. For example, the industry has tried to advertise filter cigarettes as `healthier' (USDHHS 1989). The industry has also used advertising to reach young smokers (Institute of Medicine 1994). Other tactics of the industry to leave smokers uninformed or misinformed include dissuading lay journals from reporting on smoking's health effects (Warner et al. 1992), and spon- soring biased scientific research (Bero et al. 1994). Internal industry docu- ments uncovered in recent lawsuits in the United States confirm such practices (Glantz et al. 1995). Second, consumers derive information on the costs and benefits of smok- ing primarily from their own experience and what happens to their peers, as well as from studies largely financed by the public sector. However, the obvi- ous health damage from smoking usually emerges at least 20­30 years after exposure. This differs from most other risky behaviors, such as fast driving, where the costs and benefits are more readily and immediately appreciated. The long delay between exposure and effect has also impeded the growth of scientific knowledge. In the United States, the 1960s evidence suggested that only one in four smokers died from smoking. When risks were re- assessed decades later, when the epidemic had matured, the evidence showed that the risks were actually much higher: one in two long-term smokers die from smoking (see Doll et al. 1994; Peto et al. 1999). Anyone who considered starting or continuing smoking 20 or 30 years ago in high- income countries would, therefore, have under-estimated the risks, even if he or she had based the decision on the best available information. More- over, as the list of diseases and conditions associated with smoking expands, smokers continue to under-estimate the risks. Most developing countries still do not have estimates of the health hazards of smoking for their own populations. It is, therefore, not surprising that even respectable journals, such as The Economist (1997), reveal their confusion about the scale of the true risks or the high proportion of smokers who die in middle age: ". . . most smokers (two-thirds or more) do not die of smoking- related disease. They gamble and win. Moreover, the years lost to smoking come from the end of life, when people are most likely to die of something else anyway." As Kenkel and Chen (2000) discuss, there are two key features of con- sumers' incomplete information: first, in low-income and middle-income countries, absolute awareness of the health risks is still comparatively low. For example, in China, about two-thirds of adult smokers surveyed in 1996 believed that cigarettes did them "little or no harm" (Chinese Academy of The Rationale for Government Intervention · 81 Preventive Medicine 1997). Second, consumers in all countries may not clearly internalize the risks, even when they have been informed about them, nor may they accurately judge the risks of smoking relative to other environmental exposures, such as `stress' or radiation. Children and teenagers generally know less about the health effects of smoking than adults. A recent survey of 15- and 16-year-olds in Moscow found that more than half either knew of no smoking-related diseases or could name only one, lung cancer (Levshin and Droggachih 1999). Even in the United States, where young people might be expected to have received more information, almost half of 13-year-olds today think that smoking a pack of cigarettes a day will not cause them great harm (National Cancer Policy Board 1998). In addition, teenagers--even those with good understanding of the risks of smoking--may have a limited capacity to use information wisely. Teenagers behave myopically, or short-sightedly. It is difficult for most teenagers to imagine being 25, let alone 55, and warnings about the dam- age that smoking will inflict on their health at some distant date are unlikely to reduce their desire to smoke. In developing countries, there is less awareness of the hazards of smok- ing at all ages, including among adults, for several reasons. Education levels are lower, and, since education leads to more rapid and thorough absorp- tion of information, it is reasonable to conclude that less-educated popula- tions will be less receptive to health information. There are fewer local data on the hazards of smoking and less dissemi- nation of existing data on health risks. Governments less often regulate industry information practices, such as advertising and promotion. For all these reasons, it is unlikely that current smokers and potential smokers in low-income and middle-income countries have adequate knowledge from which to make informed decisions. Inadequate information about addiction The second major information failure in the tobacco market involves inad- equate information about nicotine addiction. Smokers acquire psychological addiction to the act of smoking itself, and physical addiction to nicotine (Kessler et al. 1997). Psychological addiction to cigarettes is hardly different from habit formation with respect to other products or practices. Nicotine addiction, however, is not simply a matter of choice or taste reinforced by repetition, such as choosing to listen to certain music or keeping company with dangerous friends. Of course, as with all biologically addictive goods, 82 · Health Economics in Development many people can change their behavior and quit using nicotine, as the decline in smoking among adults in high-income countries demonstrates. However, the costs of quitting are significant, so much so that some people find quitting virtually impossible. Most smokers who quit have to make sev- eral attempts before they succeed, and former smokers remain vulnerable to resuming smoking at times of stress (USDHHS 1990). Is addiction alone reason enough for governments to intervene against smoking? If children had full information about the likelihood of becoming addicted and understood the long-run implications of their addiction, they might conceivably become `happy addicts' who are maximizing their own welfare by smoking. For example, the teenager might argue that it would be `better to suffer lung cancer at age 60 than to suffer Alzheimer's disease at age 80'. Models of so-called `rational addiction' (Becker and Murphy 1988) assume that individuals maximize utility over their lifetime, taking into account the future consequences of their choices. However, the key assumptions of the model are that people are fully rational, that they are far- sighted about their choices, and that they have full information on the costs and benefits of their choices. These assumptions are not satisfied in the case of smoking. Children are more myopic, or `short-sighted', than adults, and they typically have less information. Recent extensions to the rational addic- tion model by Orphanides and Zervos (1995) take some of this into account when looking at youthful `decisions' to become addicted. In their model, imperfect information about addiction early in life can result in seemingly rational decisions that are later viewed with regret. Other recent theoretical work emphasizes the role of `adjustment costs' for addictive goods (Suranovic et al. 1999). The presence of these adjust- ment costs, in the context of less than fully rational behavior, implies that smokers may continue to smoke while regretting this decision, given that the costs of stopping are greater than the costs of continuing. In this con- text, rather than providing benefits, continued smoking for an addicted smoker is the lesser of two evils. Some might interpret the differences between the short- and long-run price elasticities of demand for an addic- tive good as reflecting the magnitude of these adjustment costs. That is, much of the difference between the long-run and short-run consumer sur- plus may be thought to reflect the adjustment costs. Assuming a linear demand curve, and given the evidence that the long-run elasticity for cigarette demand is about double the short-run elasticity, this suggests that as much as half of perceived consumer surplus (based on short-run demand) reflects the adjustment costs associated with addiction. The Rationale for Government Intervention · 83 Perhaps most importantly, there is clear evidence that young people under-estimate the risk of becoming addicted to nicotine, and, therefore, grossly under-estimate their future costs from smoking. Among high-school seniors in the United States who smoke but believe that they will quit within five years, fewer than two out of five actually do quit. The rest are still smok- ing five years later (Institute of Medicine 1994). In high income countries, about seven out of ten adult smokers say they regret their choice to start smoking and two-thirds make serious attempts to quit during their life (USDHHS 1989). In sum, it is the combination of imperfect information about addiction and myopia that results in significant under-estimation of the risks of future health damage. In the absence of addiction, teenagers could more easily quit later, when they become aware of the health risks, as they tend to do where other risky behaviors are concerned. We discuss this further below. The risk that young people will make unwise decisions is rec- ognized by most societies and is not unique to choices about smoking, although in the case of smoking it is compounded by addiction and inade- quate information. Therefore, most societies restrict young people's power to make certain decisions. For example, most democracies prevent their young people from voting before a certain age; some societies make educa- tion compulsory up to a certain age; and many prevent marriage before a certain age. The consensus across most societies is that some decisions are best left until adulthood. Likewise, many societies consider that the freedom of young people to choose to become addicted should be restricted. It might be argued that young people are attracted to many risky behav- iors, such as fast driving or alcohol binge-drinking, and that there is nothing special about smoking. However, few other risky behaviors carry the high risk of addiction that is seen with smoking, and most others are easier to abandon or modify, and are abandoned or modified in maturity (O'Malley et al. 1998; Bachman et al. 1997). For example, teenagers often binge drink, but most grow to be responsible moderate drinkers later in life. Driving motor vehicles is risky, but most young drivers survive long enough to learn to drive more responsibly. With smoking, there is no comparable way to behave more prudently, except to quit; even cutting back somewhat on con- sumption does not reduce the risks proportionally. Also, compared with other risky behaviors, such as alcohol use, new recruits to smoking face a very high probability of premature death. These factors combined create a probability of addiction and premature death that is higher than for other risk behaviors. Using estimates from Murray and Lopez (1996) and WHO (1999), and studies in high-income countries, we estimate that of 1000 84 · Health Economics in Development 15-year-old males currently living in middle-income and low-income coun- tries, 125 will be killed by smoking before age 70 if they continue to smoke regularly. By comparison, before age 70, 10 will die because of road acci- dents, 10 will die because of violence, and about 30 will die of alcohol-related causes, including some road accidents and violent deaths. The tobacco industry has a clear incentive to subsidize or to give away free cigarettes to potential smokers, especially young people, in order to induce them to smoke and become addicted to nicotine (Becker et al. 1994; Ensor 1992). The same incentive applies to creating addiction among adults in low-income and middle-income countries by manipulating price. Thus, at best, nicotine addiction greatly weakens the argument that smokers should exercise consumer sovereignty. Given the myopia of young consumers and the likelihood of information failure for all smokers, it is inappropriate to regard an addiction-induced demand as representing gen- uine welfare gains to the smoker. External costs Consumers and producers in any transaction may impose costs or benefits on others, which are known as externalities. The costs--or benefits-- imposed by smokers on others are of three types. First are the direct physical costs for non-smokers who are exposed to others' smoke. Second are the financial externalities that cause monetary loss (or gain) for non- smokers, whether or not they are exposed to smoke. Last (and most diffi- cult to assess) are the so-called `caring externalities' or `existence value' effects of smoking, whereby non-smokers suffer emotionally from the illness and death of smokers unrelated to them personally. Physical externalities. Physical externalities from smokers involve both health effects for non-smokers, such as a higher risk of disease or death, and other effects, such as the nuisance of unpleasant smells, physical irritation, and smoke residues on clothes, and the greater risks of fire and property damage. The health effects are briefly summarized. They include, for children born to smoking mothers, low birth weight and an increased risk of various dis- eases (USDHHS 1986; Charlton 1996), and an increased risk of various dis- eases in children and adults chronically exposed to environmental tobacco smoke either at home or in the workplace (Environmental Protection Agency 1992; Wald and Hackshaw 1996). Importantly, the list of diseases and conditions associated with environmental tobacco smoke is expanding (California Environmental Protection Agency 1997). The Rationale for Government Intervention · 85 Financial externalities. Financial externalities are costs that are imposed by smokers but at least partly financed by non-smokers. In countries where there is an element of publicly financed healthcare, these include medical costs, among them the costs of treating the newborns of mothers who smoke during pregnancy. Non-smokers also help to pay for the damage from fires and the higher maintenance costs of workplaces and homes where smokers are present. Here we briefly summarize the key arguments related to healthcare costs and to pensions. In high-income countries, the overall annual cost of healthcare that may be attributed to smoking has been estimated to be between 6% and 15% of total healthcare costs. In most low-income and middle-income countries today, the annual costs of healthcare attributable to smoking are lower than this, partly because the epidemic of tobacco-related diseases is at an earlier stage, and partly because of other factors, such as the kinds of tobacco- related diseases that are most prevalent and the treatments that they require. However, these countries are likely to see their annual smoking related healthcare costs rise in the future as the tobacco epidemic matures (World Bank 1992). For those concerned with public spending budgets, it is vital to know these annual healthcare costs and the fraction borne by the public sector, because they represent real resources that cannot be used for other goods and services. For individual consumers, on the other hand, the key issue is the extent to which the costs will be borne by themselves or by others. As the following discussion shows, the assessment of these costs is complex, and therefore it is not possible yet to draw definitive conclusions about whether or how they may influence smokers' consumption choices. In any given year, on average, a smoker's healthcare is likely to cost more than that of a non-smoker of the same age and sex. However, because smok- ers tend to die earlier than non-smokers, the lifetime healthcare costs of smokers and non-smokers in high income countries may be fairly similar. Studies that measure the lifetime healthcare costs of smokers and non- smokers in high-income countries have reached conflicting conclusions. In the Netherlands (Barendregdt et al.1997) and Switzerland (Leu and Schwab 1983), for example, smokers and non-smokers have been found to have sim- ilar costs, while in the United Kingdom (Atkinson and Townsend 1977) and the United States (Hodgson 1992), some studies have concluded that smok- ers' lifetime costs are, in fact, higher. Part of this confusion stems from the fact that it is relatively easy to make actuarial estimates of the potential for smokers' earlier deaths to bring savings in public health or pension 86 · Health Economics in Development expenditures. In contrast, the external financial costs of smoking are more difficult to measure reliably, and may be considerably under-estimated (Chaloupka and Warner, 2000). Recent reviews that take account of the growing number of tobacco-attributable diseases and other factors con- clude that, overall, smokers' lifetime costs in high-income countries are somewhat greater than those of non-smokers, despite their earlier deaths (Chaloupka and Warner, 2000). There are no such reliable studies on lifetime healthcare costs in low-income and middle-income countries. Clearly, for all regions of the world, smokers who assume the full costs of their medical services will not impose costs on others, however much greater those costs may be than non-smokers'. In developing countries, higher proportions of healthcare costs are borne by private individuals, rather than by the public system (Bos et al. 1999). Nonetheless, even in low- income countries, a significant percentage of medical care, especially that associated with hospital treatment, is financed either through government budgets or through private insurance. To the extent that taxes, co- payments, or social insurance premiums are not differentially higher for smokers, the higher medical costs attributable to smokers will be at least partly borne by non-smokers. To the extent that private business healthcare costs are passed on to consumers in the form of higher prices, or to work- ers in the form of lower wages, any costs incurred by workers who smoke will similarly be partly passed on to non-smokers. However, such costs are small in low-income and middle-income countries (Collins and Lapsley 1998). Out-of-pocket payments and risk-adjusted insurance schemes do not burden non-smokers with some of the costs of smokers. For private insur- ance, where premiums for non-smokers are lower than for smokers, there may be little economic justification for public intervention. In reality, how- ever, most health insurance plans are increasingly group-based and contain no risk-adjustment for smoking. In low-income and middle-income countries, intra-household transfers of income or welfare may be as important a source of externalities as for- mal, extra-household transfers (James 1994). Manning et al. (1991) and others argue that intra-household transfers are irrelevant, since adults' decisions to smoke are made on behalf of a whole household, and reflect the preferences of all family members. This is implausible, since adults are likely to become smokers before marrying or having children. They are likely to find it difficult to quit later--even if spouses or children urge them to. Furthermore, very young children, who may be the most severely affected by exposure to others' smoke, have no voice in such decisions. The Rationale for Government Intervention · 87 Spouses may, in deciding to marry, have taken into account the addiction of their partner, and may, therefore, be said to acquiesce in the decision; but that is not the same thing as helping to make the decision or approving of it. In high-income countries, public expenditure on health accounts for about 65% of all health expenditures, or about 6% of GDP (Bos et al. 1999). If smokers have higher net lifetime healthcare costs, then non-smokers will subsidize the healthcare costs of smokers. The exact contribution is comp- lex and variable, depending on the type of coverage and the source of taxa- tion that is used to pay for public expenditures. If, for example, only the healthcare costs of those over 65 are publicly funded, then the net use of public revenues by smokers may be small, to the extent that many require smoking-related medical care and die before they reach this age. Equally, if public expenditure is financed out of consumption taxes, including cigarette taxes, or if third-party private insurance adjusts smokers' premiums because of their higher health risks, then their costs may not be imposed on others. Once again, the situation differs in low-income and middle-income coun- tries, where the public component of total healthcare expenditure is on average lower than in high-income countries, at around 44% of the total, or 2% of GDP (Bos et al. 1999). However, as countries spend more on health, the share of total expenditure that is met by public finance tends to rise too (World Bank 1993). While it is difficult to assess the relative healthcare costs of smokers and nonsmokers, the issue of pensions has proved at least as contentious, and has attracted some popular debate. For example, an editorial in The Econo- mist (1995) expressed the view that smokers `pay their way'. It continued: ". . . what they cost in medical bills, fires and so on, they more than repay in pensions they do not live to collect." This assertion is based on analyses from high-income countries that sug- gest that smokers contribute more than non-smokers to pension schemes, because many pay contributions until around retirement age and then die before they can claim a substantial proportion of their benefits (Manning 1989; Viscusi 1995). There are several problems with this assertion. First, there is an ongoing academic debate over definitions of the social costs of smoking, and particularly the extent to which `savings' from not collecting pensions should be included. Depending on differing assumptions, other studies (see, for example, Atkinson and Townsend 1977) have not found net costs for smokers to be lower. Second, the issue is not currently relevant to many of the low-income and middle-income countries where most of the 88 · Health Economics in Development world's smokers live. In low-income countries, only about one in ten adults has a public pension, and in middle-income countries the proportion is between a quarter and half of the population, depending on the income level of the country; private pension plans are less common (James 1994). Finally, and perhaps most importantly, most of these studies have followed traditional notions of economic externalities, and have not placed any value on life per se. Even if smokers do reduce the net costs imposed on others by dying young, it would be misleading to suggest that society is better off because of these premature deaths. To do so would be to accept a logic that says society is better off without its older adults (Harris 1994). Caring externalities. The third group of externalities that we consider are those that are the most difficult to assess: they are known as `existence value' or `caring' externalities (Krutilla 1967). There is evidence that people are willing to pay for another's well being, even if they do not know the person and even if they do not benefit directly themselves. Public spending on health partly reflects such externalities. Existence value is most readily applied to children, whom society typically protects more than adults. In contrast, caring externalities for adults almost directly contradict the notion of consumer sovereignty. Clearly, caring externalities differ across cultures and countries, depending among other things on the importance society assigns to individual sovereignty. Nonsmokers may be willing to subsidize efforts to prevent people taking up smoking or efforts to help smokers quit. They may also be prepared to contribute towards the care of sick smokers, even when these represent a financial burden. However, their attitudes may change over time as knowledge about the health effects of smoking becomes more widespread and non-smokers' tolerance for smokers may decline (Gorovitz et al. 1998). In any case, there is little solid information of such willingness, so it is difficult to use it to formulate public policies. In sum, there are clearly direct costs imposed by smokers on non-smokers, such as health damage. There are probably also financial costs, although it is more difficult to identify or quantify these. Government Responses to Market Failure: What, For Whom and at What Price? Given that, as we have argued, the markets for tobacco products suffer effi- ciency failures that result in premature death and illness, and costs imposed on others, it is appropriate to ask if government intervention can correct The Rationale for Government Intervention · 89 them. Here we ask whether governments have interventions available to correct these failures, and discuss the costs and effectiveness of these interventions. Below we describe briefly those interventions that respond to, or deal with, each of the types of inefficiency in the tobacco market that we have described above. Governments can use information, regulation, taxation, or subsidies to address these market failures. Government responses to incomplete or erroneous information include, specifically, mass information campaigns, warning labels, and publicly- financed research to create more, or better, or more easily assimilated, infor- mation. All are public goods, which the market is unlikely to provide adequately. Public responses to existing addiction in adults include, specifi- cally, incentives to quit, such as cessation programs (with or without pharma- cological therapies) offered free or at subsidized prices, and education campaigns that raise awareness of the risks of smoking and the benefits of ces- sation. In addition, governments can encourage deregulation of the market for nicotine replacement therapy. Public responses to preventing new addic- tion in children (discussed in more detail below) include education campaigns about the danger of addiction, restricting children's access to tobacco prod- ucts, bans on the advertising and promotion of tobacco products, and taxa- tion. Increased taxation will also increase cessation rates among adults. Government responses to direct physical externalities include education campaigns emphasizing the right of non-smokers to a smoke-free environ- ment, restrictions on smoking in public places and workplaces, and taxes. Government responses to financial externalities may include risk-adjusted health or pension premiums, or anything that restricts tobacco consumption, whether or not in the presence of non-smokers. These may include taxation, information campaigns, and restrictions on where people can smoke. Government responses to `existence value' externalities also include any intervention that restricts consumption and thereby reduces the health damage from smoking. Concern for smokers at highest risk--those already addicted who have smoked for many years--would lead to specific subsidies for cessation programs, the deregulation of nicotine replacement markets, and information campaigns emphasizing the dangers of long-term smok- ing. However, in reality, governments do not always aim interventions directly at the sources of market failures themselves, but to particular con- stituencies or population groups affected by those market failures. In the case of the tobacco market, government intervention is often designed to protect children. 90 · Health Economics in Development We turn now to a discussion of the appropriateness of the various avail- able interventions. Choosing `first-best' and `second-best' interventions Government intervention in the tobacco market is most easily justified to deter children and adolescents from smoking and to protect non-smokers. But it is also justified for the purposes of giving adults all the information they need to make an informed choice. Ideally, government interventions should address each identified problem with a specific intervention tailored to solve that particular problem and none other. These may be thought of as first-best interventions. However, a neat one-to-one correspondence between problems and solutions is not always possible, and some interven- tions may have broader effects. We discuss first-best interventions, their effectiveness, and their limitations, first for protecting children, then for correcting the physical and financial costs imposed by smokers on others, and lastly for informing adult smokers. A common theme emerges: the use of taxes, though a second-best and more blunt instrument, is more effective. Protecting children. Several economists have suggested that protection of children is the most compelling economic argument for higher taxes (Warner et al. 1995). Governments can choose to protect children for sev- eral reasons. First, childhood is when nicotine addiction is likely to begin. Second, children are not yet sovereign adults making informed choices, so the principal argument for not intervening does not apply to them as strongly as to adults. Third, there is evidence that the tobacco industry tar- gets children with glamorous advertisements and promotion. Fourth, com- pared with many consumer goods that may appear desirable to children, such as automobiles, cigarettes are generally affordable and accessible: thus the market does not spontaneously protect children from them. Finally, children have no way to become better or safer smokers as they mature, except by quitting. A priori, parents would ideally always be willing and able to protect chil- dren from tobacco themselves. If this happened, there would be little need for governments to duplicate such efforts (Musgrove 1999). Perfect parents, however, are rare. Adults may smoke themselves, thereby modeling this behavior for their children, and, even though few would actually encourage their children to start smoking, they may also fail to educate them about the risks. Parents' responsibilities on the question of smoking are not compara- ble to, say, their responsibilities to ensure their children are immunized. In the latter case, the parent or caregiver has a defined responsibility to The Rationale for Government Intervention · 91 protect the child through a fairly simple action, and the child's lack of infor- mation is irrelevant. The next best public or non-parental interventions would be to try to educate children, restrict advertising and promotion targeted to children, and to restrict their access to tobacco products. As discussed above, infor- mation campaigns have had an important impact on overall declines in smoking in high-income countries. But information campaigns targeted at children are likely to be less effective than those targeted at adults, because children discount the future more, and have difficulty considering conse- quences of today's behavior that may not take effect for three or four decades. Individual youth-centered programs, including school health pro- grams, have often been found ineffective (Reid 1996). For a specific campaign aimed at children, governments would need to ban advertising and promotion of tobacco products in the media that chil- dren are most often exposed to, such as television or radio. Empirical evi- dence (Saffer 2000) suggests that partial bans cause the tobacco industry to shift to other media, including promotional goods (such as free samples), and sponsorship of sports events, which do influence children (Charlton et al. 1997). Finally, efforts to restrict young people's access to tobacco prod- ucts in shops, restaurants, and bars appear to have had mixed success to date, given that the enforcement of bans is difficult. Moreover, youth restrictions have relatively high administrative costs (Reid 1996). In contrast to these measures, there is ample evidence that tax increases are the single most effective policy measure for reducing children's consumption of tobacco products. Young people are more sensitive to price changes than older people. Estimates suggest that a tax increase of $2 per pack in the United States would reduce overall youth smoking by about two-thirds (National Cancer Policy Board 1998). To the extent that low-income and middle-income countries have younger populations than high-income coun- tries, tax increases would be expected to be effective in these countries too. In theory, if cigarette taxes are to be used mainly to deter children and adolescents from smoking, then the tax on children should be higher than any tax on adults. Such differential tax treatment would, however, be virtu- ally impossible to implement. Yet a uniform rate for children and adults, the practical option, would impose a burden on adults. Societies may neverthe- less consider that it is justifiable to impose this burden on adults in order to protect children. Moreover, if adults reduce their cigarette consumption, children may smoke less, given evidence that children's propensity to smoke is influenced by whether their parents, and other adult role-models, smoke (Murray et al. 1983). 92 · Health Economics in Development Physical costs imposed on non-smokers. Governments can choose to protect non- smokers from the health effects of exposure to environmental tobacco smoke, including the effects on children and babies born to smoking parents. The externalities of maternal smoking for infants are less clear than for other non- smokers exposed to others' smoke, at least where mothers are assumed to have rights over fetuses, including the right to submit them to risks. However, the literature on the attitudes of pregnant women to their own health and that of their fetuses suggests that those who are informed about healthy behaviors are more likely to act to protect their fetuses' health (Charlton 1996). Costs to non-smokers' health would appear, a priori, to be easily reduced through bans on public and workplace smoking. These `clean-air' restric- tions have the advantage that they limit the conditions under which people can smoke, without directly addressing the choice of whether to smoke. It should be noted that direct physical externalities do not by themselves justify widespread government interventions, such as advertising and promotion bans, and tax increases, since what matters is not how much people smoke, but whether others are exposed to tobacco smoke. As discussed by Woollery et al. (2000), restrictions in high-income countries on smoking in public places and private workplaces reduce both smoking prevalence and average daily cigarette consumption. Data from developing countries are much less complete, but experience from South Africa suggests that restrictions do reduce smoking (Van der Merwe 1998). Such restrictions are clearly weak- ened where there is a lack of enforcement, or a reliance on self enforcement. However, a more significant problem with this approach is that the vast majority of exposure to environmental tobacco smoke is in homes, and this is where children are also more likely to be exposed. (Mannino et al. 1996; NCI 1999). In contrast to clean-air restrictions, tax increases, by significantly reducing smoking in all settings, could lower this cost to children. Financial costs borne by non-smokers would, a priori, be best reduced through adjusted risk premiums on health services or pension services. Financial costs could be calculated over short intervals, but lifetime medical costs for today's young smokers are more unpredictable. Private insurance markets sometimes include such price differentials, without requiring regu- lation; publicly-financed insurance seldom or never does. As the admini- strative costs for adjusting risk premiums are high, a less precise but more efficient method would be to simply tax cigarettes at the source. Note that in contrast to physical externalities, financial externalities would justify such general consumption-reducing measures, since what matters is how much people smoke rather than where they do it. The Rationale for Government Intervention · 93 Giving adult smokers information. Governments can use a number of meas- ures to protect adult smokers' health by inducing them to quit or to smoke less, but this most directly conflicts with the assumption of consumer sov- ereignty, except in the case of smokers who want to quit but find it difficult because they are already addicted. Public policy responses include informa- tion about the health risks, subsidization of cessation programs and tax increases. Only the last of these conflicts with permitting individuals to make risky decisions (such as playing dangerous sports, or associating with dangerous friends) on the assumption that individuals know their risks and bear the costs of their choices. Providing information, and helping individ- ual smokers who want to quit, are not in conflict with the principal of con- sumer sovereignty. Publicly financed information campaigns and research on the health risks of smoking for adults are justified as a `first-best' intervention. As Kenkel and Chen (2000) elaborate, such information has had a powerful impact on smoking in high-income countries, although the effects take time to appear. Statutory warnings on tobacco products and regulations on tar and nicotine content are also common throughout the world, but few countries use strong and varied warning labels that convey meaning- ful information on the hazards of smoking (WHO 1997). An extension of information measures are bans on advertising and promotion. Such bans can help smokers to quit or to avoid starting again (USDHHS 1990). As discussed above, historically the tobacco industry has used advertising to make misleading claims about the health risks. Thus, bans on advertising and promotion are justified as a more intrusive but effec- tive intervention. Governments may also deregulate nicotine replacement, finance, or provide cessation advice, or even subsidize cessation treatment. As dis- cussed by Novotny et al. (2000) and Gajalakshmi et al. (2000), an indi- vidual's risk of premature mortality drops sharply on quitting, especially at younger ages (Doll et al. 1994). Note that nicotine replacement prod- ucts are not public goods, and are in fact provided by the private mar- ket: smokers wanting to quit can buy private cessation-help programs and nicotine-delivering patches to ease withdrawal. The argument for public intervention is only that the private market's response may be sub-optimal, partly due to regulation that restricts the public's access to cessation aids. Taxation is also an effective intervention. Cigarettes are taxed in nearly all countries, sometimes heavily, but mainly because of the administrative 94 · Health Economics in Development ease of collecting tobacco taxes and the relatively inelastic demand. Adults are less price-responsive than children to increases in tobacco tax. The economic costs of intervening Given that the effective interventions do not neatly correspond to the mar- ket failures they were designed to correct, an important consideration is whether they also generate further economic costs that may be worse than the original market failure. This specifically applies to taxes, given that they are the most blunt, and also most effective, measure to protect children. Below we discuss the key economic costs of intervening, including the costs of foregone pleasure from smoking. Unfortunately, there are few empirical studies of the economic costs of intervening (Warner 1997). We focus on the conceptual framework of costs from various interventions, emphasizing the costs to individuals. We do not discuss costs to producers. Estimates by Peck et al. (2000) suggest that consumer satisfaction is the lion's share of any plau- sible estimate of benefits from smoking, with producers' benefits being much smaller. Ranson et al. (2000) provide estimates of cost-effectiveness from the perspective of the public sector. Control measures would cause regular smok- ers to forego the pleasure of smoking, or incur the costs of quitting, or both. A priori, this loss of consumer surplus would appear to be the same as it would be for bread or any other consumer good. However, tobacco is not a typical consumer good with typical benefits. For the addicted smoker who regrets smoking and expresses a desire to quit, the benefits of smoking are largely the avoidance of the costs of withdrawal. If tobacco control measures reduce individual smokers' consumption, those smokers will face significant withdrawal costs. Furthermore, the costs would differ between current smok- ers and potential smokers who have not yet begun. Clean-air restrictions impose costs on smokers by reducing their oppor- tunities to consume cigarettes, or by forcing them outdoors to smoke, rais- ing the time and discomfort associated with smoking, or by imposing fines for smoking in restricted areas. Such restrictions raise the individual's costs relative to his or her benefits, and prompt some smokers to quit or cut back their consumption. For non-smokers, however, restrictions on smoking in public places will bring welfare gains. Given that most regular smokers express a desire to quit but few are successful on their own, it seems likely that the perceived costs of quitting are greater than the perceived costs of continuing to smoke, such as damage to health. By making the costs of con- tinued smoking greater than the costs of withdrawal, higher taxes can The Rationale for Government Intervention · 95 induce some smokers to quit. However, smokers who quit or cut back would face withdrawal costs from higher taxes. The extent of the loss depends on levels of tax already paid, price responsiveness, and other fac- tors (see Chaloupka and Warner, 2000, for a related discussion on the dis- tributional impacts of taxes). In considering economic costs to smokers, it is important to distinguish between regular smokers and others. For children and adolescents who are either beginners or merely potential smokers, the costs of avoiding tobacco are likely to be less severe, since addiction may not yet have taken hold and, therefore, withdrawal costs are likely to be lower. Other costs may include, for example, reduced acceptance by peers, less satisfaction from the thwarted desire to rebel against parents, and the curtailment of other pleas- ures of smoking. Bans on advertising and promotion might be expected to increase the costs for smokers of obtaining information about their preferred products. However, to the extent that tobacco advertising focuses more on establish- ing brand loyalty among the new smokers it attracts rather than on provid- ing information of value to current smokers, even established adult smokers would suffer little information loss or search costs if advertising and pro- motion were banned (Chapman 1996). In sum, interventions in the smoking market vary by specificity to the market failure and groups most affected. It is obvious that some interventions are fairly specific to particular problems. This is notably the case for bans on smoking in public places, which are intended to control physical externalities. It is also the case for measures to make smokers pay any additional medical costs due to their behavior, which are intended to control financial external- ities. But measures that are aimed at reducing cigarette consumption, rather than controlling where it occurs or who pays the associated costs, are much more general. Taxation and information campaigns are both measures of this type. When it comes to protecting or affecting particular population groups, there is similarly a mixture of more specific and more general connections between an intervention and the group(s) it is meant to affect. Government Interventions to Protect the Poor Aside from government interventions to correct for market failures, inter- vention to protect the poor is a well-recognized government role (Musgrove 1999). Investing in health is one method, but another is to reduce poverty 96 · Health Economics in Development or alleviate its consequences (World Bank 1993). We examine next the issues of how smoking burdens are distributed and the equity implications of some of the interventions analyzed above. In most countries of the world, tobacco consumption is highest among poorer socioeconomic groups, and, accordingly, so is the incidence of tobacco-related disease. Comparison between countries reveals that the poor have higher death rates from smoking-related diseases. Moreover, the poor spend a considerable amount on tobacco as a percentage of their household income, which adversely affects household consumption of items beneficial to children's health (Cohen 1981; World Bank 1993). To some extent, the market failure of incomplete information is more pronounced among the poor (Townsend et al. 1994). Government interventions to reduce the impact of smoking among the poor include taxation, information, and subsidizing access to cessation advice or nicotine replacement therapies (NRT). Differences in the relative importance of different problems imply that the optimal combination of interventions should probably be different for poor and non-poor popula- tions. Several studies suggest that information is less effective in reducing smoking among poor groups than among richer groups (see, for example, USDHHS 1989; Townsend 1998). Smoking prevalence has declined much faster among higher socio-economic groups than among lower groups. The provision of information (such as mass information campaigns and warning labels), and bans on advertising and promotion are justified on efficiency grounds. There is little doubt, however, that the poor would use such infor- mation less, or less quickly, than would the rich. Another strategy would be to finance or provide cessation advice and cessation aids to help the poor quit smoking if they could not afford to pay for them (Musgrove 1999), pro- vided the effects justify the costs. Delivering these services may be costly or difficult, however, since the poor tend to have less access to basic health services than the rich, and the costs of expanding these services to reach the poor might be considerable. In contrast to information, tax increases on tobacco reduce consumption more among the poor and less educated than among the rich and more edu- cated. Evidence from the United Kingdom and the United States (CDC 1998; Townsend 1998; Chaloupka 1991) suggests that price elasticities in the lowest income groups are significantly higher than in the highest income groups. Tobacco taxation would thus narrow the difference in con- sumption between rich and poor (Warner et al. 1995). In high-income countries, the poor usually spend a larger share of their incomes on tobacco The Rationale for Government Intervention · 97 than do the rich. Thus, a tax on tobacco is necessarily regressive among those who continue to smoke. Whether the overall effect of tax increases is regres- sive, depends on what share of each group, poor and non-poor, would react to the higher price by quitting. If more of the poor quit, then the tax effect could even be progressive. Tobacco taxes, like any other single tax, need to work within the goal of ensuring that the entire system of tax and expendi- ture is proportional or progressive (Townsend 1998; Chaloupka and Warner, in press). Studies of tobacco taxation in the United States and the United Kingdom suggest that tax increases are less regressive than pre- sumed, and may even be progressive. In contrast to the taxation of other goods, when the poor reduce their consumption of tobacco they gain a health benefit in return for the tax burden they continue to pay. Finally, the poor may benefit in another way from increased tobacco taxes, if health and social services are targeted to the poor and financed by those taxes (Saxenian and McGreevey 1996; WHO 1999). It might be argued that taxes and other tobacco control measures would impose bigger costs on poor individuals. But if this is true for tobacco, it is not unique in public health. Compliance with many health interventions, such as child immunization or family planning, is often more costly for poor households. For example, poor families may have to walk longer distances to clinics than rich families and may lose income in the process. Yet health officials do not hesitate to argue that the health benefits of most interven- tions, such as immunization, are worth the cost, provided the costs do not rise so high that poor individuals are deterred from using services. In summary, the fact that the poor devote relatively more of their income to tobacco does not provide any strong equity-based argument against the tobacco control measures analyzed here. Conclusion We have described specific failures in the tobacco market: first, inadequate information about the health risks of smoking; second, inadequate informa- tion about the risks of addiction (and particularly the youthful onset of use of an addictive product); and, third, the external costs of smoking. We argue that because of these market failures, government intervention is justified on economic grounds. However, the interventions themselves are often non-precise and impose costs on even informed adult smokers. What then do these findings imply for public policy? 98 · Health Economics in Development First, the public health arguments and the economic arguments for tobacco control differ on goals. Public health goals would, rationally, be to eradicate smoking if possible, given that tobacco hazards increase with increasing exposure and overwhelm any possible beneficial effects on health. In contrast, the economic arguments suggest that the socially opti- mal level of consumption of tobacco would not be zero. Ideally in economic terms, children would not smoke, but adults who knew their risks and bore their costs entirely themselves could smoke (Warner 1998). Such a situation would involve considerably less smoking than at present, but would stop well short of eradication. Preventing children from smoking could, in theory, eventually lead to the epidemic disappearing. In reality, slightly older cohorts may take up smoking, and it is unlikely that the recruitment of new smokers would cease. Several of the interventions dis- cussed here, particularly those designed to prevent smoking in youth, pro- tect non-smokers from externalities, and leave smokers better informed. However, a major problem for the `economically optimal' view of smok- ing is the fact that nicotine is addictive. This undermines the consumer- sovereignty argument against intervention, because all evidence suggests that the conditions for a rational choice to become addicted are not met, and the addicted smoker is to some degree a different person from the one who decided to start smoking. If addiction is taken into account, a `middle- ground' rationale that is justifiable by both economic and public health arguments becomes feasible. It still falls short of eradication, but is more realistic and justifiable than a purely economics-led view that defines adult consumers as rational and informed. The economic rationale for interven- tion described here largely involves information and regulation, and not direct public finance or the provision of private goods, except perhaps to the poor. As such, it leaves much room for private choice. As with other areas of public policy, governments have to make choices, drawing here on economics, epidemiology, and public health. Even limited reductions in the prevalence of smoking, achieved as the result of interven- tions to correct market failures, would, by any measure, constitute an enor- mous public health victory, avoiding millions of deaths per year. References Atkinson, A. B. and Townsend, J. L. (1977). Economic aspects of reduced smoking. Lancet, 2(8036), 492­5. The Rationale for Government Intervention · 99 Bachman, J. G.,Wadsworth, K. N., O'Malley, P. M., Johnston, L. D., and Schulenberg, J. (1997). Smoking, drinking, and drug use in young adulthood: The impacts of new freedoms and new responsibilities. Mahwah, NJ: Lawrence Erlbaum Associates. Barendregt, J. J., Bonneux, L., and van der Maas, P. J. (1997). The health care costs of smoking. New England Journal of Medicine, 337(15), 1052­7. Becker, G. S. and Murphy, K. M. (1988). A theory of rational addiction. Journal of Political Economy, 96(4), 675­700. Becker, G. S., Grossman, M., and Murphy, K. M. (1994). An empirical analysis of cigarette addiction. American Economic Review, 84(3), 396­418. Bero, L. A., Glantz, S. A., and Rennie, D. (1994). Publication bias and public health policy on environmental tobacco smoke. JAMA, 13, 133­6. Bos, E. R., Hon, V., Maeda, A., Chellaraj, G., and Preker, A. (1999). Health, Nutrition, and Population Indicators: A Statistical Handbook. Washington, DC: World Bank. California Environmental Protection Agency (1997). Health Effects of Exposure to Environmental Tobacco Smoke: Final Report. Office of Environmental Health Hazard Assessment (OEHHA). http://www.oehha.org/scientific/ets/ finalets.htm Centers for Disease Control and Prevention (CDC) (1998). Response to increases in cigarette prices by race/ethnicity, income, and age groups­United States, 1976­1993. Morbidity and Mortality Weekly Report, 47(29), 405­9. Chaloupka, F. J. (1991). Rational addictive behavior and cigarette smoking. Journal of Political Economy, 99(4), 722­42. Chaloupka, F. J. and K. E. Warner (2000). The economics of smoking. In The Handbook of Health Economics (ed. J. Newhouse and A. Culyer). Amsterdam: North Holland. (In press.) Chapman, S. (1996). The ethics of tobacco advertising and advertising bans. Br. Med. Bull., 52(1), 121­31. Charlton, A. (1996). Children and smoking: the family circle. Br. Med. Bull., 52(1), 90­107. Charlton, A., While, D., and Kelly, S. (1997). Boys smoking and cigarette-brand- sponsored motor racing. Lancet, 350(9089), 1474. Chinese Academy of Preventive Medicine (1997). Smoking in China: 1996 National Prevalence Survey of Smoking Pattern. Beijing: China Science and Technology Press. Cohen, N. (1981). Smoking, health, and survival: prospects in Bangladesh. Lancet, 1(8229), 1090­3. Collins, D. and Lapsley, H. (1998). Estimating and disaggregating the social costs of tobacco. In The Economics of Tobacco Control: Towards an Optimal Policy Mix (ed. I. Abedian, R. Van der Merwe, N. Wilkins and P. Jha), pp. 155­78. Cape Town, South Africa: Applied Fiscal Research Centre, University of Cape Town. 100 · Health Economics in Development Doll, R., Peto, R., Wheatley, K., Gray, R., and Sutherland, I. (1994). Mortality in relation to smoking: 40 years observations on male British doctors. British Medical Journal, 309(6959), 901­11. The Economist (1995). An anti-smoking wheeze: Washington needs a sensible all- drugs policy, not a `war on teenage smoking'. 19 August, pp. 14­15. The Economist (1997). Tobacco and tolerance. 20 December, pp. 59­61. Ensor, T. (1992). Regulating tobacco consumption in developing countries. Health Policy and Planning, 7, 375­81. Environmental Protection Agency (1992). Respiratory Health Effects of Passive Smoking: Lung Cancer and Other Disorders. EPA, Office of Research and Development, Office of Air and Radiation. EPA/600/6­90/006F. Glantz, S. A., Barnes, D. E., Bero, L., Hanauer, P., and Slade, J. (1995). Looking through a keyhole at the tobacco industry. The Brown and Williamson docu- ments. Journal of the American Medical Association, 274(3), 219­24. Gorovitz, E., Mosher, J., and Pertschuk, M. (1998). Pre-emption or prevention?: lessons from efforts to control firearms, alcohol, and tobacco. Journal of Public Health Policy, 19(1), 36­50. Harris, J. E. (1994). A Working Model for Predicting the Consumption and Revenue Impacts of Large Increases in the U.S. Federal Cigarette Excise Tax. Working paper no. 4803. Cambridge (MA): National Bureau of Economic Research. Hodgson, T. A. (1992). Cigarette smoking and lifetime medical expenditures. Milbank Quarterly, 70(1), 81­125. Institute of Medicine (1994). Growing Up Tobacco Free. Washington, D.C.: National Academy Press. James, E. (1994). Averting the Old Age Crisis: Policies to Protect the Old and Promote Growth. Oxford and New York: World Bank and Oxford University Press. Kessler, D. A., Barnett, P. S., Witt, A., Zeller, M. R., Mande, J. R., and Schultz, W. B. (1997). The legal and scientific basis of FDA's assertion of jurisdiction over cigarettes and smokeless tobacco. Journal of the American Medical Association, 277, 405­9. Krutilla, J. V. (1967). Conservations reconsidered. American Economic Review, 57, 776­86. Leu, R. E. and Schaub, T. (1983). Does smoking increase medical care expenditure? Social Science and Medicine, 17(23), 1907­14. Levshin, V. and Droggachih, V. (1999). Knowledge and Education Regarding Smoking Among Moscow Teenagers. Paper presented at the workshop on Tobacco Control in Central and Eastern Europe. Las Palmas de Gran Canaria. February 26. Manning, W. G. (1989). The taxes of sin: do smokers and drinkers pay their way? Journal of the American Medical Association, 261(11), 1604­09. Manning, W. G., Keeler, E. B., Newhouse, J. P., Sloss, E. M., and Wasserman, J. (1991). The Costs of Poor Health Habits. Cambridge, Mass.: Harvard University Press. The Rationale for Government Intervention · 101 Mannino, D. M., Siegel, M., Husten, C., Rose, D., and Etzel, R. (1996). Environmental tobacco smoke exposure and health effects in children: results from the 1991 National Health Interview Survey. Tobacco Control, 5(1), 13­18. Murray, C. J. L. and Lopez, A. D. (eds.) (1996). The Global Burden of Disease: a Comprehensive Assessment of Mortality and Disability from Diseases, Injuries, and Risk Factors in 1990 and Projected to 2020. Cambridge, Mass.: Harvard School of Public Health. Murray, M., Swan, A. V., Johnson, M. R., and Bewley, B. R. (1983). Some factors associated with increased risk of smoking by children. Journal of Child Psychology and Psychiatry, 24(2), 223­32. Musgrove P. Public spending on health care: how are different criteria related? Health Policy 1999, 47(3):207­23. National Cancer Institute (NCI) (1999). Health Effects of Exposure to Environmental Tobacco Smoke. The Report of the California Environmental Protection Agency. Smoking and Tobacco Control Monograph no. 10. Bethesda, MD. US Department of Health and Human Services, National Institutes of Health, National Cancer Institute, NIH Pub. No. 99­4645. National Cancer Policy Board (1998). Taking Action to Reduce Tobacco Use. Washington, DC: National Academy Press. O'Malley, P. M., Bachman, J. G., and Johnston, L. D. (1988). Period, age and cohort effects on substance use among young Americans: a decade of change, 1976­86. American Journal of Public Health, 78(10), 1315­21. Orphanides, A., and Zervos, D. (1995). Rational addiction with learning and regret. Journal of Political Economy, 103(4), 739­58. Pekurinen, M. (1991). Economic Aspects of Smoking: Is There a Case for Government Intervention in Finland? Helsinki: Vapk-Publishing. Peto, R. (1986). Influence of dose and duration of smoking on lung cancer rates. In Tobacco: a Major International Health Hazard. (ed. R. Peto, and D. Zaridze), pp. 23­34. International Agency for Research on Cancer, 1986 (IARC Scientific Publications, no. 74). Peto, R., Lopez, A. D., Boreham, J., Thun, M., and Heath, C. Jr. (1994). Mortality from Smoking in Developed Countries 1950­2000. Oxford: Oxford University Press. Peto, R., Chen, Z. M., and Boreham, J. (1999). Tobacco: the growing epidemic. Nature Medicine, 5(1), 15­7. Reid, D. (1996). Tobacco control: overview. British Medical Bulletin, 52(1), 108­20. Saxenian, H. and McGreevey, W. P. (1996). China: Issues and Options in Health Financing. World Bank Report No. 15278-CHA,Washington, DC. Suranovic, S. M., Goldfarb, R. S., and Leonard, T. C. (1999). An economic theory of cigarette addiction. Journal of Health Economics, 18, 1­29. Sweda, E. L. Jr. and Daynard, R. A. (1996). Tobacco industry tactics. British Medical Bulletin, 52(1), 183­92. 102 · Health Economics in Development Townsend, J., Roderick, P., and Cooper, J. (1994). Cigarette smoking by socioeco- nomic group, sex, and age: effects of price, income, and health publicity. British Medical Journal, 309(6959), 923­27. Townsend (1998). The role of taxation policy in tobacco control. In The Economics of Tobacco Control (ed. I. Abedian, R. van der Merwe, N. Wilkins, and P. Jha), pp. 85­101. Cape Town, South Africa: Applied Fiscal Research Centre, University of Cape Town. US Department of Health and Human Services (1986). The Health Consequences of Smoking For Women. Rockville, Maryland: US Department of Health and Human Services, Public Health Service, Office of the Assistant Secretary for Health, Office on Smoking and Health. US Department of Health and Human Services (1989). Reducing the Health Consequences of Smoking: 25 Years of Progress. A Report of the Surgeon General. Rockville, Maryland: US Department of Health and Human Services, Public Health Service, Centers for Disease Control, Center for Chronic Disease Prevention and Health Promotion, Office on Smoking and Health. DHHS Publication No. (CDC)89­8411. US Department of Health and Human Services (1990). The Health Benefits of Smoking Cessation: A Report of the Surgeon General. Rockville, Maryland: US Department of Health and Human Services, Public Health Service, Centers for Disease Control, Center for Chronic Disease Prevention and Health Promotion, Office on Smoking and Health. DHHS Publication No. (CDC) 90­8416. Van der Merwe, R. (1998). The economics of tobacco control in South Africa. In The Economics of Tobacco Control (ed. I. Abedian, R. van der Merwe, N. Wilkins, and P. Jha), pp. 251­71. Cape Town, South Africa: Applied Fiscal Research Centre, University of Cape Town. Viscusi,W. K. (1995). Cigarette taxation and the social consequences of smoking. In Tax Policy and the Economy. (ed. J. M. Poterba). Cambridge, MA, MIT Press. 9, 51­101. Wald, N. J. and Hackshaw, A. K. (1996). Cigarette smoking: an epidemiological overview. British Medical Bulletin, 52(1), 3­11. Warner, K. E., Goldenhar, L. M., and McLaughlin, C. G. (1992). Cigarette adver- tising and magazine coverage of the hazards of smoking. A statistical analysis. New England Journal of Medicine, 326, 305­9. Warner, K. E. (1997). Cost-effectiveness of smoking cessation therapies: interpreta- tion of the evidence and implications for coverage. PharmacoEconomics, 11, 538­49. Warner, K. E. (1998). The economics of tobacco and health: an overview. In The Economics of Tobacco Control (ed. I. Abedian, R. van der Merwe, N. Wilkins and P. Jha), pp. 55­75. Cape Town, South Africa: Applied Fiscal Research Centre, University of Cape Town. The Rationale for Government Intervention · 103 Warner, K. E., Chaloupka, F. J., Cook, P. J., Manning W. G., Newhouse, J. P., Novotny, T. E. et al. (1995). Criteria for determining an optimal cigarette tax: the economist's perspective. Tobacco Control, 4, 80­6. Weinstein, N. D. (1998). Accuracy of smokers' risk perceptions. Annals of Behavioral Medicine, 20(2), 135­40. World Bank (1992). China: Long-term Issues and Options in the Health Transition. Washington, DC. World Bank (1993). The World Development Report 1993: Investing in Health. New York: Oxford University Press. World Health Organization (1997). Tobacco or Health: a Global Status Report. Geneva, Switzerland. World Health Organization (1999). The World Health Report 1999: Making a differ- ence. Geneva, Switzerland. Zatonski, W. (1996). Evolution of Health in Poland Since 1988. Warsaw: Marie Skeodowska-Curie Cancer Center and Institute of Oncology, Department of Epidemiology and Cancer Prevention. CHAPTER 4 Measurement of Equity in Health Equity, according to the Plan of Action of the Pan American Health Orga- nization (1), is one of three essential qualities of a system of health services, efficiency and effectiveness being the other two. None of these concepts is simple to define or measure, so it is not surprising that no indicators of progress towards them have been adopted, in contrast to the situation for a number of more specific goals such as high life expectancy or low infant mortality. However, a number of simple indicators can be used to tell some- thing about equity, even if there is no single measure of it. This article dis- cusses the logic behind various such indicators, drawing on recent data from Peru for empirical illustration. Equity as Equality of Treatment The fundamental idea of equity is that of equal treatment for all the popula- tion. The intention is to assure good health for all, and as far as possible, equally good health, to be pursued through preventive or curative treatment. Equity cannot simply be identified with equality in general, because of dif- ferences in needs, but it can be judged by considering certain kinds of equal- ity. The idea may be "to provide 100% of the population with access to health services" (1), but the crucial notion is that whatever the level of access, it should be the same for all. Inequity results from differences in the ability to obtain health care, whatever the reasons may be, that prevent some people and not others from getting medical assistance (2). Since illness and accident are randomly and non-uniformly distributed, it cannot be expected that Reprinted, with permission, from World Health Statistics Quarterly 39 (4), 1986. 105 106 · Health Economics in Development everyone will see a doctor equally often, or that spending per person on health care will be the same for all groups in the population. Treatment and resources should go where they are most needed. It is to be expected, how- ever, that if the health-care system is equitable, certain probabilities will be equal across population groups for a given set of health problems. Probabilities of illness, treatment and recovery This idea is shown schematically in Fig. 4.1. Within a given interval of time, a person may or may not become sick or hurt; he or she may receive treat- ment or may not; and there may or may not be a cure as the result of treat- ment, or a recovery in the absence of treatment. The complete passage from an initial state to a final state (health, illness, disability or death) can be described by just four probabilities, indicated by the bold arrows in Fig. 4.1. These are: P(S) the unconditional probability of needing medical care P(T/S) the conditional probability of receiving treatment, given the need for it P(C/T) the conditional probability of being cured by treatment P(R/T*) the conditional probability of recovering without treatment If the chance of being treated when there is no need for it--shown by the dashed arrow--is excluded from consideration, then every other probability, including those of final good health and ill health, P(H) and P(H*), is just some combination of these four basic probabilities (3). Figure 4.1 Schematic Stages of Illness or Accident, Treatment and Outcome S(Yes) T(Yes) C*(No) H*(Ill health, disability or R*(No) death) Initial Symptom, Treatment, Cure (C) or Final state sickness, consultation untreated state accident recovery (R) C(Yes) T*(No) R(Yes) S*(No) H (Health) Measurement of Equity in Health · 107 Different Dimensions of Equity This scheme simplifies a great deal, but it serves to emphasize several issues to consider in attempting to judge the equity of a health service system. Among these are: Which stage of the sequence is analyzed. The system might, for example, provide roughly the same chance of being cured to all patients who receive treatment, but be inequitable in reaching some people for treatment much more readily than others. The difference between prevention, which acts on P(S), and curative treatment, which involves P(T/S) and P(C/T). Some differences in the likelihood of getting sick or hurt should not be considered inequitable, since they are associated with age or other risk factors largely outside the control of the health care system. Other differences in P(S)--for example, in the chance of getting diphtheria or poliomyelitis--are inequitable because prevention is within the power of the system and can be applied equally to virtually the entire population. The particular condition, illness or need, studied. A health care system may provide everyone with the same chance of emergency care after a motor- vehicle accident, but maintain marked inequalities in the treatment of can- cer or tuberculosis. The level or quality of treatment. Does equity require the same type of care for everyone who is treated for a particular condition? Or should the sys- tem be regarded as equitable if every patient gets at least a minimal ade- quate level of care, even though some receive more elaborate, prolonged or expensive treatment (3)? A similar issue arises in analyzing the distribution of income: does equity require equality of incomes, or is what matters a decent minimum income for everyone? This issue becomes particularly important when analyzing expenditure on medical care. The interaction of supply (physical availability of services) and demand (individual perception of need) in determining who does and who does not get treatment. For example, is it fair to describe a system as inequitable if it provides relatively lit- tle care to a cultural group which is more stoic than average, or more likely to rely on home remedies or traditional healers? How much responsibility falls to the government to change such cultural patterns so as to relate medical need and individual demand in the same way for the whole population? All these questions show clearly why there cannot be a single measure of how equitable a health care system is: the same system may be quite fair by 108 · Health Economics in Development some indicators, and grossly inequitable according to others. It is no more possible to judge a country's health services as to equity by just one number than it is to summarize the population's health status in one indicator. Finally, there is one important issue which is not clear from Fig. 4.1. In order to estimate and compare probabilities, the population must be divided into groups on some basis. How this is done may greatly affect the appar- ent inequality. Just as income typically is distributed more unequally among educational classes than among geographical regions, the health care system may look much more equitable in one dimension than in another. An ade- quate evaluation will probably require that the population be divided in more than one way, for example, by socioeconomic criteria as well as by geographical location. Several of the issues discussed above can be illustrated by recent data from Peru. These include the usual administrative data collected by the Ministry of Health and analyzed by the country's Central Reserve Bank1, household level information obtained in the 1984 National Health and Nutrition Sur- vey, and analyses conducted as part of the 1985-1986 National Health Sector Analysis by various public and private agencies in Peru, with financing from the United States Agency for International Development (USAID)2. Not all the possible measures of equity can be computed--there are, for example, no data on the probability of cure or of spontaneous recovery-- but a variety of indicators is available. Distribution of health-related resources In comparing different countries, it is common to refer to the share of the population which has access to safe water or sewerage connections, or to compute the ratio of population to such health care resources as physicians or hospital beds. The same analysis can be carried out within a country, as illustrated in Table 4.1. This comparison shows that physicians are very highly concentrated, especially in the department of Lima, which has just over a quarter of the population but two-thirds of all the physicians in Peru. Nurses are much more equally distributed, although still tending to con- centrate in the same departments as physicians. Hospital beds are slightly more equally distributed, and access to safe water still more so. In general, expensive resources seem to be more concentrated than cheaper ones (physicians vs. nurses, and sanitation vs. water supply). These rather easy indicators, which need no information about the pop- ulation except place of residence, suffer three important limitations as Measurement of Equity in Health · 109 Table 4.1 Measures Related to Equity in Health Care, Peru, 1982: Health Care Resources and Sanitation Services (Percentages of National Total) RESOURCES DWELLINGS HOSPITAL WITH WITH DEPARTMENT POPULATION PHYSICIANS NURSES BEDS TOTAL WATER SEWERAGE Amazonas 1.5 0.0 0.1 0.3 2.4 0.6 0.4 Ancash 4.8 1.2 3.4 2.7 5.1 4.5 3.7 Apurímac 1.9 0.1 0.5 0.7 2.2 0.5 0.2 Arequipa 4.2 5.3 8.2 6.8 4.3 5.7 6.0 Ayacucho 2.9 0.2 1.3 0.9 2.6 1.5 0.7 Cajamarca 6.1 0.5 0.6 0.9 6.3 2.0 1.4 Callao 2.7 6.3 4.5 4.9 2.3 4.4 5.3 Cuzco 4.9 1.0 2.2 2.9 5.4 2.6 2.1 Huancavelica 2.0 0.1 0.4 0.7 2.4 0.6 0.4 Huánuco 2.8 0.6 0.6 1.5 2.9 1.0 0.9 Ica 2.5 2.8 3.8 3.7 2.5 3.1 2.4 Junín 5.0 1.5 4.0 4.5 5.3 4.3 3.1 La Libertad 5.6 4.9 5.5 4.9 5.5 6.5 6.1 Lambayeque 4.0 2.6 5.4 3.8 3.5 4.5 4.3 Lima 28.1 66.6 50.7 48.4 26.4 45.1 52.8 Loreto 2.6 0.8 0.7 1.5 2.2 2.0 1.8 Madre de Diós 0.2 0.1 0.1 0.2 0.2 0.0 0.1 Moquegua 0.6 0.7 0.2 1.1 0.2 0.7 0.7 Pasco 1.2 0.6 1.2 1.7 1.3 0.8 0.6 Piura 6.6 2.3 2.7 3.2 6.1 5.2 4.2 Puno 5.2 0.5 1.5 1.6 6.4 1.5 0.8 San Martín 1.9 0.2 0.5 0.8 1.7 0.9 0.3 Tacna 0.9 0.6 1.3 1.4 0.9 1.5 1.4 Tumbes 0.6 0.2 0.3 0.3 0.5 0.1 0.1 Ucayali 1.2 0.3 0.3 0.6 1.0 0.4 0.3 National total 100.0 100.0 100.0 100.0 100.0 100.0 100.0 Gini coefficient 0.51 0.38 0.34 0.32 0.41 of inequality Source: [Health map of Peru]. Lima, Central Reserve Bank of Peru, 1984. measures of equity. Firstly they are restricted to geographical comparisons; without knowing who actually consults the physicians, one cannot know their distribution according to other dimensions of the population, such as income. Secondly, it is implicitly assumed that needs are uniformly dis- tributed. This is probably a reasonable assumption for sanitation facilities, but is questionable for health needs. The incidence or prevalence of health problems may differ substantially from one region to another, so that an equitable distribution of resources would in fact be unequal, not uniform. Third, it is presumed that the resources or facilities analyzed answer the 110 · Health Economics in Development needs of the population. Again this is a more reasonable assumption for sanitation facilities, although the health effects of safe water may differ considerably depending on the pattern of illness in a region and the hygienic behavior of the population. For health care the assumption is still more questionable: treatment by a physician may not be necessary in many cases; many physicians may be engaged in teaching or research; patients with more difficult problems may travel to other departments for treat- ment; etc. Thus while these measures say something about equality in the distribution of resources, and consequently in expenditure or health care, they should not be assumed to be the only or the best indicators of the equity of the system. Probabilities of need and of treatment A principal weakness of the resource vs. population comparisons in Table 4.1 can be overcome by referring to the morbidity and utilization data obtained from a large sample of the Peruvian population in the 1984 survey, and given in Table 4.2. With one dramatic exception, among the urban population of the country's mountainous central region, the prob- ability of presenting with some illness or symptom was fairly uniform, at Table 4.2 Measures Related to Equity in Health Care in Peru, 1984: Morbidity and Medical Attention PERCENTAGE SEEKING MEDICAL PREVALENCE (%) OF SYMPTOMS ATTENTION WITH SYMPTOMS ALL RESPIRATORY PARASITIC ALL UNDER OVER AREA KINDS DISEASE INFECTION TOTAL AGES 1 YEAR 1-4 YEARS 4 YEARS Coast 34.89 16.11 0.21 12.67 30.86 51.41 30.68 28.64 Urban 34.87 16.34 0.20 13.36 31.54 55.74 32.69 29.96 Lima 36.57 17.68 0.15 14.88 33.49 59.43 35.87 31.62 Slums 37.32 16.85 0.19 14.14 32.05 64.84 32.85 29.78 Rural 35.00 14.38 0.28 7.50 18.76 27.14 16.52 18.71 Mountains 30.05 11.95 0.16 5.53 15.89 24.13 15.47 15.49 Urban 21.50 9.88 0.11 7.38 28.20 41.33 25.63 28.09 Rural 33.75 12.84 0.18 4.73 12.49 19.32 12.33 12.10 Jungle 36.03 12.06 1.78 7.65 18.39 27.33 17.20 17.94 Urban 33.72 11.59 1.36 10.36 25.59 42.63 22.34 24.97 Rural 37.35 12.32 2.02 6.11 14.69 18.42 14.46 14.41 National total 33.31 14.20 0.36 9.60 24.18 38.89 23.83 23.35 Urban 32.51 14.93 0.26 12.15 30.77 52.64 30.90 29.46 Rural 34.62 13.01 0.53 5.45 14.01 20.55 13.51 13.69 Source: National health and nutrition survey, Peru, 1984 (unpublished). Measurement of Equity in Health · 111 about one-third, during the two-week reference period. The uncondi- tional probability of seeking medical care (including a visit to a pharmacy, but excluding home care), however, ranged from almost 15% in Lima down to less than 5% in rural mountainous areas. Consequently, the condi- tional probability P(T/S) ranged from over one-third in Lima to 12.5% in the rural mountainous areas. If equity means equal likelihood of attention-- not necessarily the certainty of attention, since many illnesses and symp- toms do not require more than home care--then the Peruvian health care system shows dramatic geographical inequity. Several other features of these estimates merit consideration. For one thing the "geographical dimension" has several possible meanings. The data in Table 4.2 are presented according to the survey area, in order to bring out urban/rural differences. For Peru as a whole, P(T/S) is more than twice as high in urban areas as it is in rural areas. The results could instead be shown according to the "health regions" of the Ministry of Health, which are used in Table 4.3. Neither of these classifications matches the departmental boundaries used in Table 4.1, although the health regions correspond approximately to departments or combinations of them. Table 4.3 Measures Related to Equity in Health Care in Peru, 1984 Consultations, Hospitalizations and Expenditures (Percentages of National Total) POPULATION MINISTRY OF HEALTH HEALTH WITH PATIENT-RELATED REGION TOTAL SYMPTOMS CONSULTATIONS HOSPITALIZATIONS EXPENDITURES Ancash 4.92 7.26 3.25 7.25 3.57 Arequipa 4.33 2.19 7.63 5.98 6.07 Cajamarca 4.48 5.08 1.57 1.38 1.80 Chiclayo 7.30 7.71 3.62 3.40 2.79 Cuzco 7.00 2.64 3.64 5.70 4.42 Huancayo 7.07 3.95 4.01 5.46 7.30 Huánuco 4.75 4.52 2.67 4.64 2.65 Ica 3.40 2.58 4.04 5.16 4.55 Iquitos 3.01 2.95 1.88 2.84 2.06 Lima 32.12 35.81 52.79 41.25 46.71 Moyobamba 2.00 3.30 2.18 2.20 2.51 Piura 7.44 10.28 4.29 4.81 3.90 Puno 5.09 5.26 1.36 2.13 2.81 Tacna 1.51 1.39 1.880 2.27 3.03 Trujillo 5.59 5.08 5.33 5.51 5.83 National total* 100.0 100.0 100.0 100.0 100.0 *Excluding the Ayacucho health region; percentages have been adjusted to 100% over the rest of the country. 112 · Health Economics in Development Differences by illness and by age. The relative uniformity of total morbidity hides larger variations for specific illnesses or symptoms. Respiratory dis- ease affected from under 10% to over 17% of the population, while para- sitic diseases, although unimportant overall, were extremely frequent in the jungle, and especially in rural areas. Conditional probabilities of treatment can be calculated separately for different causes such as these. Table 4.2 also shows that P(T/S) varies considerably by age, because infants under 1 year have a much higher likelihood of receiving medical care than do older chil- dren. This presumably reflects the greater risk of a serious illness in the first year of life. Inequality, however, appears to be worse where treatment of infants is concerned: for example, the urban/rural differential in the proba- bility of care is greater for infants than for other age groups. Thus higher overall coverage or utilization of services need not imply greater equity or less inequality. Equity in vaccination coverage. As the comparison of infants and older children emphasizes, illnesses and symptoms differ in danger or severity, and this may account for much of the variation in the likelihood of obtaining medical care. Not all such differences can be interpreted as indications of inequity. In order to control for this source of variation, it may be advisable to study equity with respect to a single well-defined need or condition. This is done in Table 4.4, for vaccination against the six target diseases of the Expanded Programme on Immunization, for all children under 5: first by type of vaccine and then according to the schooling of the child's mother. (This is one of the few analyses so far pre- pared using a socioeconomic classification; eventually the survey data will allow classification by income, or a proxy variable for it, as well as by edu- cational level.) Comparison across the four types of vaccines shows clearly that as the total coverage drops, the inequality of coverage increases. For BCG vac- cine, the urban/rural differential is less than 2:1, reaching about 2:1 for measles vaccination and close to 4:1 for protection against poliomyelitis, diphtheria, whooping cough and tetanus. This national pattern is repeated within the mountain and jungle regions of the country. The dif- ferential is considerably less in the relatively favoured coastal region. Of course, as the national coverage approaches 100%, inequality necessarily disappears; but the great inequity occurring in rural areas when coverage is low reflects the very poor protection against these diseases in those areas. Measurement of Equity in Health · 113 Table 4.4 Measures Related to Equity In Health Care in Peru, 1984: Vaccination Coverage (%) of Children under 5. BY TYPE OF VACCINE AREA BCG POLIO DPT MEASLES ALL TYPES Coast 84.02 54.29 52.24 67.08 46.43 Urban 87.40 58.07 56.01 70.05 50.14 Lima 90.31 62.49 60.22 73.96 53.88 Slums 91.24 58.35 55.50 74.42 50.02 Rural 62.44 30.14 28.13 48.10 22.71 Mountains 49.99 18.31 17.94 39.31 12.99 Urban 77.16 41.56 41.86 59.70 30.29 Rural 39.42 10.11 9.51 32.13 6.89 Jungle 57.06 32.42 31.23 50.44 27.63 Urban 84.20 61.46 60.01 74.02 55.82 Rural 43.16 17.48 16.42 38.30 13.13 National total 67.62 37.81 36.51 54.40 31.31 Urban 85.36 55.47 53.87 68.56 47.14 Rural 44.34 14.66 13.75 35.82 10.57 BY MOTHER'S EDUCATION NONE SOME PRIMARY COMPLETE PRIMARY SECONDARY HIGHER Coast 31.59 41.97 50.60 60.23 69.66 Urban 37.19 46.70 53.29 61.28 69.56 Lima 37.37 59.16 55.10 62.67 69.45 Slums 38.29 66.99 59.30 65.23 62.01 Rural 23.47 27.76 32.70 42.38 81.09 Mountains 10.778 14.18 25.68 35.81 51.39 Urban 27.02 30.51 30.57 39.30 50.37 Rural 9.32 10.70 23.05 27.88 58.50 Jungle 16.42 23.94 40.18 60.52 79.70 Urban 36.08 44.58 56.08 66.74 83.43 Rural 15.04 17.33 29.29 43.85 50.00 National total 16.17 26.62 42.23 55.73 65.28 Urban 34.36 43.46 50.23 58.49 65.47 Rural 11.82 15.05 26.79 36.02 61.01 Source: National health and nutrition survey. Peru, 1984 (unpublished). A somewhat similar pattern emerges when vaccination coverage is com- pared across educational levels. As the educational level rises, so does vac- cination coverage, and the geographical inequality diminishes. It is also true that the lower the coverage in a given geographical area, the greater is the inequality across educational groups. Thus the most marked socioeco- nomic inequity occurs in the rural mountainous areas, where overall 114 · Health Economics in Development protection is extremely low, and the least inequity is found in Lima. Chil- dren of less educated mothers, presumably in poor families, are the last to be reached, except for urban slums. This pattern probably reflects demand for vaccination on the part of parents, and not simply differences in avail- ability: better educated mothers are more likely to bring their children to be vaccinated, without requiring a public campaign to persuade them to do so, and are better able to pay a private physician for service if necessary. While the inhabitants of the slums of Lima may be very poor, Tables 4.2 and 4.4 also show that they are not a particularly underprivileged group in access to health care. Summary measures of inequality The emphasis in Tables 4.1, 4.2 and 4.4 is on all the differences among geo- graphical regions or socioeconomic groups, since any of these differences may be particularly important for the interpretation of how equitable or inequitable a health care system is, and where its inequities are concen- trated. It is also possible to calculate summary measures of inequality, and to use them to form overall judgments about whether one distribution represents a more or less equitable situation than another (3). All such measures, however, discard information, and the way a particular statistic summarizes a distribution corresponds to assumptions that may or may not be appropriate when judging equity. Because it has an easy geometric inter- pretation, the Gini coefficient is often used for this purpose; it is applied, for example, in the Central Reserve Bank's analysis for Peru. The coefficient is calculated by first drawing the Lorenz curve, which relates the cumulated population across groups to the cumulated resource or utilization measure studied, ranking the groups from lowest to highest values of resources per capita. The Lorenz curve corresponding to the departmental distribution of physicians in Peru, calculated from Table 4.1, is shown in Fig. 4.2 It indi- cates, for example, that the departments where physicians are most scarce have 70% of the country's population but only about 28% of its physicians. The diagonal line corresponds to perfect equality in the ratio of physicians to population. The Gini coefficient is the ratio of the area between the diag- onal and the Lorenz curve, to the total area under the diagonal, a measure which increases from 0 to 1 as inequality increases. As Table 4.1 shows, this measure is higher for physicians than for any of the other health-related resources considered. The Lorenz curve can also be used to represent inequality or inequity without calculating the Gini coefficient or any other summary statistic. Measurement of Equity in Health · 115 Figure 4.2 Lorenz Curves of Inequality of the Distribution of Physicians Relative to Population (by Department) and of Ministry of Health Patient- related Expenditure Relative to Population with Symptoms (by Health Region), Peru, 1982 Cumulated percentage of physicians and of expenditure 100 90 80 70 60 50 40 Physicians per capita 30 Expenditure per person 20 with symptoms 10 0 0 10 20 30 40 50 60 70 80 90 100 Cumulated percentage of population (total, or with symptoms) Financial Measures of Equity Physicians, nurses, drugs, vaccines and other resources and supplies are used to produce health services, but none alone is an adequate measure of resources used in health care. The best overall measure of the resources ded- icated to health services is the cost of producing those services. This raises two equity-related questions. The first concerns the equality of expenditure across population groups; the second concerns the relation between the cost of providing health services and the contributions different groups make to 116 · Health Economics in Development that cost, either by direct payment (fees for service) or by taxation. Both questions are particularly important for the equity of ministry of health serv- ices, which are typically financed largely from general government revenues and which are usually intended to cover the population groups too poor to pay for private medical care, and unlikely to be protected by the medical services of the social security system. Equity in the distribution of expenditure Variations in expenditure per capita on health care can be separated into dif- ferences in the likelihood of being sick or needing care, in the likelihood of receiving care, and in the cost per patient of the care given. Estimates of total cost by health region in Peru have been made for the years 1982­1984; these can be separated into costs related to individual patients, and those attributable to preventive and maintenance activities. Costs for the former can then be compared to total population, to the population presenting ill- ness or symptoms and needing care, or to the population which actually receives care. Some of these estimates for 1984 appear in Table 4.3, classi- fied by health region. The first of these measures--cost per capita--is often used as an indicator of equity, but like the indicators in Table 4.1, it suffers from the untested assumption that needs are everywhere proportional to population. At the other extreme, cost per consultation leaves out much that is important for equity; the cost per consultation could be uniform, yet the health system would not be equitable if the chance of getting attention var- ied widely among population groups. Thus the best measure of equity in spending on health care seems to be expenditure per person needing assis- tance--ideally this indicator would not include persons with symptoms too minor to require medical attention. Expenditure and morbidity. It appears from Table 4.3, that patient-related expenditures are distributed more equitably in Peru than are some of the resources which help account for those costs. The Lima health region, for example, has about one-third of the country's population, a slightly larger share of those with illness or symptoms, two-thirds of Peru's physicians, but less than half of spending attributable to individual treatment. This percep- tion is confirmed by Fig. 4.2, which compares the Lorenz curve for expen- diture per sick person to that for physicians per capita. The former distri- bution is systematically much more equitable: the very unequal distribution of physicians undoubtedly overstates the inequity of the Peruvian health care system. Differences in unit costs contribute to total inequality, but their Measurement of Equity in Health · 117 impact is small compared to the effect of differences in the likelihood of receiving treatment: costs per consultation, for example, vary over a range of only 2:1 in most of the health regions, and even those differences may depend on regional variations in the incidence of particular diseases or con- ditions, some of which in fact cost more to treat than others. These results only reinforce the conclusion that the chief source of health care inequity in Peru is variation in the probability of getting medical attention when sick. Inequality without inequity. Equity is related to, but not identical with equality: thus, as remarked earlier, there can be inequality which is not nec- essarily inequitable. Suppose, for example, that while the cost per consulta- tion in 1984 ranged from 8,845 to 23,843 soles among Peruvian health regions, adequate care could be given for a unit cost of 15,000 soles. That is, higher unit costs represent inefficiency, use of overqualified personnel, etc. Then, adequate medical care could have been given to all those patients attended, for only 85% of what was spent. This situation is shown in Fig. 4.2, where the hatched area represents real inequity, and the dotted area represents inequality. The former is due mostly to differences in the probability of receiving care, plus some inequality in unit costs below 15,000 soles, suggesting inadequate care, on average, in some regions. The latter is due entirely to average expenditures exceeding that level in some health regions. While the choice of a unit cost of 15,000 soles is arbitrary and used here only for illustration, the argument that not all inequality repre- sents real inequity in health care is quite general. The logic of this compar- ison is similar to that involved in drawing a "poverty line" in the distribu- tion of income, and regarding as inequitable the fact that some incomes are below that line, without being concerned with differences in income among those above the poverty line. Equity as getting what you pay for The comparison of costs or expenditure to population or to medical needs still does not ask how those expenditures are paid for. To medical profes- sionals, equity is usually interpreted in terms of the satisfaction of needs, and questions of payment arise only as possible obstacles to obtaining medical care. This is a somewhat different matter from the satisfaction felt by the patient or consumer, because his demand for health care may not coincide with his need as determined by a physician. Equity could also be considered to require the satisfaction of wants or demands. To economists, however, equity has another meaning: that the distribution of health care expenditure 118 · Health Economics in Development should be less unequal than the distribution of income, or that there should be a net subsidy (expenditure minus tax contributions) to population groups with low incomes, and a net contribution by groups with high incomes. Equity then becomes a question of the amount and direction of net subsi- dies. Studies with this orientation commonly find that public health care spending is progressive when compared to the distribution of income, and also that it is more equitable than some other kinds of public social expen- diture such as spending on education (4). That is, compared to what they pay in taxes, the poor usually get a net benefit from public health care, whereas they may suffer a net loss in certain other categories of government expen- diture. This kind of analysis has not yet been conducted for Peru, although studies have been completed in several other Latin American countries. The first of these recent studies to be published concerns Chile3. It is important to note that these two concepts of financial equity in health care--expenditure relative to need, and expenditure relative to pay- ment--may, but will not necessarily, coincide. If a particular population group contributes nothing and receives little or nothing, that will be equi- table according to one criterion but not the other. This question becomes particularly important if the object is to judge the equity of a public health service system, when there is also private fee-for-service medical care avail- able. Is the public system equitable if it provides access to those who cannot afford private care, ignoring those who can and do buy medical attention from private providers? Or does equity require the public system to reach, and subsidize, those poor consumers who now pay for private services because they are more accessible, or believed to be of better quality, than public services? The answers to these questions determine which indicators to construct, and how to interpret the available information. Concluding Reflections Equity is too complex a concept to be reduced to a single indicator; to ana- lyze it necessarily requires a great deal of information and some subjective judgment as to what kinds of inequality in fact constitute inequity. However, many indicators can be constructed which are related to equity, or which help to measure it. The empirical discussion of the Peruvian case illustrates both the difficulties of analyzing equity in health care and the possible uses of administrative, financial and household data to form some overall judg- ment and to identify where inequity is concentrated, or with what factors it Measurement of Equity in Health · 119 is associated. This is much more valuable than arriving at some single over- all measure of how equitable a particular health care system is. The need to compare medical consultations and expenditures not just to population but to morbidity and perceived needs for assistance, indicates that relatively full assessments of equity must draw on population-based data and cannot be constructed only from the kind of information normally available to a ministry of health. The infrequent collection and high cost of such population data mean that equity is more easily studied in the cross section than in year-to-year changes. However, given a baseline assessment of how equitable a system is and where its principal problems are, changes over time in the distribution of resources and effort can give a good idea of whether the system is becoming more or less equitable. It is less important to calculate elaborate statistical measures such as summary coefficients of inequality, than to have a clear view of the range or variation in the proba- bilities chosen for study. All such efforts, to be useful, must take account of the diversity of health conditions and of responses to them, and of the dif- ferent dimensions of the population according to which equity can be evaluated. Notes 1Central Reserve Bank of Peru. Mapa de Salud en el Perú. Lima, 1984. 2Chirinos, 0. et al. [Health sector financing and expenditure. Study 6.2, Financing and costs sector, ANSSA-Peru, Taller seminar III]. Lima, 1986 (in Spanish). 3Rodriguez Grossi, J. [The distribution of revenue and social expenditure in Chile, 1983]. Santiago, Latin American Institute of Social Sciences and Studies (ILADES), 1985. (In Spanish). References [1] PAN AMERICAN HEALTH ORGANIZATION. Health for all by the year 2000: plan of action for the implementation of regional strategies. Washington D.C., PAHO, 1982. (Official document no. 179). [2] PRESIDENT'S COMMISSION FOR THE STUDY OF ETHICAL PROBLEMS IN MEDICINE AND BIOMEDICAL AND BEHAVIORAL RESEARCH. Securing access to health care: the ethical implications of differences in the availability of health services. Volume 1: report. Washington D.C., US Department of Health and Human Services, 1983. 120 · Health Economics in Development [3] MUSGROVE, P. Equity in health system services--concepts, indicators and interpretation (summary in English). Boletin de la Oficina Sanitaria Panamericana, 95 (6): 525-546, 1983. [4] JIMENEZ, E. The public subsidization of education and health in developing countries: a review of equity and efficiency. The World Bank Research Observer, I (1): 111-129, 1986. CHAPTER 5 What Should Consumers in Poor Countries Pay for Publicly-Provided Health Services? Acknowledgments An earlier version of this paper was presented at the National Council on International Health Conference, Washington, D.C., June 1982. Nothing said here represents the position of the Pan American Health Organization or its Member Governments. I am grateful to Robert Robertson, Dayl Donaldson, David Dunlop and anonymous reviewers for a large number of helpful comments, some but not all of which have been taken into account in the revised version; responsibility for errors of fact or interpretation is mine alone. Definition of the Problem The question to which this paper is addressed was defined for me by the Coordinator of Health Services of the Ministry of Health and Education of one of the smaller Caribbean countries. Publicly-provided health care in that country is free, in accord with the government's ideological commit- ment that no one should be denied medical attention for economic reasons. However, it is extremely difficult to raise the revenue required to operate the existing health care system, especially in the current economic situation, and the difficulty will only be compounded if coverage is expanded. Under Reprinted from Social Science and Medicine 22 (3), 1986, with permission from Elsevier Science. 121 122 · Health Economics in Development these circumstances, the government would like to recover some part of the cost from the consumers or beneficiaries, while doing the least violence to its principles; the question is, what prices or fees for service should they adopt? Thus expressed, the question is not the same as, What should the price(s) of health service(s) be? In particular, prices are not assumed to determine supply; the government decides on the number and type of facilities to build and staff and the services to offer, on other grounds. Prices are not even assumed to determine resource allocation within the existing or planned level of supply, in the short run. Consumers will have to respond to prices, but suppliers in the public sector will not. This does not mean that it is of no interest to operate the public health system more efficiently, only that prices to patients are not to play a role in that effort. In these circumstances, what is sought is clearly a second-best solution. It is irrelevant to a ministry of health--and most if not all ministries of health in Latin America and the Caribbean find themselves in this same sit- uation--that a first-best solution might involve restructuring the entire tax system, or changing a great many other prices in the economy. The min- istry of health does not control taxes: it simply faces pressure both from within the government and from international donor agencies and from the International Monetary Fund, to recover some part of its costs, either so as to permit expansion of its services or so as to make smaller demands on the budget. When allotments to all ministries are cut as part of an austerity pro- gram, imposing fees may be the only way to avoid reducing services. In any case, the question of what, if any, fees to charge would still be present even if the tax system were modified and the economy otherwise made to oper- ate more efficiently. Given the restricted role of prices in this view of the problem, and the short-run pressures on the public health system, two other limitations of the question are important. The first is that there is no requirement that fees cover any particular fraction of total cost or bear any particular relation to marginal costs. The second is that equity considerations are vital: the gov- ernment does not want to purchase efficiency at the expense of its view of equity. Its goal is essentially that espoused by the U.S. Presidential Commis- sion, which studied the issue of securing access to health care [1, pp. 20, 111], that of an adequate level of health care for everyone, without economic diffi- culties causing people who need care to forego it or to postpone it until their medical problems are much worse. I have argued elsewhere that medical equity should be interpreted as an equal probability of receiving care when it What Should Consumers in Poor Countries Pay? · 123 is needed [2], and it is this probability which the government wants to reduce as little as possible as a consequence of charging fees. This is distinct from concepts of equity which consider how health care is to be financed, but the government is also interested in financial equity. This issue is addressed later. The question of what the government wants is taken up in more detail in the next section. Given some strong assumptions about the welfare function implicit in the ministry of health's actions--a welfare or objective function which I assume is characteristic at least of most governments in Latin America and Caribbean countries and perhaps in poor countries generally--it is possible to derive conditions for the optimum level of prices or fees to charge. Because this level depends on the demands of consumers for health care, the following section considers what assumptions are reasonable to make about those demands. Given the emphasis on equity, the relation of health care demands to income may be particularly important. Other determinants of demand are considered more briefly. Although it is convenient to speak of `the' price for medical attention, in fact medical care consists of a great variety of goods and services. Moreover, these are not independently produced and consumed; one level of utilization (consultation) may be a prerequisite to another (treatment). Even for a par- ticular element of care, the price may vary among consumers, geographic areas or other dimensions; the next section therefore considers price dis- crimination, particularly with respect to income and to the other (non-fee) costs which consumers must meet in order to obtain medical attention. Conditions for an Optimal (Single) Price Under the assumption that the price charged for medical care is not to determine the supply, the optimum price is to be determined taking account not of the cost of production but only of the government's welfare function, i.e. its objectives in providing care. Governments do not usually specify wel- fare or objective functions in such terms as to permit the derivation of opti- mum policies, for prices or for anything else. The member governments of the Pan American Health Organization, for example, have subscribed to a mixture of objectives [3, p. 15], some of which specify measurable results in terms of death rates, immunization coverage, etc., while others refer to gen- eral goals of "reorganization", "improvement", etc. to achieve higher levels of (undefined) "equity, efficiency and effectiveness". 124 · Health Economics in Development I will assume as a first approximation that for the purpose of deciding what fees to charge, the government's welfare function may be regarded as having just two components, which do not interact with one another. These are N(p), the number of consultations or utilizations by patients, expressed as a function of price; and pN(p), the total revenue obtained from these con- sultations. The first element is a reason for wanting fees to be low, while the second may justify high fees, depending on the elasticity of demand for medical care. In order to form a welfare function, W(p), from these two elements, a parameter is required for comparing the volume of service to the monetary value of revenue. If this parameter is called , and is assumed to be con- stant--each consultation is treated as of equal monetary value, although its value may not be the same as its price--then W1p2 = N1p2 + pN1p2. It is assumed that for a given need, all consumers benefit equally from treatment; there is no discrimination either according to individuals' sub- jective valuation of how much they benefit from care, nor according to any more objective criteria such as the effect on earning power. Note that the same formulation would result if the ministry were assumed to try to maximize N(p) subject to a budget constraint in which total expenditure were equal to resources from the treasury plus the revenue generated by fees. Differentiating W(p) with respect to p yields, after some manipulation, 0 7 N = -1>11 + >p2 7 -1 where N is the price-elasticity of demand. If demand is elastic (falling more than proportionally as the price is raised) at a zero price, then the price should be raised until demand becomes inelastic. Just how inelastic it should be depends on the relative sizes of and p. If the ministry cared only about revenue ( 0), the elasticity should be exactly 1; whereas a high value on attending to patients ( large com- pared to the price p) means a low elasticity. At the extreme of caring only about consultations, price should be zero. This very simple form of the objective function requires only that the ministry decide for itself how much a consultation is worth, compared to the revenue obtained from it--i.e., establish a value for --and then learn enough about how demand depends on price to adjust price and therefore N(p) to the optimum level. The analysis can be complicated in a number of What Should Consumers in Poor Countries Pay? · 125 ways, such as by introducing other elements into the objective function or by distinguishing different types of demand. Two of these possibilities are considered next. The government's concern with equity, for example, might be explicitly incorporated into its objective function by including a third term, aUi1Yi,p 2 i C where Ui is the utility of the ith consumer, expressed as a function of his income Yi and the price p. Charging that price will reduce the utility of any patient who actually is attended, compared to treating him for free; the sum is therefore taken only over the set C of patients who are seen when the price is p. The parameter compares the value (to the ministry) of these utility losses, to the revenue obtained: the loss for any customer is greater as p is higher or as Yi is lower, so if a concave utility function is assumed, the total loss of utility will be greater as the distribution of income is more unequal, leading to a lower price. It is assumed here that interpersonal com- parison of utilities is appropriate, and that income, properly defined, is a reasonable basis for such comparisons. This assumption is necessary to the `ability-to-pay' approach to price setting [4, Chap. 5], however dubious the idea is; and more generally, it is the `old welfare economics' assumption for necessities such as food, shelter and health care [5], for which everyone's needs can be assumed to be very similar if not identical, in the same physi- ological circumstances. Prices usually are not given the role of protecting equity in the economy, and for the producer of a typical good or service the distribution of income matters only as it affects sales and therefore revenues and profits. Ministries of health do not appear to think this way, however; and as they have some control over prices but no control over their clients' incomes, they may try to use prices to affect welfare apart from the effect on the number of patients cared for. If this term is included, and W(p) is differentiated, then so long as the same set C of customers is considered, the condition for the optimum price becomes -1 - = N 1 + >p E1dU>dp2 where E refers to the average or expected value of the change in utility, dU/dp, when the price is increased. Since this is negative, the effect is to 126 · Health Economics in Development bring the optimum elasticity closer to zero and to lower the optimum price. Changing the price will have the further effect of shrinking or expanding the set C of clients who use the medical service, so that some people's util- ity losses due to the price will cease to be, or will become, of interest to the ministry. Formulating the objective this way means representing the min- istry as caring both about the total number of consultations N(p) and about the group of patients who account for those consultations, whereas the previous formulation is concerned only with utilization and does not distin- guish one patient from another. It may seem like double-counting to con- sider both the utilization N(p) and the welfare loss that results from paying the price p for that utilization, but public health officials do express both objectives, at least rhetorically. This formulation of their objectives attempts to reflect the cost associated with a highly inelastic demand, so that a high price can be considered undesirable even if it does not reduce utilization. In fact, not all consultations are equally necessary or desirable on med- ical grounds. Some can be considered frivolous, so that while they are welcome for the revenue they generate they should not be considered to contribute to welfare, and they still represent a cost to patients. The min- istry's objective function can be rewritten to separate necessary (N) and friv- olous (F) demand, both as functions of price, as W1p2 = N1p2 + p5N1p2 + F1p26; Differentiating yields, after some manipulation, the condition N11 + N 2 + F11 + F2 + dN>dp = 0 where N and F are respectively the price elasticities of N and F, which may be different. Whether N and F will be different functions of price will depend on whether consumers' views of the necessity for care coincide with doctors' opinions. The formulation shown allows for different elasticities with respect to p, but--since patients may be poor judges of how much they need medical care--does not require it. In order for the maximum condition to be satisfied at a positive price, 1 N or 1 For both must be positive; otherwise all three terms in the expression are negative. This means that if demand is elastic (falling more than proportionally as price is raised) at zero price, the price should be raised until at least one of the demands becomes inelastic. If necessary demand is inelastic to start with, as is likely if patients correctly recognize What Should Consumers in Poor Countries Pay? · 127 the more serious needs, then any increase in price will have little effect on medically needed utilization, but may still reduce less-needed demand F. Whether in that case price should be raised above zero will depend on the levels of both kinds of demand; it is not necessarily the case that price should increase until Falso becomes inelastic. Rewriting the optimum con- dition to separate N , as in the previous analysis, yields -1 - 1F>N211 + F2 = N 1 + >p which depends on the relative amounts of frivolous and necessary demand and on the elasticity of the former. Since F can be more or less than one in absolute value, the optimum price to charge can go either up or down com- pared to the case where all demand is necessary. If both kinds of demand exist but react the same way to price, then 1 + >p + F>N -11 + F>N2 = = N F reflecting the fact that a higher price has the good effect of reducing frivo- lous demand but the disadvantage of also reducing needed consultations, and the balance of these two effects depends on F N. Introducing patients' utilities explicitly affects the optimum levels of elasticity and price, but not the need for at least one of the demands to be inelastic. Consumer Demand and its Determinants The demand for health care, whether or not it is separated into necessary and frivolous components, can be presumed to depend on the price charged for care; on the costs in time of reaching a medical facility and waiting for attention; on the income foregone by seeking care; on income; on one's general condition of health; on the specific conditions or symptom(s) of ill- ness or accident; on the consumer's ability to diagnose and treat himself, which depends among other things on education; and on his opinion of the quality of care he is likely to receive [6]. With respect to price and income, I am assuming that the consumer's demand function has the shape shown in Fig. 5.1. Two consumers are represented, with incomes OY1 (poor) and OY2 (rich). Facing a relatively high commercial price PC such as a private doctor would charge, the amount of care demanded by the richer consumer, C2, is much larger than that demanded by the poorer one, C1. This corresponds 128 · Health Economics in Development Figure 5.1 Demand for Medical Care as a Function of Income and Price PRICE INCOME Pc Y2 P' Y1 C2 N2 D2 C1 M2 O N1 D1 M1 QUANTITY (UTILIZATION) to the empirical finding that total private spending on medical care in Latin American countries typically shows an elasticity with respect to income of slightly more than 1.0 [7]. (For specific components of spending the elas- ticity may be quite different: for drugs, for example, it starts high but quickly declines as demand approaches saturation.) At a zero price, the poor consumer's demand M1 may actually exceed that of the richer consumer, M2, because the former is less healthy or less well able to treat himself, or because his time is less valuable. It is not obvious which consumer will exhibit the greater necessary demand (N1 and N2) when the price is zero, but it is to be expected that except at low prices, rais- ing the price will reduce the necessary demand more for the poorer con- sumer. For both consumers, the horizontal distance between M and N rep- resents frivolous demand; this is presumed to be more sensitive to price than necessary demand, and is likely to be greater for poorer consumers, at least if the non-fee costs of seeking medical care (travel and lost income) are low. Under these conditions, the optimum price discussed above is something like p*, established at a level where the demand curve has become inelastic; compared to a zero price, necessary demand is reduced only slightly and What Should Consumers in Poor Countries Pay? · 129 frivolous demand considerably more. Exactly where the optimum is depends also on the importance (if any) of consumers' disutility from having to pay the fee, and on the weight ( ) assigned to consultations in the gov- ernment's objective function; these are not shown in the figure. The demands resulting from p* are D1 and D2. Two further observations concerning the assumed demand function and the consequent optimum price are in order. The first is that the suggested rule for determining price is in one sense the opposite of what a private monopolist would follow. Here the price is set in relation to where the orig- inally elastic demand becomes inelastic, as price rises; a monopolist would set price at the other end of the inelastic range, where the price has become so high that raising it more begins to reduce even necessary demand sharply and so revenue falls. The second observation is that if the optimum price were determined separately for each consumer, it is conceivable that it would be lower for the richer one, despite his greater ability to pay. This is because of his presumed more inelastic demand. Setting the optimum rela- tive to all consumers together takes account of the distribution of income among them and also of the possibility of extracting more revenue from those with higher incomes. Finally, it should be noted that both the government's supposed welfare function and the consumer's demand refer to instances of medical care, not to health in general or to other kinds of behavior which influence health. Medical care is undoubtedly sought to maintain or improve one's (per- ceived) health, and the government's ultimate objective is a healthy popula- tion, but models of the `demand for health' [8, 9] are too general and too long-term in focus to help determine what fees should be charged. Both consumers and the government take actions designed to promote health by other means than the provision of medical services; these actions can best be thought of as shifting the demand curve through better self care, more prevention and less cure, and otherwise reducing the number of instances in which medical care is necessary. Optimum Price Discrimination The argument thus far has assumed that all consumers will pay the same price, but it is probably preferable to discriminate in price-setting. The most obvious reason is the disparity of incomes: richer patients can afford to pay more, without significantly reducing their necessary demand. Calculating 130 · Health Economics in Development the optimum price separately by income level will lead directly to prices being higher for the rich, provided the utility-loss term reflects decreasing marginal utility of income at a rapid enough rate. If no account is taken of utility loss as a function of income, the elasticity analysis alone could lead, as indicated above, to the poor being charged more. (Alternatively, can be considered to depend on income, but then all patients are not considered equally valuable in medical terms, which seems unethical.) Introducing price discrimination among consumers of course raises two problems: patients have to be classified by some sort of means test, which creates some admin- istrative expense, and it has to be decided for which services to charge dif- ferentiated prices. Governments do often recognize the greater ability of some patients to pay than others, and introduce higher fees for the former, but even then the price may remain very low, hardly justifying the adminis- trative burden and bearing little relation to incomes. This is perhaps espe- cially likely to occur with hospital charges [10]. At the other extreme, dis- crimination may mean treating indigent patients completely free of charge. Discrimination among consumers by income level may be of limited potential in many countries, simply because the better-off are unlikely to be clients of the ministry of health in the first place; if there is medical cover- age under social security, it is the better-paid workers who are most likely to benefit from it. Of course, the arguments in favor of fees to reduce friv- olous demand apply equally well in a social security medical care system, but the revenue argument is unimportant if the system is already financed by payroll taxes which the beneficiaries regard as their contribution to paying for the service. As for where to apply price discrimination, the simplest procedure is probably to set fees for particular services and then to apply uniformly a scale of adjustments based on income. However, the administrative burden would be reduced, and a considerable revenue could still be collected, if a uniform fee were charged for consultation (set low enough not to have much effect on necessary consultations by poorer consumers), and price discrimination were introduced only at the stage of treatment. In particular, hospitalization seems an appropriate stage at which to set differential fees, both because of the large share of hospital costs in total government health budgets, and because higher-income consumers sometimes use private doc- tors for consultations, but go to public hospitals for treatment. One of the arguments for setting fees at all is that this will reduce unnec- essary demand, ideally with minimal effect on necessary utilization. It is more correct to say that the total cost of obtaining medical care will reduce What Should Consumers in Poor Countries Pay? · 131 frivolous demand; and if the cost of travel, waiting time and income lost are high enough, most or all unnecessary utilization will already have been eliminated. In that case there will be no role for fees except to raise revenue, and so the optimum fee will in general be lower. (This argument is offset, to the extent that a given fee represents a lower share of total cost when other costs are high, so that the price-elasticity of necessary demand, N , is reduced, permitting the optimum fee to be somewhat higher.) This suggests a second form of price discrimination may be desirable: fees should be lower for consumers for whom the other costs of obtaining care are higher. In par- ticular, fees should perhaps be lower in rural areas of dispersed population and difficulty of access than in urban areas with good transportation and rel- atively easy access. This kind of geographic basis for discrimination has the advantage that it does not require classifying individual patients. Moreover, since the population of thinly-settled rural areas is typically poor, the lower fees would not usually benefit high-income consumers. The argument is sometimes made that introducing fees is a way to make the revenue of a particular health facility reflect the quality of service it offers its clientele, and therefore to stimulate managers of hospitals and clinics to operate more efficiently and to satisfy their patients better; it not only permits but rewards decentralization. Among facilities serving the same class of consumers, this is a reasonable defense of fees, but it ceases to be valid once the non-fee costs of obtaining care differ. Applied to the sort of urban/easy-access and rural/difficult-access facilities just described, it would mean either that the urban facility would have extra funds with which to expand service where it is already better than average while the rural facility would not, or else that necessary demand would be reduced more in the already-under-served rural area. Equity in the raising and use of revenue from patients' charges probably requires either that fees raised in one place be (partly) transferred to less-favored facilities and populations, or that regular (tax-financed) budgets go preferentially to those locations where the optimum fee is lower. References [1] President's Commission for the Study of Ethical Problems in Medicine and Biomedical and Behavioral Research. Securing Access to Health Care: the Ethical Implications of Differences in the Availability of Health Services, Vol. I (Report): 20, 111. Washington, 1983. 132 · Health Economics in Development [2] Musgrove P. La equidad del sistema de servicios de salud. Conceptos, indicadores e interpretación. Boletín de la Oficina Sanitaria Pan-Americana. 95, 525, 1983. [3] Pan American Health Organization. Plan of Action for the Implementation of Regional Strategies: 15. Official Document 179, 1982. [4] Musgrave R. The Theory of Public Finance. McGraw-Hill, New York, 1959. [5] Cooter R. and Rappoport P. Were the ordinalists wrong about welfare eco- nomics? Journal of Economic Literature 22, 507, 1984. [6] Newhouse J. P. The demand for medical care services: a retrospect and prospect. Health, Economics and Health Economics (Edited by Van der Gaag J. and Perlman M.). North-Holland, New York, 1981. [7] Musgrove P. Family health care spending in Latin America. Journal of Health Economics. 2, 245, 1983. [8] Grossman M. The demand for health: a theoretical and empirical investiga- tion. NBER Occasional Papers 119, 1972. [9] Muurinen J. M. Demand for health: a generalized Grossman model. Journal of Health Economics 1, 1, 1982. [10] Pan American Health Organization and Agency for International Development. Belize Health Sector Assessment, 1982. CHAPTER 6 Compensatory Finance in Health: Geographic Equity in a Federal System Acknowledgments I am indebted to Rena Eichler for much useful information about the United States Medicaid Program; to colleagues from the World Bank, the Inter-American Bank and the Brazilian Ministry of Health for the discus- sions which eventuated in the formula for assignment of federal funds to states in the two Banks' REFORSUS project; and to Barjas Negri of the Ministry for his urging to put these comparisons together and draw con- clusions from them. Atilio Alsogaray of the University of La Plata provided the stimulus to write this paper. Introduction: Relations Between Nations and Localities The health of the population is often considered a responsibility of the national government, whether specified by law or constitution or through political consensus. At the same time, it is common for sub-national gov- ernments to control the public provision of health services, and even to assume the responsibility of financing private providers. This situation cre- ates the need to determine the relation between one and another level of government, particularly with respect to financing. Two characteristics are relevant: the differences among parts of the country, and the specific capac- ities of the national government. Translated and revised version of a paper in Spanish presented to the Third Inter- national Seminar on Fiscal Federalism, La Plata, Argentina, 24 April 1998. 133 134 · Health Economics in Development Local inequalities and inequity In general, a federal country is characterized by differences among regions, states or provinces, and municipalities. Relative to health, these may be dif- ferences in: The degree of poverty, which limits private spending on health; Local fiscal capacity for public spending on health; Total health expenditure; Needs for health care (the epidemiological situation); and finally, The overall state of health. These differences, especially in health status and access to services, are often considered unjust or inequitable, requiring compensatory or redistributive interventions. The role of the national government At least in comparison to the poorest states or provinces in the country, the national government generally has a greater capacity to generate tax revenues. The tendency--more marked, the more public functions are decentralized--is for the federal government to raise resources and transfer them to sub-national governments. In addition to the purely fiscal respon- sibility, the nation also is expected to assure equity or even equality among its citizens. This implies, among other things, reducing geographic differ- ences. This function can be exercised either for universal programs, to which everyone has a right, or for targeted programs. An Example: The Medicaid Program of the United States The United States depends, more than is typical for a rich country, on pri- vate financing of health care, largely through employment-related insur- ance. Nonetheless, there are two large federal health programs: Medicare for the elderly, which is financed by a specific tax, and Medicaid for the poor. Both are shared between the federal government, which legislates the rules under which they operate and contributes to financing them, and the state governments, which administer the programs and also spend on them. (In contrast to what happens in some other countries such as Brazil and Chile, municipalities do not figure in the definition, financing or operation of these programs.) Compensatory Finance in Health · 135 Logic and objectives of the program The poor, by definition, can hardly pay for health care out of pocket. On the other hand, they seldom have jobs that allow access to adequate insurance at reasonable cost. The chief objective of Medicaid is to provide subsidized insurance coverage for poor people. Another goal is to allow considerable freedom to state governments, not only to administer the program but to define the beneficiaries, the benefits to which they are entitled, and the cor- responding cost. In contrast to some other programs, such as those described below, Medicaid does not operate by distributing a fixed federal budget. Determining expenditures and their distribution The state government (s) defines the beneficiary population (Bs) to be eli- gible for the program, subject to the regulations in the national legislation. Similarly, and also subject to federal rules, it determines the set of health services to be covered by the insurance. The combination of beneficiary characteristics and the list of services determines the quantity of care (Qs) to be utilized. On the other hand, and also subject to federal approval, the state gov- ernment determines, or negotiates with providers (mostly private), the prices ( s) it will pay. These prices should be the lowest necessary to guar- antee a supply of services of adequate quality. A provider who agrees to serve Medicaid beneficiaries promises not to require co-payments or addi- tional fees from patients. Together, the prices and quantities determine the total cost of the program in the state, Cs Qs* s. What the federal government determines is the fraction ( s) of that cost that it will contribute in that state. This depends on state per capita income; at incomes above about $ 22,000 (in 1995), the share s reaches its mini- mum value of 50 percent. That is, in the richer states the federal govern- ment finances half the cost of the program. In the poorer states the federal share is larger, reaching a maximum of 83 percent. Federal or national expenditure in the state is therefore CNs s*Qs* s, while the state government's expenditure is CSs (1 s) *Qs* s. The relation of federal to state spending is then CNs [ s/(1 s)]CSs, with a minimum of equal expenditure at the two levels and a max- imum of federal spending being five times that of the state government. Observations: who decides what The federal government influences the determination of the beneficiaries, the services to be subsidized, and the prices to pay for them. However, the 136 · Health Economics in Development only parameter it directly determines is the share it contributes to financing-- on the basis of income per person, which neither the national nor the state government can influence appreciably. The other decisions are taken by the states. This has an important consequence for equity: two equally poor peo- ple in two different states can be treated differently. This can happen because of inclusion among the beneficiaries in one state and not the other, or because more services are covered in one state than in the other. To put it another way: the federal Medicaid program is redistributive-- it contributes much more to paying for health in poor states than in rich ones--and thereby compensates for differences in the number of poor peo- ple as well as in the capacities of states to assure them adequate health care. But the program does not impose equality of treatment, leaving it to state governments to decide how "generous" to be in applying the rules. This represents a relatively high degree of decentralization of decisions. It also has the consequence that the decisions by the states influence the total expenditure to which the federal government is committed. Finally, as occurs with some other programs, especially that of Aid to Families with Dependent Children, differences from one state to another may motivate migration from a less-generous to a more-generous state. There are no estimates of whether this happens on a significant scale. A General Example: Explicit Redistribution Another way to define a redistributive health financing scheme is to begin with a uniform level of expenditure per capita, or a uniform set of services, for the entire beneficiary population of a public insurance. Then one deter- mines how the cost is divided between different levels of government. This logic corresponds more to the Canadian health system than to that of the United States, particularly as the former operated in Canada a decade or more ago, when there were specific federal transfers to the provinces to finance health. (Nowadays the transfers also pay for education and other programs, and it is left to the provinces to decide how to distribute the funds among different uses.) In general, such a system can be universal, as in Canada, or it can exclude part of the population because it is rich, or has private health insurance, or for some other reason (as occurs in the Nether- lands). The definition of a public system which takes into account the cov- erage of private insurance, avoiding subsidy to those who do not need it, while assuring equity among the whole population, is of particular interest Compensatory Finance in Health · 137 for countries with mixed health systems such as the majority of countries in Latin America. The logic of a national health insurance What follows describes a system for defining compensatory transfers from the national government to sub-national governments, which allows the lat- ter some liberty in spending their own resources but imposes common national norms for the beneficiary population, the services covered, and their costs. Another difference from the Medicaid program is that the shar- ing of costs depends not on per capita income but on the revenues of the states--including any non-tied transfers which the state or province receives from the federal government for spending as it chooses. This leaves more control to the state for determining its own taxes and consequent rev- enues, but simplifies the determination of how much the state will have to spend on health. Revenues are known more quickly and more precisely than incomes, especially in countries with less-developed information systems. The central objective is to guarantee a uniform minimum coverage: to impose a floor but not a ceiling. The other objective is to control federal spending, although not rigidly--that expenditure is still influenced by the decisions of sub-national governments. A redistributive formula The national government defines (perhaps through negotiation with the states or provinces, as in Canada) the population which, because it is already privately insured (A), is not eligible for the full benefits of the public insur- ance. If private insurers cover fewer services than the public program, that insured population can participate as partial beneficiaries and still receive transfers. If Ps is the population of a state, the fully-covered beneficiary population is Bs Ps As. The national government also determines the minimum level of per capita expenditure (C), or alternatively the minimum set of services (Q) to be financed. (Trying to specify both quantities and expenditures leads to rationing, if the spending is insufficient, or to expenditure which can buy more services than the minimum defined.) This determines the committed level of spending by each state, which can be expressed as Cs C*Bs, or Cs Q* *Bs, where are the prices, specified nationally. (This ignores the potential complexity of allowing different federal prices in different states, because providers' costs differ. However, that complication is often 138 · Health Economics in Development relevant for hard-to-reach rural areas or dispersed populations.) The state, for its part, retains the right to offer more services than are negotiated with the federal government, Qs Q, or to pay higher prices for them, s . In either of these cases, it has to spend its own resources: the federal com- mitment is limited to financing only Q* . If the privately insured population receives a more limited set of services than the national minimum, Qas Q, the formula for the total expendi- ture committed has to be modified to include the services not guaranteed by private insurance. The result is Cs = C*Bs + (Q - Qas)* *As, or Cs = Q* *Bs + (Q - Qas)* *As, which can be expressed as Cs Q* *Ps Qas* *As. The program pays the cost of the minimum set of services for the whole population of the state, less the cost represented by the services provided by private insurers to their clients. The last element of the system is the determination of how this total cost is shared between the federal and sub-national governments. There can be many different formulas, the simplest being that the state or province is required to spend on health, a fraction of its revenues Rs (own tax rev- enues plus non-tied federal transfers). This way the state contribution becomes CSs *Rs, and that of the federal government is CNs Cs CSs. Federal expenditure is purely compensatory, guaranteeing the mini- mum stipulated spending or services for the whole population, including any shortfall of private insurance coverage relative to the public insurance. Observations on political decisions In a system like that just described, the states or provinces can participate in two types of decisions. If the parameters C or Q, and are not fixed by law or the constitution, but negotiated (for example, annually), the sub- national governments share in determining them. The best way to fix the parameter values depends on the country's political situation, and in partic- ular on the relative power of different states or provinces to benefit from these decisions. In the case of Brazil, the constitution was amended, first for education and later for health, to specify the responsibilities of the different levels of government (federal, state and also municipal). This was done to prevent the states with greater wealth and political power from continuing to obtain inequitably large federal transfers because the ceilings on federal Compensatory Finance in Health · 139 resources were determined more on the basis of the distribution of pro- viders than in relation to needs for health or financing. In the second place, each state determines its own taxes, which affects the revenues Rs of which a share must be spent on health. There is some risk of moral hazard, in that a state can pass more of health costs on to the fed- eral government by reducing its tax rates. This danger is not entirely imag- inary: it happened with a corrupt and irresponsible government in the state of Alagoas, in Brazil, which went effectively bankrupt by reducing taxes on the sugar industry and waiting for the federal government to pay for public services such as health. However, in general there are two strong protec- tions against such moral hazard. One is that reducing taxes so as to pay less for health means losing larger revenues for other uses, especially if is low, say 10 or 15 percent. The other protection is relevant if besides its own taxes, state revenues Rs include non-tied or general transfers from the fed- eral government. In a poor state, these transfers can be a large share of total revenue, which effectively assures that the federal government will finance much of health expenditure and thereby protect the neediest. However, this requires that the non-tied transfers themselves not be very inequitable, to the point of annulling the equity due to the tied transfers for health. Equity in Investment: The REFORSUS Project in Brazil The two examples just discussed refer to recurrent spending in health, that is for the purchase of services. The problems of geographic inequality or inequity and the need for compensatory actions also can arise for investment spending, that is for the creation of capacity to produce services. The REFORSUS investment project in Brazil, begun in 1996 and financed partly by the World Bank and the Inter-American Development Bank, was designed to complete, equip or remodel hospitals and other health facilities. It required some arrangement to distribute the resources of US$590 million among the states of the country in an equitable and politically acceptable way. It therefore offers an illustration of how the ideas discussed here can also be applied to investments. Background and objectives Federal health expenditures in Brazil at the outset of the project (before the constitutional amendment mentioned above, which specifies values of for states and municipalities) were distributed very unequally among 140 · Health Economics in Development states--largely because the public system (Sistema Único de Saúde) paid according to the production of services, and capacity for provision was very unequal and concentrated in the richest states. The resulting inequity was limited, at that time, by federal spending ceilings established for each state, besides a general ban on paying for the hospitalization of more than eight percent of a state's population in any one year. The ceilings were negotiated annually, but because of the unequal political power of the state govern- ments, did not succeed in equalizing expenditure per capita, much less to adjust it according the fiscal capacities of the states for financing health from their own revenues. The goal in the REFORSUS project was to establish ceilings for invest- ment in each state, which would correct part of the existing inequity. They would also have to respect three important limitations: according to budgetary law, at least 50 percent of the funds would have to be assigned on the basis of population; no state should receive more funds than could be absorbed in justified investments; and the criteria had to be transparent and the formula simple. Determination of state-by-state ceilings Many criteria were proposed for distributing the total investment fund F among the states or determining the values Fs. Among these were the level of income, the prevalence of poverty, age-standardized mortality, and many other indicators. In principle all were relevant, but they would have yielded a formula too complex to be understood and accepted. In consequence, and after much discussion, it was decided to base the allocation of funds solely on the state's population, Ps, and recurrent federal health spending per capita in the state, Cs. In addition, it was evident that the formula should be linear in these two variables, so as not to give very high or low allocations to states with extremely high or low population or per capita expenditures much above or below the mean. (Note that according to the rules for the Medicaid program, the fraction s is linear in per capita income, up to the maximum income; and in the second example above, the state's contribution is linear in its revenues since is constant.) A "gap" was defined between federal expenditure per person in a state, Cs, and the maximum per capita spending among all the states, Cm (in the state of Paraná). The total gap or deficit of federal spending for any other state is calculated as Ds (Cm Cs)*Ps. This is the additional spending Compensatory Finance in Health · 141 the state would need to receive in order to reach the same per capita level as the most-favored state m. The sum of these deficits is the total additional spending needed in Brazil to bring per capita federal expenditure in all states up to the level in Paraná: that is, D Cm*P s(Cs*Ps), where P is total population. To complete the allocation formula, one arbitrary parameter was needed, which represents the relative weight of population. The fraction 1 is then the relative weight of the state's deficit in recurrent spending. This parameter is arbitrary in the same sense as s or in the previous examples; it must be decided politically. The difference is that instead of determining federal and state contributions, it fixes the assignment among states of a purely federal fund. Incorporating , the formula becomes Fs/F *(Ps/P) (1 )*(Ds/D). The fraction of the total fund assigned as a ceiling to a state, Fs/F, depends on the state's share of the country's population and on its share of the total deficit or gap in recurrent federal health expenditure. Only the second term is explicitly redistributive, but in the circumstances of Brazil, even proportionality to population improves equity for various states. In the end, a value of 0.70 was chosen for , so the redistributive adjustment has a relative weight of 30 percent. Potential long-term adjustments Obviously, if a formula such as this were applied during a long interval of investment, eventually the value of federally-financed installed capacity per person in health would become equal among all the states. If payment con- tinued to be based on production of services, this would also tend to equal- ize recurrent spending among states. Eventually this might not be desirable, because differences in the epidemiological pattern among regions and states would justify unequal expenditures. Moreover, this adjustment toward greater equity says nothing by itself about the distribution of spending between the federal government and the states (not to mention the munic- ipalities, which in Brazil operate much of the network of health facilities and sometimes contribute substantially to financing it). The subsequent consti- tutional amendment is aimed at equity in this larger sense. Not only should federal investment resources be directed more toward those states where spending has been lower in the past, but states should, by assuming respon- sibility for spending according to their own resources, liberate federal recurrent expenditure for a more equitable distribution. 142 · Health Economics in Development Concluding Reflections These three cases have several features in common which are worth empha- sizing. First, they deal only with geographic equity, leaving for lower levels of decision the allocation of funds among particular interventions, health con- ditions, or population groups. Geographic equity is at least partly a necessary condition for more detailed distributional justice according to needs, but not a sufficient condition. Second, the variables which go into the formulas for distribution--per capita income, revenues, population, or expenditure--are relatively easy to observe, more so in the case of revenue and spending, and do not require additional efforts to gather information. Third, these variables are subject very little or not at all to manipulation by either federal or sub- national governments. (Fraud may be possible, and estimates of population or income can always be controversial, but moral hazard is essentially nonex- istent.) No government at either level can obtain more funds for itself, or reduce its financial contribution, by playing with variables under its control. Fourth, each formula contains one and only one arbitrary parameter: s, and , respectively. As a result, the unavoidable political negotiation can focus on a single number rather than dealing with a multitude of effects or relative weights of several different variables. The resulting formula is transparent and politically comprehensible. It seems reasonable to suggest that these four characteristics are valuable for any allocation formula in pursuit of equity. The case for such simple formulas with a minimum of arbitrary parameters is strengthened when information is limited or of doubtful accuracy, or when relations between national and sub-national governments are characterized by mistrust or incomprehension. CHAPTER 7 An Ounce of Prevention Is Worth How Much Cure? The conventional wisdom in health care is that an ounce of prevention is worth a pound of cure: preventing illness or injury not only avoids pain and suffering, it is also cheaper than treating those ill or hurt. Sometimes this is clearly true, but in other cases it is questionable. At the least, the right balance of prevention and cure in health care spending must depend on the costs of the two approaches, and those costs in turn may depend on the numbers of preventive and curative actions taken. Unless these costs are properly taken into account there is no way to tell how far to emphasize prevention, and at what point to give up and treat the cases that were not prevented. Still less is it possible to judge a priori how expenditure on health care ought to be divided between prevention and cure--and yet it is com- mon to hear the complaint that 80 percent of a government's health budget goes for curative care and only 20 percent for prevention. If prevention really is so much cheaper than treatment, that may even be the right out- come: many cheap preventive actions and fewer, but more expensive, cura- tive treatments. The fact that rice is much cheaper than meat does not by itself imply that rice will, or should, take up more of a family's food budget. In order to minimize the combined costs of prevention and cure--the total cost of keeping people in good health or restoring them to it, suppos- ing that such restoration is possible--the latter should take over from the former at the point where their marginal costs are equal. Here marginal cost means the cost of providing for the next person, whether what is provided is preventive or curative. So long as prevention costs less than cure, everything should be spent on prevention; but if at some point curing Reprinted, with permission, from "An Ounce of Prevention Is Worth How Much Cure? Thinking about the Allocation of Health Care Spending". A View from LATHR. Latin America and the Caribbean Technical Department, the World Bank, 1990. 143 144 · Health Economics in Development people becomes cheaper, spending from that point on should go for cura- tive care. What matters is not average but marginal cost, and these will not be equal if the cost per patient is not constant but depends on the number of patients for whom service is provided. Figure 7.1 illustrates this argument, for the simple case in which preven- tion is always cheaper than cure. The population at risk is N, of whom some number NV receive preventive care: V is the marginal cost of prevention, shown by the curve from VO out to V. The symbol V is chosen to suggest vaccination, which is close to the ideal of prevention in both effectiveness and cost. This cost may initially decline as more people receive preventive care, reflecting the fixed costs and consequent economies of scale as coverage starts to expand. Before the whole population can be covered, however, the marginal cost is likely to rise again, surpassing its initial level. This reflects the increasing difficulty of reaching people as they are geographically scattered, resist being treated, or respond poorly to the Figure 7.1 Marginal Costs of a Curative and a Preventive Intervention, When Prevention is Always Cheaper and Everyone Receives One Intervention or the Other Marginal Cost of the Intervention C C' CO V VO V' 0 NC NV N Number of People Receiving the Intervention An Ounce of Prevention Is Worth How Much Cure? · 145 measure. The total cost of reaching NV people is the area under the V curve, from zero out to NV, and the marginal cost at that point is V . The other, higher curve in Figure 7.1 shows the marginal cost of cura- tive treatment for the same health problem, C. If NC patients are treated the total cost is the area under the curve out to NC, and marginal cost at that point is C . This cost curve may also show an initial decline and a subsequent rise, but it need not be shaped anything like the preventive- cost curve. The analysis is unchanged if the marginal cost C includes the non-medical cost of being sick or hurt, such as loss of income, or even an allowance for pain and suffering. In this example, the cost of prevention never becomes as high as the cost of treatment, even if the whole population is covered. The right policy is therefore to reach everyone possible with preventive care even though this will raise the marginal (and average) cost of prevention above what it would be with lower coverage, since the last people reached are relatively expensive to provide for. The share of preventive spending is 100 percent, whether cov- erage is provided to the whole population or only to a small fraction of it. This probably is the model most people have in mind when they argue that (almost) everything should be spent on prevention. And there are doubtless cases where it applies: measles is easier to prevent than treat, and in an unvac- cinated population, essentially every child runs a high risk of getting measles. This example is far from universal, though. It contains two features which make prevention very attractive. First, the cost of prevention stays low even when everyone is covered: there is no technological barrier to reaching the whole population and the expense of doing so is bearable. Second, it is assumed that everyone who does not receive preventive care will, in conse- quence, require treatment--every individual gets one or the other. Even for measles, this is not strictly true. As more children are vaccinated the likeli- hood that an unvaccinated child will get sick is reduced, because transmis- sion of the disease may be interrupted with less than full coverage by the mechanism of herd immunity. And for most diseases and injuries the prob- ability of requiring treatment is well below 100 percent, even in the absence of any preventive action. Not everyone got polio before a vaccine was found. If the probability of needing curative treatment when one has not received any preventive care is P, then the right comparison to make is not between V and C, the marginal costs per patient, but between V and PC. The latter is the expected expenditure on an individual who gets no preventive care, allowing for the possibility that he may or may not get sick or hurt; put another way, it is the cost per person of treatment when only a fraction of persons become patients. (If only a fraction of the entire population are 146 · Health Economics in Development candidates for the preventive measure--are at risk for the disease or injury it is desired to prevent--there is no need for a second concept of probabi- lity to apply to V; instead, N is simply defined as the population at risk, so that the probability for V continues to be unity.) Now when P is very low, cure must be much more expensive than prevention in order to justify pre- ventive action, 1/P times as high, in fact. This will sometimes be the case-- the cost of treating polio, for example, appears to be high enough to offset the low probability of paralysis among untreated children--but not always. Figure 7.2 illustrates the two complications just discussed. First, the pre- ventive cost curve VO-V starts out low, just as in Figure 7.1, but then begins to rise so steeply that it becomes essentially impossible to cover the whole population at risk. Second, the curative cost curve is shifted down by the fraction P, to the position PCO-PC. The result is that the two curves now Figure 7.2 Marginal Costs of a Curative and a Preventive Intervention, When Only Part of the Population Requires the Curative Intervention Marginal Cost of the Intervention PC V' PCO VO 0 NV' N Number of People Receiving the Intervention An Ounce of Prevention Is Worth How Much Cure? · 147 cross: beyond the point NV it would be cheaper to start curing patients than continue to work at prevention, even though it is much cheaper to prevent NV cases than to give curative care to the same number of people. For simplicity, the Figure is drawn supposing that P does not itself depend on the number of preventive actions. (Allowing for that possibility would generate a whole family of curative cost curves, each corresponding to one level of preventive coverage, which could cross the preventive cost curve at different places. Alternatively, the diagram would need to be three-dimensional, with the third dimension showing how P is related to preventive coverage.) Figure 7.2 may seem to show that NV is the correct level of preventive care, but of course that is not the case: costs are compared as though all individuals received both preventive and curative measures with the latter becoming cheaper for the next person after NV . In fact, the expected cost of treating the first person who is not covered by prevention is shown by the height of the PC curve at PCO at the left margin of the Figure. This will be equal to the marginal cost V at NV only if the cost curve is flat, with marginal and average costs equal. To find the correct (cost-minimizing) level of preventive action when the curative curve is not flat, the curve PCO- PC is shifted (added) horizontally to the right until its starting-point meets the preventive cost curve. This is done in Figure 7.3, which shows the correct level of prevention to be NV*. To the left of V* the curve VO-V* shows preventive costs and to the right the curve V*-PC shows curative costs. The first person for whom curative care is expected to be needed will cost exactly as much to take care of as the last person covered by preventive measures. In this example, prevention turns out to be the cheaper approach for much more than half the population at risk, but the total cost of cura- tive care may still exceed total expenditure on prevention: the area NV*-V*- PC-N may be greater than the area O-VO-V*-NV*. This distribution of resources is not evidence of inefficiency; it simply reflects the low probabil- ity of needing cure, and the sharply rising marginal cost of prevention. (The shaded, almost triangular area between the curve V*-V and the curve from V* toward PC shows the additional cost that would be incurred by carrying preventive action too far, to the intersection of costs shown in Figure 7.2. A similar triangle to the left of V* would show the extra cost associated with stopping the preventive measure too soon.) Figure 7.3 shows how both total expenditure and its allocation between prevention and cure vary, as the total number of people reached by one measure or the other increases. It is important to note that so far as cost min- imization is concerned, every point on the curve VO-V*-PC is an optimum, 148 · Health Economics in Development Figure 7.3 Cost-minimizing Distribution of Effort between a Curative and a Preventive Intervention Marginal Cost of the Intervention V' PC V* PCO VO 0 NV* NV' N Number of People Receiving the Intervention representing the cheapest way of reaching a given number of people. It is only in terms of health benefits, that one point can be said to be better than another--benefits increase steadily as one moves from VO toward PC. Whether those benefits are sufficient to justify the costs incurred is a larger question, which is considered briefly below. At this point, what matters is that the right balance of spending between preventing a health problem and curing it, depends on how much in total is spent on that problem. At low lev- els of total expenditure, the correct allocation may be 100 percent for pre- vention, but when enough is spent to do something for everyone at risk, the optimum share for prevention may be much lower, as in the Figure. The argument that too much is being spent on curative care, particularly in poor countries, is often implicitly an argument either that the situation in Figure 7.1 prevails, and nothing should or need be spent to cure people; or, that in the situation depicted in Figure 7.3, the country cannot afford to help An Ounce of Prevention Is Worth How Much Cure? · 149 everybody, that budgetary constraints impose an optimum close to--perhaps a little to the right of--V*, and far from PC. Does the real world ever look like what Figure 7.3 shows? In at least two important instances it does. One case is exemplified by diseases like malaria, where the preventive measures available work imperfectly--eradication of vectors is possible only in unusual circumstances, and is becoming more difficult, it is hard to avoid the vectors, and there is as yet no effective, cheap vaccine--and costs can become exorbitantly high as coverage is expanded. These problems do not mean that prevention should be abandoned, only that curative care is also part of a cost minimizing solution, and may have to absorb a large share of total costs. (Just how large depends also on the defini- tion of the population at risk, N.) The other typical case is exemplified by vehicular accidents. It is known how to reduce death and injury drastically by very inexpensive means such as always wearing seat belts and never driving while intoxicated. But these measures require repeated individual compliance: there is no vaccine yet against pigheadedness or drunk driving. Some people comply readily with the required preventive actions, producing a very low initial cost V, while others are impervious to safety propaganda and can be reached only by much more expensive measures such as air-bags or saturating the roads with police who will enforce the (appropriate) laws. It is these drivers who make the preventive cost curve rise so steeply. So do other elements outside the control of individual motorists, such as the high cost of making roads safer and cars more crashworthy. Nobody even knows what the cost curve looks like with any certainty: giving more money to the Ministry of Health (or anyone else) to reduce vehicular accidents does not translate into fewer vic- tims as readily as more money for polio immunization translates into fewer paralyzed children. Hence the high, and growing, share of expenditure to treat accident victims. Both these examples suggest that what is needed is not just more expendi- ture on existing preventive measures--and certainly not just less expendi- ture to treat victims--but the discovery of better prevention, cheaper and/or surer to work. That is why so much effort goes into developing new vaccines, and why governments experiment with laws, regulations and educational campaigns to reduce drunk driving, smoking, and other risks for which the curative costs can be very high, even, as in the case of AIDS, infinitely high. New preventive measures may eventually transform the optimum balance between prevention and cure, just as vaccines have done for measles and polio; and that in turn can be expected to redistribute 150 · Health Economics in Development spending between the two types of care. But at least when total expendi- ture is high enough, the outcome may be that curative care will properly get a larger, not a smaller, share of total health care resources. That will be the right solution, if effective prevention becomes very cheap for almost everyone, but still cannot reach some people--while curative treatment remains costly. Meanwhile, the relative spending on prevention and cure by itself says nothing about whether preventive efforts are inadequate, excessive, or about right: one must get beyond budget shares to relative marginal costs. Strictly speaking, this analysis is valid only for a given health problem, which can be prevented (with more or less success) and for which some effective treatment exists (with greater or lesser probability of restoring full health). For that particular problem marginal cost analysis can in principle help determine the balance of preventive versus curative spending, to mini- mize the total cost of a given health outcome or to maximize the number of cases prevented or cured, for a given expenditure. Under these conditions, costs do not have to be compared to any measure of benefits, if one takes as given the budget for total spending on that problem. The preventive/cura- tive distribution of total health care expenditure, however, depends also on the distribution of health problems on which money is spent. It is not pos- sible to determine how much to spend in total on each kind of care without comparing the seriousness of different problems or, equivalently, the bene- fits of preventing them or curing them. If vehicular accidents are more of a problem than measles, there will be more spent on curative care, in the opti- mum situation, than if measles were the more serious problem. Judging how serious a problem is means explicitly considering some measure such as loss of life, of years of life, or of quality-adjusted life years (which can also take account of non-fatal morbidity). The combination of such a criterion for comparison among different health problems and the marginal-cost criterion for allocation for each particular health problem, is in principle sufficient for specifying the optimum allocation of total health care spending between prevention and cure--again, for a given total expenditure on all types of health care. Arguments that too much is spent on curative care are often, implicitly or explicitly, arguments that too much is spent on certain problems, for which effective preventive measures do not exist or for which the known curative treatments are very costly. Such judgments can only be based on some comparison of benefits. For a given health problem, prevention which is cheap and effective but cannot--for whatever reason--reach everyone in the population is likely to An Ounce of Prevention Is Worth How Much Cure? · 151 lead to an optimum in which curative care for a small number of people requires spending more than preventive measures for a much larger number. "Success" does not necessarily raise the preventive share in total spending, unless the prevention covers everyone at risk. The same phenomenon prob- ably operates even more powerfully between health problems: successful pre- vention of one problem, such as measles, means that some other problem, such as vehicular accidents, becomes more important, with the consequence that the curative share of total spending should go up--at least until vehicular injury becomes as easy and cheap to prevent as measles is. At each moment, the correct balance of prevention and cure depends on total spending, on the epidemiologic situation and on the state of technical knowledge about how to prevent and how to cure health problems. The "epidemiologic transition" is largely just the product of successful prevention, whether by medical means (vaccination) or others (better diet, safer water), and it has cost conse- quences. How much should be spent in total on health care depends on com- paring health benefits to the--incommensurable--benefits from spending on other things: there is no consensus on how to make that comparison. It may be unwise to guess at a long-run equilibrium of health care cost levels and composition, since technical progress is rapid and there is little consensus on the seriousness of different health problems. But it is at least plausible that the correct distribution of spending in the long run will be even more weighted toward curative care than current spending is-- because almost no one will die or get very sick at early ages, thanks to pre- vention, whereas the health problems people experience at later ages will not be amenable to prevention or even to cheap treatment. This might be the case even if society spent a smaller share of total product on health care--for example, if it stopped spending on "heroic" measures to prolong life by a few days or months, and spent nothing on people in comas or with Alzheimer's disease. A high proportion of spending on curative care may be the inevitable consequence of the fact that certain diseases and injuries can be prevented, but death can only be postponed. CHAPTER 8 DALYs and Cost- Effectiveness Analysis Acknowledgments This note is the direct result of a request from Maria Paalman, the senior author-editor of the Review1 {to which this article responded}, to read and comment on it. I am grateful to her for that impetus, and to Abdo Yazbeck, who worked on the 1993 World Development Report, for information and suggestions. Any misinterpretations are my sole responsibility, and in particular do not necessarily represent the views of the co-authors of the WDR nor of the World Bank. Introduction The article by Paalman et al. (1998) constitutes a fairly thorough and gen- erally helpful guide to what is meant by the `burden of disease' presented in the World Bank's 1993 World Development Report, Investing in Health (WDR), and how it is calculated. To a slightly lesser extent, the article is also useful for understanding how cost-effectiveness was estimated and used for setting disease control priorities,2 and how these results were used in the WDR. As the title indicates, the piece is a critical review of these exercises and not simply an explanation of them. In this respect it is less acceptable, since some of the criticisms are misguided, a few are couched in somewhat irresponsible or evasive language, and some of them are plain wrong. This rejoinder proposes to sort out which criticisms are sound and which are not, Reprinted from "A Critical Review of `A Critical Review': The Methodology of the 1993 World Development Report, `Investing in Health'." Health Policy and Planning 15(1), 2000, pp. 110­115, by permission of Oxford University Press. 153 154 · Health Economics in Development or less sound, and thereby to clarify further what the WDR says and why it says it. I had the good fortune to be one of its authors, and so cannot claim neutrality in this discussion; but the experience of working on the Report and enduring much criticism of it even before it was published,3 as well as in the 6 years since, may illuminate some of the issues raised. It is conven- ient to group the observations of the review by Paalman et al. (henceforth referred to as the Review) under four headings: · the choice of cost-effectiveness, cost-utility or cost-benefit analysis; · the various subjective parameters used to calculate Disability-Adjusted Life Years (DALYs); · using the burden of disease and cost-effectiveness estimates (and other criteria) to set priorities; and · things the WDR might have undertaken, but did not. Choice and Appropriateness of Analysis There are three distinct ways to compare the cost of some program or activ- ity to its outcome: the analysis of cost-effectiveness (CEA), cost-utility (CUA) or cost-benefit (CBA) ratios or, in the case of CBA only, differences. The Review says that `of the different economic evaluation techniques, the method used in the WDR is cost-utility analysis, a type of cost-effectiveness analysis used in the health sector' (p. 19). This is mostly right, although CUA need not be restricted to the health sector, and CUA is not so much a type of CEA as it is an extension of the latter that moves from purely objective meas- ures of effects such as cases of disease, or lives, or people, to measures such as Potential Years of Life Lost (PYLL), Quality-Adjusted Life Years (QALYs) or DALYs, which explicitly introduce one or more subjective parameters in order to value the outcome. The distinction is clear and useful (Hurley et al. 1998), and it is perhaps regrettable that the WDR did not employ it; but there is a long tradition of using CEA to mean both things, and there is no con- venient adjectival form of CUA comparable to `cost-effective'. The choice of CUA for the WDR meant that only health outcomes would be considered among the effects of an activity, precluding fuller evaluation of `other sector interventions which are known to impact on health, such as girls' schooling, water supply and sanitation, and increased DALYs and Cost-Effectiveness Analysis · 155 food consumption', as well as evaluation of the non-health effects of family planning (the Review, p. 14). The Review says of these interventions that `in terms of total welfare they may be very cost effective', but this is a con- fusion of terms, just as it is to say that `this has the effect of reducing the overall impact of an intervention, and thus reducing the resulting cost- effectiveness ratio' (p. 21). Such comparisons can be made only by assigning monetary values to the outcomes and applying CBA, as the Review recog- nizes elsewhere (p. 20). That raises problems of valuation far more vexing than those involved in CUA, such as whether to value results according to people's willingness to pay for them, which is dependent on the distribution of income and other factors that determine whether people act as though the intervention were worth its cost. Such comparisons are difficult enough within a single country, and would become almost meaningless across coun- tries with very different incomes. The WDR authors chose to avoid these problems and accept the resulting limitations, and that choice cannot be criticized without implicitly favoring the monetization of benefits. Subjective Parameters: Life Expectancy, Disability, Discounting and Age-weights Even asking how much life is lost by a premature death, as in PYLL, requires choosing some life expectancy to define `premature'. And any sum- ming-up of mortality and morbidity, disability, or other notion of non-fatal health loss requires assigning weights to states between death and perfect health. Thus at least two of the four subjective parameters incorporated in the burden of disease calculations are inescapable in any such exercise, and the only question is what values they should take. There is a more subtle reason why the future should be discounted to some degree, which also makes a third parameter essential to avoid problems. Life expectancies The Review admits that it is `ethically appealing' to assign potential lives of 80 or more years to everyone in the world (p. 16), but complains that `the use of high standard life expectancies leads to very high burdens of disease in countries which presently have considerably lower life expectancies', so that `it might be better to use realistic life expectancies when setting national priorities'. This is, regrettably, a common confusion: people dying 156 · Health Economics in Development young is clearly a `burden of disease', or the phrase does not mean anything, and it is wrong to imply that the high burden of disease in poor, short-lived populations is somehow an overestimate. Probably what the authors meant, but did not distinguish, is that the amount of healthy life that could be gained by eliminating that disease alone might be relatively small, because in an unhealthy environment the people whose lives were saved would still die young from something else. That is, the effectiveness of interventions might be exaggerated, even if the burden were correctly estimated. This is a plausible-sounding argument, and would imply the need for some adjustment when passing from the current burden to the potential gain from intervention. But it runs into a paradox which hardly any critics of the WDR have noticed: clearly if all the diseases contributing to the burden were controlled to the same extent that they are in Japan, life expectancy would rise to 80 or more years, just as in that country. (It does not matter that the cost of achieving such control, and the best choice of activities to bring it about, might differ from country to country.) Thus the gain from controlling all the diseases together would greatly exceed the sum of the gains from controlling each one separately, using the current short life expectancy. This adding-up paradox is a further reason, independent of the equity argument or the desirability of uniform international compar- isons, for using the same long life expectancies everywhere when estimating the burden of disease. Citing Anand and Hanson (1995), the critical review proceeds to another and much blunter confusion: the argument that if high life expectancies are used, `the global burden of disease not only measures the burden of disease, but also indicates a burden of `underdevelopment'.' It should be superfluous to point out that by definition, every cause of death is classified as some sort of `disease', taking the term to include perinatal conditions, malnutrition, and all forms of violence or external causes. Nobody dies directly of under- development. The notion that there is a burden of disease up to a country's current life expectancy, and that the difference in years of life lost from there out to an expectancy of 80 years or more is due to something else, is simply wrong. The right way to think about underdevelopment is as a collection of risk factors, and it is true that disaggregating the burden of disease by risk factors is a much more complex undertaking than distinguishing the contributions of individual diseases. That is why the WDR attempted such an analysis for only a handful of risk factors: malnutrition (WDR Table 4.3), poor household environments (Table 4.5), and some other environmental threats (Table 4.6). These estimates are DALYs and Cost-Effectiveness Analysis · 157 more preliminary and less thorough than those for particular diseases, and may be substantially revised when the World Health Organization finishes publishing the complete Global Burden of Disease series now in progress. Disability weights In contrast to the criticism of the other subjective parameters, the Review does not quarrel with the individual weights assigned to different disabilities-- there are, after all, many of them, grouped into six classes--but focuses on how those weights were arrived at. In particular, the authors question the use of expert opinion, rather than that of patients or the general population, and complain that `some of the criticisms might have been prevented if the process by which disability weights were established had been published and openly debated' (p. 18). It is certainly true that people with different relations to a disability--having suffered it, or treated it, or only thought about it--may evaluate it differently, but that does not establish whose view is more to be accepted; by definition, there are no right answers to these subjective elements, only answers that are reasonably consistent and defen- sible. And given the very tight deadline for producing the estimates that went into the WDR, there was no possibility of debate thorough enough to satisfy all the potential critics of the procedure. There was really no alter- native to publishing the results and thereby opening them for discussion. The Review points out that there is no allowance for conditions worse than death, but while one might admit there are such cases, they are surely rare enough, and often so short-lived as to make little contribution to the overall burden of disease. A more serious criticism is that severity and dura- tion of disability are estimated separately, and the burden is simply the prod- uct of these two characteristics; there is no allowance for learning to live with a disability so as to suffer less from it, or for suffering more because there is no hope of cure or remission. This is an instance of a general issue that is confronted repeatedly in the definition of DALYs, which is that of linearity. The burden of disease adds linearly across individuals of the same age who suffer death or the same disability at the same time: this has the great com- putational advantage of permitting one to work with totals or averages. In contrast, DALYs are not linear with age, because of age weighting, nor through time, because of discounting; a year of disability differs in value according to when it occurs and the age of the person to whom it happens. Rather than imposing linearity everywhere, which would greatly simplify the calculations, or allowing everywhere for non-linear relations, the creators of 158 · Health Economics in Development the burden of disease estimates introduced nonlinearity wherever there seemed to be a strong specific justification for it, and not otherwise. Finally, it is true that the assigned disability weights are uniform across cultures and circumstances, even though the ease with which someone can cope with a disability differs from place to place and person to person. Uni- formity was imposed here partly for computational convenience, but also for equity reasons, so as to make no distinction among people according to where they happen to live and what resources they have for dealing with disabilities. Discount rate Since the burden of disease incident in any one year includes all the future effects of death and disability occurring in that year, it is necessary to decide how to value the future. The effects of choosing a high or low discount rate on the composition of the burden of disease are correctly described, as is the reason for not using short-term rates on risky investments. However, the discussion of this issue in the Review is unfortunately rather evasive. Thus, discounting `purportedly' converts future lives to their present values, and `is said to' avoid the time paradox (p. 18)--it is not recognized that it does avoid that paradox--and the paradox itself is consigned to a footnote. In a world where resources can be invested at a positive interest rate, if one does not discount future benefits, it will never pay to undertake an intervention because each year's delay means the benefit can be increased, for example by reaching more beneficiaries. The rhetorical question, `Whose perspective and values should be considered?' ignores the fact that `the long-term [aver- age] yield on investments' is determined not by a cabal of experts but by all of society, which behaves as though it does discount the future at some rate not too different from 3%. Age weights These are easily the most controversial of the subjective parameters used in the calculation of DALYs, and as the Review says, `neither universally accepted nor valued' (p. 17). This is partly because age weights are not intrinsically necessary for comparing mortality and disability, measuring years of life lost, or avoiding time paradox: this makes them the most optional of the subjective choices involved in going from CEA to CUA. However, a substantial part of the rejection they encounter is due to mis- understanding, in particular to supposing that the weights are determined DALYs and Cost-Effectiveness Analysis · 159 by relative economic productivity. If they were, they would stay close to zero up to age 15 or so, reach their peak much later than age 25, and drop sharply at retirement age. They would also discriminate according to peo- ple's income-earning ability, so that their use to set priorities for health care would favour the rich--thus the `obvious inequity' to which Murray (1994) refers. There is nothing inconsistent about rejecting such an economic measure and instead using age-weights that are designed to take account of the emotional loss to other people from an individual's death. The weights reach their peak at 25 years, because a person that age is likely to have a spouse and young children, and parents still living. They are low at very early ages because parents can often replace a lost infant, before they age much themselves, and low at high ages because one's parents are usually deceased and one's children grown by then. It is inconsistent to criticize parameters that try to incorporate the loss to those related individuals, and at the same time complain that DALYs do not include any of `the burdens which fall on households' (p. 18). More generally, treating all ages equally is also a subjective choice, and while it is simpler and more `classical', it is not obviously better. Part of the sensitivity analysis conducted by Murray et al. (1994, p. 103) involves vary- ing the age weights from the pattern used in the WDR, through three inter- mediate values with the same general shape but smaller age differences, to equal weight for all ages--not, as the Review says, `shifting to more unequal age weights' (p. 19). Changing the age weights has complex effects on the composition of the burden of disease by age and by disease, effects partly compensated by the discount rate, but with essentially no effect on the com- position by sex. There is no clear right answer to how to treat different ages, since it depends on how much one wishes to take into account emotional losses suffered by people close to the disease victim. Actual Choices in Burden of Disease Estimates As of mid-1998, there had been, in addition to the global estimates used in the WDR, 15 exercises in national priority setting in health in 17 countries and one large Indian state, which included estimates of burden of disease (Bobadilla 1998, pp. 12­15). Another such study was in preparation, and there had been another four efforts to determine priority interventions without estimating disease burden. These exercises covered as few as 10 major diseases and as many as 112. Several of them ignored morbidity and 160 · Health Economics in Development disability and estimated only the (much larger) health damage from prema- ture mortality. `Although the indicator to assess the burden of disease most commonly used was the DALY, half of the countries made significant mod- ifications, and one (Russia) used a different indicator . . . Five countries did not discount future health losses, and . . . Five countries did not apply any age weights to health losses or used different weights from those proposed in the 1993 World Development Report. Finally, three of the countries that studied disability used a different disability scale to assess severity. In short, only seven of the 19 study results are comparable . . .' (Bobadilla 1998, pp. 17­18 and Table 3). This varied experience shows that what is proposed in the WDR is not universally imposed, for at least three reasons--a disease burden dominated by mortality from a small number of diseases, inadequate data (a problem on which the Review rightly insists) and subjective choices which depart from the logic outlined above. While this may be interpreted as a victory for local views and values, there is no reason to think it makes the different national studies any better than if they had all used the same methods and numerical values. Priorities for Intervention The Review summarizes well (p. 14) the idea that an intervention is a `good buy' if it is cost-effective and also deals with a disease or risk factor that accounts for a sizeable burden of disease. In strict economic terms, the lat- ter condition should be irrelevant: if at the margin, an intervention deals cost-effectively with a small disease problem, then it is worth buying. At least, this will be true so long as there are no large fixed costs to confronting small disease burdens, and the health care system can deal simultaneously with a large number of interventions. The reason for including the size of the burden as a criterion for priority setting is precisely that health systems in poor countries often cannot efficiently administer a large collection of programs, and dissipate their resources trying to do so. In consequence, it makes sense to maximize the health gains from a small number of interven- tions, economizing on scarce managerial and administrative capacity. There are, as the Review notes (p. 22), some very difficult issues of whether disease burden and cost-effectiveness criteria can be used for large-scale reallocations or only for marginal increments of resources, and of what the health system is currently doing or trying to do with its existing infrastruc- ture, which cannot easily be re-directed. These difficulties mean that a DALYs and Cost-Effectiveness Analysis · 161 simple ranking of cost-effectiveness cannot be treated as a list of priorities, even ignoring other considerations, unless the health system can--subject to overall resource limitations--actually offer any of the interventions on the list without interfering with the capacity to administer others.4 The Review observes, correctly, that when a disease is expanding rapidly, the current burden it causes is a poor guide to priority (p. 23, citing Bobadilla et al. 1994). But it does not recognize that the WDR applied this reasoning to argue `Why AIDS is a special case', requiring urgent preven- tive measures to contain its spread (pp. 99­106). Nor do the authors of the Review take any account of the discussion in the WDR (pp. 54­59) of the rationales for government action in health other than that of obtaining value for money. Cost-effectiveness is, almost tautologically, the criterion for getting the most health gain out of a given quantity of resources, how- ever the effect is measured. It is not necessarily the appropriate criterion for how public resources should be spent, except in the case of public goods for which markets do not exist, or services with such large externalities that markets will work quite imperfectly (Musgrove 1999). The emphasis in the WDR on disease burden and cost-effectiveness constitutes its principal novelty, and, not surprisingly, has drawn much more attention from the public health community than the more conventional discussion of public finance criteria and of equity. Equity is a troublesome issue, because cost-effectiveness is concerned only with efficiency, and `the fact that the most efficient interventions . . . tend to specifically benefit the poor is more a result of coincidence than of principle' (p. 24), a finding that has since been more fully substantiated (Gwatkin and Guillot 1998). And neither horizontal nor vertical equity is necessarily compatible with cost-effectiveness (Musgrove 1999); societies must choose how far to favor one or the other goal. Nonetheless, govern- ments can do much to promote equity by how they finance health care, even if the interventions are chosen on the basis of value for money. And the essential package which the WDR recommends was designed specifically for very poor countries. At higher income levels, it becomes much harder to say what should go into a package, and cost-effectiveness becomes less relevant as a guide. Giving some degree of priority to the poor, either because they are sicker than the non-poor or because they are less able to pay for health care, is ethically acceptable and is explicitly favored by the WDR. It does not follow that other distinctions among potential beneficiaries are equally justified. Thus giving `greater weight to mothers and children than to adult males', 162 · Health Economics in Development as the Review says was done in the Oregon exercise to determine what interventions Medicaid would finance (p. 17), is prima facie inequitable, par- ticularly since the greater life expectancy of children already gives them a measure of priority over adults, where death or long-term disability are concerned. And to judge that, ceteris paribus, a year of death or disability matters more for a woman than for a man is breathtakingly unfair. Priori- ties among population groups should not be imposed as part of the estima- tion of disease burden, but should emerge from the analysis of what health gains can be achieved, at what cost, given those groups' disease profile. That is the greatest conceptual advance in the WDR, the insistence that on grounds of both equity and efficiency `it is appropriate to . . . prioritize health care interventions rather than population groups' (p. 23 of the Review). If this contradicts the common emphasis on `vulnerable groups', so much the worse for that emphasis: everyone is vulnerable to some health problems, and not every vulnerability has a solution. The way to combine information on the main elements of the disease burden of particular age and sex groups, with information on the availability of cost-effective inter- ventions that could significantly reduce that burden, is well illustrated in Appendix Tables B.6 and B.7 of the WDR. That exercise shows that chil- dren and women do indeed deserve priority for certain health problems, but not universally or independently of what diseases they suffer. Thus control- ling childhood diarrhoea and respiratory infections should have priority, but dealing with congenital malformations probably should not. The Review criticizes the idea of an essential package for poor countries, including a small number of interventions, on the grounds that it `is not in line with the Alma Ata approach to comprehensive care, which has been pushed by WHO and donors for many years now', and that `the call for spe- cific packages could lead to vertical programs' (p. 24), as though it were demonstrated that vertical programs are a bad thing and that the Alma Ata approach has been shown to work adequately.5 This sounds more like cling- ing to poorly justified positions than like real analysis. A more serious criticism is that `the package arrived at by experts might not be acceptable to the public, and hence less cost-effective in the end' than if people's wishes were consulted in formulating priorities. This raises the general, and severe, problem of needs and wants not necessarily coinciding, and points to the urgency of finding out whether and why people may not use existing health services and reforming those services so as to attract the beneficiaries they are meant to help. The Review correctly emphasizes that the actual cost- effectiveness of an intervention depends on the scale of utilization and the DALYs and Cost-Effectiveness Analysis · 163 degree of compliance by patients, which will vary with local cultural, eco- nomic and other conditions. The Review is the result of conflating and editing separate papers by the four authors.6 This may explain why it says in one place (p. 24) that `Once the most cost-effective interventions are in place the burden of disease is expected to decrease', and then elsewhere (p. 25) claims that `implementa- tion of the advocated good buys will most likely not result in a reduction of the burden of disease, but in the transition of disease patterns and increased health care costs'. Certainly the former view is correct, or it makes no sense to do anything about health problems at all. The fact that the cost of gain- ing a DALY will rise if the cheapest DALYs are bought first is an indication of success, not of failure; it would still be true that the greatest number of DALYs could be bought for any given expenditure, or that for a given reduction in disease burden, the cost would be minimal. All that the WDR intended to do in this regard was to emphasize the huge differences in cost- effectiveness among known interventions--differences that are often much larger than the likely errors in estimation--and the large disease burdens that could be controlled at relatively low cost (although possibly requiring expenditures significantly larger than the poorest countries are now dedi- cating to health). What the WDR Might Have Done, But Did Not Unlike most publications from the World Bank or elsewhere, the annual World Development Report is produced on a timetable that does not allow even one day's deviation, and with a rigid ceiling on the number of pages. These constraints precluded doing a great many things that would have made for a more comprehensive volume.7 A good deal of the criticism in the Review concerns such desirable but infeasible improvements as: · generating more and better data on disease incidence, and especially on intervention costs and effectiveness, since the existing base of informa- tion is incomplete and includes many estimates and approximations; · performing more extensive sensitivity analysis to see whether any of the conclusions or recommendations would be overturned by plausible changes in empirical data or in the values of the subjective parameters; · incorporating more views into the estimation of disability weights and age weights, and forming a wider consensus on their values; and 164 · Health Economics in Development · providing more help to countries wishing to use the recommendations of the WDR to restructure their health systems and reallocate resources, including `practical advice for Ministries of Health on how to deal with political barriers' (p. 24). The WDR was not intended as the last word on any of these issues; on the contrary, it was meant to stimulate more and better data and analysis, by showing what could be done even with the limited information available. And it could not, without becoming a multi-volume encyclopedia, deal with the specific economic, political and cultural circumstances of all the soci- eties where its recommendations might be applicable. (A flood of country studies and project documents in the last six years have been dedicated partly to that end.) In particular, trying to give global political advice would be even more open to criticism than the global suggestions based on narrow economic criteria. Many of the recommendations in the Review for further research and refinement of recommendations are entirely compatible with the aims of the WDR and the views of its authors. However, no volume of data and no amount of analysis will do away with the need for subjective judgment about key parameters. More sensitivity analysis can at best show whether conclusions hold up under different numerical values--it cannot determine the right values. And wider debate about those values will not necessarily lead to agreement, as appears from the variety of assumptions adopted in the burden of disease exercises described above. More precision and clarity are possible, at a cost, and very desirable; certainty never will be. That is why the effort launched by the World Development Report is likely to continue for years and may gradually transform how people think about health sector priorities, without ever providing final answers. Concluding Remarks The authors of the Review are nothing if not critical; I hope that this reply will clarify where those criticisms are well-founded and where they are not, including the few instances of outright error. Still, the Review concludes that `the World Bank is to be commended for this unique initiative, conducted in a self-critical spirit' (p. 25). In the same spirit, I believe the authors of the World Development Report can commend the thoroughness of the critical review which their work has received, and the obvious willing- ness of the reviewers to understand, explain, and improve on it. DALYs and Cost-Effectiveness Analysis · 165 Notes 1. Maria Paalman, Henk Bekedam, Laura Hawken and David Nyheim. A critical review of priority setting in the health sector: the methodology of the 1993 World Development Report. Health Policy and Planning, 1998, 13(1): 13­31. 2. In Jamison et al. 1993. 3. It is probably safe to say that every single issue raised by Paalman et al. was debated extensively, and sometimes acrimoniously, within the Bank and WHO; arguments between orthodox economists and public health specialists were often particularly sharp. 4. This is one of the crucial issues treated in Better Health in Africa (World Bank 1994), which is in many ways a companion volume to the 1993 World Development Report. 5. Quite the opposite occurred in Bangladesh, where the development of an essential package of services led to the consolidation of more than 100 vertical pro- grams, 66 of them supported by donors (Abdo Yazbeck, personal communication, June 1999; the details are in reference World Bank 1998). 6. Maria Paalman, personal communication, June 1999. 7. But as the author of Parkinson's Law (Parkinson 1957) says, referring to many things that were left out of his book, `such a volume would take longer to read and cost more to buy', besides taking much longer to write. References Anand S, Hanson K. 1995. Disability-Adjusted Life Years: a critical review. Working paper series Number 95.06. Harvard Center for Population and Development Studies. Cambridge, USA: Harvard School of Public Health. Bobadilla JL. 1998. Searching for Essential Health Services in Low and Middle-Income Countries. Policy Background Study Number Soc­106. Washington, DC: Inter-American Development Bank. Bobadilla JL, Cowley P, Musgrove P, Saxenian H. 1994. Design, content and financ- ing of an essential national package of health services. In: Murray CJL, Lopez AD (eds). Global Comparative Assessments in the Health Sector: Disease Burden, Expenditures and Intervention Packages. Geneva: World Health Organization. Gwatkin D, Guillot M. 1998. The Burden of Disease among the Global Poor: Current Situation, Trends, and Implications for Research and Policy. Prepared for the Global Forum for Health Research. Washington, DC: World Bank. Hurley J, Feeney D, Giacomini M, Grootendorst P, Lavis J, Stoddart G, Torrance G. 1998. Introduction to the Concepts and Analytical Tools of Health Sector Reform and Sustainable Financing. Centre for Health Economics and Policy Analysis, 166 · Health Economics in Development McMaster University. Toronto, Canada. Written for the Economic Development Institute of the World Bank. Murray CJL. 1994. Quantifying the burden of disease: the technical basis for dis- ability-adjusted life years. In: Murray CJL, Lopez AD (eds). Global Comparative Assessments in the Health Sector: Disease Burden, Expenditures and Intervention Packages. Geneva: World Health Organization. Murray CJL, Lopez AD, Jamison DT. 1994. The global burden of disease in 1990: summary results, sensitivity analysis and future directions. In: Murray CJL, Lopez AD (eds). Global Comparative Assessments in the Health Sector: Disease Burden, Expenditures and Intervention Packages. Geneva: World Health Organization. Musgrove P. 1999. Public spending on health care: how are different criteria relat- ed? Health Policy, 47. Parkinson CN. 1957. Parkinson's Law. Cambridge, UK: Houghton Mifflin. World Bank. 1994. Better Health in Africa: Experience and Lessons Learned. Washington, DC: World Bank. World Bank. 1998. Bangladesh Health and Population Program Project, Project Appraisal Document, Report No. 17684-BD, June 1. Washington, DC: World Bank. CHAPTER 9 Criteria for Public Spending on Health Care Acknowledgments This note grew out of an impromptu lecture during Module 5 (`Designing a Cost-Effective Benefit Package') of the Flagship Course in Health Econom- ics and Sustainable Financing, organized by the Economic Development Institute (now the World Bank Institute) of the World Bank in October- November 1997. I am grateful to Ricardo Bitrán, who developed Module 5, for the opportunity to participate in that part of the course and for his reac- tions to the ideas presented here; to my colleagues Paul Shaw, Anne Johansen and Hadia Samaha; and to the audiences for the Flagship Course in Wash- ington and at seminars or courses in Moscow, Santiago, Bogotá, Brasília and Riyadh during the next several months, for opportunities to pursue these ideas and improve their logic and presentation. Julio Frenk, Lant Pritchett and an anonymous reviewer also provided helpful comments. Any remaining errors or confusions are no one's fault but the author's. Introduction It is standard practice in courses on health economics to explain each of the several justifications for the state to intervene in this sector. These include the different kinds of market failure that can occur in health care interventions and insurance, together with the equity reasons for inter- vention, such as to assure some minimum coverage of health insurance or Reprinted from "Public Spending on Health Care: How Are Different Criteria Related?" Health Policy 47, 1999, pp. 207­223, Copyright 1999, with permission from Elsevier Science. 167 168 · Health Economics in Development access to health care for the poor or for some other group. What is almost always missing from such a course is any explanation of how these differ- ent criteria are related to one another. In particular, the participant seldom gets any help understanding two kinds of potentially very important rela- tions between one criterion and another. One question is whether they are compatible with each other or are in conflict, requiring one to choose between them in deciding how to use public resources. The other question is whether they are connected sequentially or hierarchically, so that one should examine one criterion before asking if another one is applicable. This note asks and answers these two questions--about compatibility and about hierarchy--for nine frequently used criteria. Working out the connections among these criteria leads to a partial decision tree concerning three possible outcomes: a particular service should be financed publicly (or subsidized in part), it should be left to private markets to provide or it should not be produced at all. Helping to reach such decisions is one of the main goals of public finance theory [1], and making the decisions is among the principal tasks of government. Which Criteria Matter? Fig. 9.1 displays the nine criteria, and classifies four of them as being pri- marily economic (about efficiency), four primarily about equity, and one as neither of these. To keep the analysis as simple as possible, each criterion is considered chiefly as a reason for spending public money on health care, or choosing how to spend it, rather than as a justification for other, non- financial kinds of government intervention. The economic efficiency criteria start with cost-effectiveness, or the rela- tion between the cost of an intervention and the resulting outcome or health gain. Two reasons for market failure which occur primarily in the market for health care are also included: health interventions which are public goods and therefore have no market or which produce such marked externalities that markets will not produce the efficient amount of the serv- ice. The fourth efficiency reason is the catastrophic cost of some health care, which is an argument for risk sharing via insurance. The equity criteria include both horizontal and vertical equity. Both these concepts of fairness are considered only in relation to health care, not to how health services are financed, although the horizontal/vertical distinction also applies there. Poverty is another criterion in this group; that is, the emphasis Criteria for Public Spending on Health Care · 169 Figure 9.1 Nine Criteria for Public Spending on Health Care Catastrophic Cost Poverty Externalities Vertical Public Equity Goods Cost-Effectiveness Public Horizontal Demands Equity Rule of Rescue Key: Efficiency Criteria Equity or Ethical Criteria Political Criterion is on the equitable or moral aspects of providing preferential treatment to poor people, not on whether that would be efficient because it made them more productive, lifted them out of poverty, and so on. The fourth criterion of equity is the `rule of rescue', the admonition to give priority to saving lives over interventions that do not make a life-or-death difference. Finally, there is something rather lamely called `public demands', which is meant to encompass an important political aspect missing from these more technical criteria: that is, what does the public think its money should be used for? Public beliefs and wishes may be many and contradictory, but probably include a mixture of equity and efficiency concerns, and it is worth asking how they also may be related to cost-effectiveness. (Other demands and pres- sures for public expenditure arise from providers of health care and from sup- pliers of equipment, drugs and other inputs; these are not considered here.) The full set of possible binary relations among the nine criteria would require an unwieldy 36 comparisons. Only eight of these are considered, 170 · Health Economics in Development comparing every other criterion with that of cost-effectiveness. The reason is that this is arguably the most misunderstood criterion, partly because it is usually not treated in relation to other reasons for intervening in the health sector. In consequence, it is thought by some people to be the only test that matters, and by others to be irrelevant or nearly so. Concentrating on the relations between other criteria and cost-effectiveness helps to clarify the role of cost-effectiveness in making decisions about public intervention in health care. Cost-Effectiveness and Public Goods This is the simplest relation among those considered: cost-effectiveness is the criterion for choosing which of a number of public goods are worth financing. Simply being a public good is not reason enough for the govern- ment to finance a health care intervention, because the result in improved health might not be worth the cost--the same resources could be better used for another health service or for some non-health activity [2]. But if something is a public good, there is no private market for it, and so there is no risk that government finance will crowd out private purchases. In conse- quence, the output that is paid for publicly is all the output there is, and there is a direct relation between public finance and the gain in welfare. How large that gain is depends on the health improvement and on the welfare evaluation of the health gain. Maximization of welfare under these conditions leads to cost-effectiveness as the way to choose among different public goods [3]. All that is necessary is that total welfare be monotonically related to total health gain, so that obtaining the largest possible improve- ment in health--which is what cost-effectiveness does, for a given amount of resources--also yields the largest obtainable total welfare. The relation between public goods and cost-effectiveness is not only straightforward, it is also hierarchical. That is, one asks first whether some health care service qualifies as a public good, and then whether it can be provided cost-effectively. It might seem just as well to select all the cost- effective health services first, and then to ask which ones of them are public goods, but there are two reasons not to proceed in this way. First, too little is known about the cost-effectiveness of hundreds of health ser- vices to draw up such a set and choose from among them [4]. It reduces the demands on information to ask about cost-effectiveness only after apply- ing some other, more easily determined, criterion. Second, the sequence Criteria for Public Spending on Health Care · 171 public good? cost-effective? is the only pair of questions that need to be asked where public goods are concerned. In contrast, the question of whether a particular health service is cost-effective also arises in relation to several other criteria, as indicated below, and it is easier to deal with each of them in turn. Externalities, or Partly Public Goods If an intervention does not qualify as a pure public good, because private purchasers are willing to pay for it, there may still be significant effects on non-purchasers: that is, externalities. Measures to treat communicable dis- ease can fall into this category, because while a person with the disease may be willing to pay to be cured, treating him or her also reduces the risk of transmission and thereby protects others. (Chemotherapy for tuberculosis is an often-cited example.) The service is then partly private, because there is a private or individual gain that can be bought in the market, but is also partly public because of the externality. For this reason it seems intuitively clear that cost-effectiveness should also be used to choose which services with externalities deserve to be financed publicly. But there are two other issues to deal with before asking about cost-effectiveness. The first is whether the externality is significant: that is, whether it really makes for a difference between the private benefits to purchasers and the total benefits, including those accruing to non-purchasers. If the difference is quite small, the service has only a little public character, and it can prop- erly be treated as if it were a purely private good. If however there is a large externality, it becomes appropriate to ask whether the private demand for the service is sufficient to assure realization of (nearly) all the potential social benefit. That is, will private individuals buy enough of the interven- tion that no public inducement or subsidy is needed? Or will they, for what- ever reason (poverty, high cost, under-valuation of the benefits), fail to seek or to continue treatment, so that the preventive potential is not realized? If the answer is that the private demand appears inadequate, then cost- effectiveness becomes relevant. That is, one should ask whether the poten- tial social gain from publicly subsidizing the service with the externality is large enough to justify the cost. Since there is some private demand, the subsidy may be only partial, rather than total as in the case of pure public goods. The chain of questions becomes longer: significant externalities? adequate private demand? cost-effective? 172 · Health Economics in Development For both public goods and services with externalities, the hierarchical relation to cost-effectiveness means that the question of whether the differ- ent criteria are consistent does not arise. One does not get into hypotheti- cal problems such as, that a particular intervention ought to be financed publicly because it is a public good, but should not have public money spent on it because of low cost-effectiveness. The second decision dominates the first, and the service is not worth buying. What If the Cost is Catastrophic? Two health care interventions may look equally justified because they have the same ratio of cost to effectiveness--the same cost per life saved, per year of healthy life gained, or some other measure of results--but one may be an order of magnitude more expensive than the other ([5] Figure 3.2). People paying out of pocket will afford the cheaper service, but the catastrophically costly service will be available only if the financial risk is shared. The fact that some services cost too much for individuals to buy is often regarded as a reason for public finance, but it is really a reason for insurance. Whether that insurance should be public or private or a mixture of the two is a com- plex question, involving the risks of failure in the private insurance market and the costs of limiting those risks or eliminating them by public regula- tion, finance or other measures. If the cost of a service is not catastrophic, however--that is, if most people can pay for it out of pocket without being impoverished--then there is generally no argument for it to be financed publicly. (The exceptions are services which people could buy, but do not, because of incomplete information, such as preventive screening that would allow early treatment and avoid later catastrophic expenses. Subsidizing such services is one way of overcoming this market failure.) This leaves two conclusions. One is that costs matter by themselves, and not only in relation to results. The other is that if the cost is low enough, individuals who contemplate buying the service can make their own deci- sions about cost-effectiveness. The two criteria, cost and cost-effectiveness, are independent, so they are compatible. Only when the cost is catastrophic, and there are good reasons for public finance of the service, does cost- effectiveness need to be separately taken into account. Then the two crite- ria may be in conflict, as will appear later. There is one more question to ask about the criterion of catastrophic costs, and that is, catastrophic for whom? Some individuals, after all, can Criteria for Public Spending on Health Care · 173 afford much more medical care than others. The easiest answer to this ques- tion seems to be to regard a cost as catastrophic if it cannot be paid by someone who is non-poor, without it making him or her poor. That is, the distinction between catastrophic and non-catastrophic depends on the dis- tinction between being poor and non-poor, where poverty is defined by some criterion other than health, such as income, consumption of necessi- ties such as food, etc. This leads directly to the next criterion: poverty as an ethical reason for public finance of health care. Poverty and Cost-Effectiveness Since poverty is defined independently of either the costs or the outcomes of health care, there would seem to be no obvious or necessary relation between these two criteria. The situation is not quite so simple, however, to the degree that the poor are not only sicker and die younger than the non- poor, but are afflicted by different diseases. Both the level and the compo- sition of the burden of disease differ somewhat between poor and non-poor populations, with the poor suffering more from communicable diseases and from premature mortality, compared to non-communicable causes and dis- abilities ([1] Appendix B, Tables B.5 and B.8), [6]. It happens that some though not all of the diseases which differentially affect the poor are also diseases for which relatively cost-effective interven- tions exist. There is therefore a fairly marked relation between poverty and cost-effectiveness, but it is also an accidental and transitory association. It exists because the non-poor either do not need those interventions, or would benefit less from them, or have already benefited, while the poor still suffer a large reducible burden. Measles is a good example of this difference: the non- poor of the world are protected by immunization, but many of the poor still are not. And since the poor are more likely to be malnourished, and since malnutrition greatly increases the risk of death from measles [7], immunizing the poor prevents even more health loss than among the non-poor. This suggests that poverty and cost-effectiveness are often compatible criteria: doing something to improve the health of the poor has a better than average chance of also being cost-effective--but not always. The poor also suffer from many health problems which do not now have cost- effective solutions, and it does not automatically follow that public money should be spent ineffectively, just because the intended beneficiaries are poor. The same question of value for money arises here as with the 174 · Health Economics in Development efficiency reasons for public expenditures. This is the basis for the idea of using public funds to guarantee access to a package of `essential' or `basic' services for the poor, and choosing those services partly on the basis of their cost-effectiveness ([51, Tables 3.2 and 3.3), [8]. Some of these services are public or partly public goods, so public spending on them is justified for the whole population at risk. The idea of a limited package of subsidized serv- ices for those in poverty is often attacked as being unethical, as represent- ing `poor health care for poor people', but that criticism is misguided on two grounds. First, the poor would, under such an arrangement, receive on average more cost-effective care than the non-poor. Second, there is noth- ing unethical about taking into account the wishes of those who pay the taxes to support the subsidy, for whom cost-effectiveness may be a reason- able criterion [9]. This issue is taken up again when considering what `pub- lic demand' means. To sum up: poverty and cost-effectiveness are somewhat related, but the association is partial and, historically, probably transitory. It did not exist a century or more ago, when the rich and the poor got sick and died of the same diseases and there was little effective health care, and it may not hold in the future, when the known cost-effective interventions will, one hopes, have been more fully applied to the benefit of the poor. Because the associa- tion is only partial, it is appropriate to ask first whether people are poor, and then whether a particular intervention justifies its costs. Horizontal Equity in Health Care This ethical principle implies giving equal treatment to people with equal health problems, that is, not discriminating among them as to how much or what kind of care to provide. If two people are really equal in the nature and severity of a disease or injury, and they get the identical treatment, then the results should also be the same. So horizontal equity implies equal effective- ness. In practice, outcomes often differ among patients who appear alike both in their problem and in their treatment, but the conclusion will still hold on average. The health gains will also differ if two people differ in age and therefore in life expectancy, with the younger beneficiary having more to gain from treatment. Similarly, there will be differences if one patient has an additional health problem (co-morbidity) which limits the effectiveness of care for their common problem. Other possible differences among people--in income, education, location, and so on--are irrelevant for this purpose unless they affect the outcome of all interventions. Criteria for Public Spending on Health Care · 175 All this implies that horizontal equity and cost-effectiveness are perfectly compatible, so long as the costs are equal, or nearly so, because then the cost-effectiveness of treatment will be the same for everyone with the same problem. But if the costs of treatment differ significantly between one patient and another, the two criteria are in conflict, and, as often happens [10], a difficult choice has to be made between equity and efficiency. Fig. 9.2 illustrates this conflict: one population suffers from a problem treatable by Intervention 1, and another population from a problem for Figure 9.2 Cost-effectiveness versus Horizontal Equity: Comparison of Two Interventions Where All Beneficiaries are Identical Except for Costs Effectiveness of an Intervention Marginal Cost per Unit of Health Gain { } Intervention 2 Intervention 1 C** C* Number of People Affected 176 · Health Economics in Development which Intervention 2 is appropriate. The effectiveness of an intervention is the same for everyone who would benefit from it, but the cost of treatment is not the same. Some people live in remote areas, or are otherwise more difficult to reach or to treat, so that the marginal cost of an intervention, per unit of health gain, rises as more people are treated. Intervention 2 is uni- formly more effective--yields more health gain--than Intervention 1, but even for the least costly patients, the cost is so much higher that the cost-effectiveness is initially lower than for Intervention 1. Horizontal equity says nothing about which intervention to finance first, and cost- effectiveness clearly gives priority to starting with Intervention 1. But before everyone in the first population is treated, the marginal cost rises to the level C*, so high that Intervention 2 begins to give more value for money. Cost-effectiveness would require switching some expenditure to that intervention, slowing the rate at which Intervention 1 is extended. If C** is the upper limit to cost per unit of health gain--any less cost-effective treatment is judged not worth buying--the result is to treat everyone with Intervention 2 but to leave part of the first population without treatment, in violation of horizontal equity. This kind of conflict can arise no matter which intervention is more effective; the problem is purely that costs differ among patients who are otherwise alike. Difficulties of this sort doubtless lie behind numerous conflicts between equity and efficiency in health care, such as those associated with urban/rural differences: it is usually easier and cheaper to treat urban residents than scattered rural populations, particu- larly if the intervention requires hospitalization. Vertical Equity in Health Care Horizontal equity presents a straightforward possibility of conflict with the criterion of cost-effectiveness, in which only one variable is involved--the cost of the intervention. The case of vertical equity and its relation to effi- ciency turns out to be much more complicated, because three variables are relevant. These are the cost and the effectiveness of different treatments, and the severity of different health problems. Since vertical equity concerns preferential treatment for people with worse problems, severity cannot be ignored the way it can be when dealing with horizontal equity. What particularly complicates vertical comparisons is that the effectiveness of a service need not bear any relation to the severity of the condition or disease it is meant to prevent or treat. Effectiveness, however it is measured, Criteria for Public Spending on Health Care · 177 corresponds to the improvement in health from an intervention, which is the same thing as the reduction in the health damage caused by the disease. (The sum of this damage over a population is called the burden of disease: the same name is not commonly used, but the same concept applies, to the health damage for an individual.) One intervention may lead to a larger health gain than another, but cause a smaller proportional reduction in dis- ease burden. For example, surgery following major trauma can be highly effective because it saves the patient's life, but still leaves him or her with severe disabilities, while surgery for some minor condition restores the patient to perfect health but causes less absolute improvement. Fig. 9.3 illustrates some of these complexities. Four diseases or condi- tions are shown, affecting different populations. For each one, the figure also shows the size of the health loss (for an individual, this is the height above the horizontal axis); the effectiveness of treatment (the portion of health loss prevented or relieved by an intervention); and the cost of treat- ment (the length of the vertical bar below the horizontal axis). For each intervention, the cost is assumed to be uniform for all potential patients or beneficiaries, since the issue of costs differing among individuals has already been considered in discussing horizontal equity. Interventions 1 and 2 deal with the entire burden of the corresponding diseases, but Inter- vention 3 is effective for only some of the people with that health problem and will do no good at all for others, while Intervention 4 helps everyone, but only partly--the effectiveness falls far short of eliminating all the health damage. Vertical equity requires giving preference to the sufferers from the sec- ond condition, since it causes the most health loss. Moreover, there is a fully effective intervention, which reinforces the choice. But the cost is much higher than for any other intervention, so that the cost-effectiveness of Intervention 2 is the worst of the four. Even Intervention 1, against the dis- ease which does the least harm to people's health, looks better on the cost- effectiveness criterion. That relation can be changed only by changing how one measures the disease burden and the results of an intervention, so that for example Disease 2 looks not twice as awful as Disease 1, but several times worse. This issue is taken up again later. Other kinds of conflict between equity and efficiency also arise in Fig. 9.3. Disease 4 causes an individual burden nearly as large as Disease 2, so equity according to severity would rank it second. And the treatment is no more costly than Intervention 1. The problem is that the service is, as indi- cated, not very effective, so that it is also not very cost-effective. Again, 178 · Health Economics in Development Figure 9.3 Cost-effectiveness versus Vertical Equity: Comparison of Four Interventions Differing in Disease Severity, Intervention Effectiveness and Cost Disease 2 Severity (health loss from the disease) Disease 4 Effectiveness (health gain from the intervention) Disease 3 Disease 1 Cost of the Intervention Disease 1 looks like the more efficient place to use resources. Disease 3 presents an even more complicated case, because the treatment is as cost- effective as Intervention 1 provided it is given only to those who will bene- fit from it, but yields much lower value for money if it is given to all suffer- ers, including those for whom it will do nothing. This discrimination of who should and who should not receive a treatment that will help some people but not others is one of the difficult choices medical professionals Criteria for Public Spending on Health Care · 179 regularly make, usually without complete knowledge of the likely outcome. Erring on the side of equity then means wasting some resources. It might seem that most or all of these problems could be resolved by redefining vertical equity, so that instead of "do more for those with worse problems", the principle were understood as "do more for those who can be helped more". That eliminates the severity of the condition as a relevant variable, leaving only the cost and effectiveness of treatment. But that does not solve the equity/efficiency conflict, because more effective treatments often cost proportionally more than less effective ones, which makes them less cost-effective. There is simply no systematic relation between cost and results, just as there is none between results and severity. Is the Rule of Rescue Efficient? This last ethical criterion is much simpler to deal with than those just dis- cussed, because it does not involve comparisons among individuals, except of the simplest form--between those who will die without an intervention and those for whom the appropriate health care will not make such an all-or-nothing difference. This choice is the basis of triage, the custom of dividing patients into those whose lives can be saved by intervening, those who will die even if given treatment, and those in between because their lives are not immediately threatened. And that seems so obviously the rational thing to do that it appears there should be no conflict between effi- ciency and ethical considerations. This conclusion is correct, with one important proviso: the rule of rescue and the criterion of cost-effectiveness are compatible, so long as `saving a life' means keeping a person alive for long enough, and in good enough health, that the effect justifies the cost. This is what is usually meant by saving some- one's life--rescuing him or her from drowning or some other accident, or curing a potentially fatal disease before it has done irreparable damage. The payoff can be decades of healthy life, so it does not matter that there may be a high cost. That is not the same thing as postponing death briefly, or keep- ing a person alive but terribly disabled. The effectiveness of any intervention depends on the instantaneous improvement in health, and also on how long that improvement lasts. Just as brief illness with full recovery contributes lit- tle to the burden of disease, no matter how severely one is sick, brief health gains, even if dramatic, do not yield much effectiveness. That is why, among other things, heroic measures to stave off death from chronic conditions are 180 · Health Economics in Development usually not cost-effective, and why the effectiveness of treatments for cancer is measured not by whether the patient leaves the hospital alive, but by the 5- or 10-year survival rate. This idea is explicitly recognized in any measure of health status that has a time dimension, such as potential years of life lost (PYLL) or more complex measures such as disability adjusted life years (DALYS) [11], or quality adjusted life years (QALYS) [12]. And What Do the People Want? The rule of rescue is the last of the seven relatively technical, explicit crite- ria to be contrasted to cost-effectiveness. It remains to consider briefly how all this relates to what the public may think or want, and in particular to whether cost-effectiveness is likely to be compatible with those views and demands. This question cannot be answered conclusively, because what the public thinks varies from place to place and time to time, and is often either amorphous or polarized. There may be no such thing as a clear demand from people about which criteria should be used to determine how their money is spent. (In the case of public goods, there is no demand in the mar- ket sense, and there may or may not be one in the sense of public agreement as to what the government should do.) So it is possible only to speculate about how public wants fit in with the criteria under consideration: two hypotheses seem particularly relevant to this discussion. First, the public is likely to mix up criteria or to misunderstand them. This is not surprising or reprehensible, in view of how complicated the rela- tions can be between one criterion and another. To take one example, the decision of which health services to subsidize for the poor depends both on an assessment of who is poor and on a notion of how cost-effective some- thing must be in order to justify public expenditure--and these are related to the distinction between costs that are catastrophic even for people who are not poor, and those that the non-poor can afford but the poor perhaps cannot. It is a combined technical and political task to decide on these matt- ers clearly enough that a public subsidy for medical care for the poor can be put into operation at all, and its design is likely to involve many choices in both the medical and the financial spheres, as in the Medicaid program in the United States [13]. One of the reasons for separating a number of possible criteria and looking at their connections, as has been done here, is precisely to facilitate public discussion and understanding of the issues involved. Criteria for Public Spending on Health Care · 181 Second, most of the public would probably agree with the general proposition that one should get value for money when spending their taxes; but whether they would support a particular criterion of cost-effectiveness, and the allocations that result from it, might depend very much on how effectiveness is defined. And the relative weights to give to effectiveness and to costs might not be so simple as taking the ratio of the two measures and using it to rank different services. There is little explicit experience with the issue, but the development of the `Oregon Plan', the list of health care procedures that the state of Oregon decided to include in its Medicaid plan, is instructive in this regard [14, 15]. The process began with an expert ranking according to perceived cost- effectiveness, using a measure that--as with PYLL, DALYs and QALYs-- is linear across individuals. That is, 10 years of healthy life lost are valued the same whether they are all lost by 1 individual, or ten people each lose 1 year, or 1,000 people each lose 3 or 4 days. When this ranking was proposed to the voters of Oregon, they ended by modifying it in various ways, one of which was equivalent to regarding a concentrated loss for one person as worse than the same loss distributed among many [16]. This non- linearity of disease burden, and therefore of effectiveness of health care, makes calculation harder because one cannot simply use totals or averages when comparing different health problems. But it is perfectly defensible as an ethical view. And it affects some of the other criteria considered here. In particular, changing how one measures effectiveness must, at the least, affect the comparisons involved in trying to apply horizontal and vertical equity, although that will not change the possible conflicts between equity and efficiency discussed above. One conclusion from this experience seems to be that one cannot simply settle all the other connections among criteria and take up what the public wants at the end, because what the public thinks, or wants, or is prepared to support, can modify some of the other criteria and the way they are related. Putting Everything Together: A Guide to Decision-Making As promised in the Introduction, the object of this note is not only to exam- ine the compatibility of different criteria, but to relate them sequentially or hierarchically when that is appropriate. If the different criteria can be taken up in some logical sequence, some of the problems of conflict among them are reduced, and it becomes easier to decide how to choose whether to 182 · Health Economics in Development finance a health care service out of public funds, leave it to the private mar- ket, do some of each, or conclude that the service will not and should not be paid for either publicly or privately. Fig. 9.4 summarizes the results of the comparisons and connections stud- ied here, in the form of a decision tree. Four conclusions deserve emphasis. First, it is possible to put together a clear sequence of questions, the answers to which ultimately determine in which of the three possible outcomes a given health care intervention falls. No single criterion ever suffices to justify public expenditure; however, depending on the character of an intervention, Figure 9.4 Decision Tree for Public Resource Allocation in Health Care Yes Public Good? No Significant Yes No Externalities? Adequate Catastrophic Yes No Demand? Cost? Insurance No Yes Appropriate? Beneficiaries No Yes Poor? Cost Yes No Effective? No Yes Public Private Leave to Do Not Finance Regulated Provide Publicly Private Market Criteria for Public Spending on Health Care · 183 one or more criteria may be irrelevant, which greatly simplifies decision- making. Second, it is easier to arrange the efficiency criteria in this way, than those related to equity. Poverty is easy enough to locate in the decision tree, but horizontal and vertical equity are not, because they involve explicit com- parisons among people and (in the case of vertical equity) among services. The question of how far to respect these two principles, in particular when they are in conflict with efficiency, probably has to be dealt with repeatedly rather than fitting neatly into one branch of the decision tree. The last eth- ical criterion, the rule of rescue, does not cause similar difficulties, since if `rescue' is properly interpreted it is usually consistent with cost-effectiveness. Third, there are some questions to ask about the use of public resources for health care that do not correspond precisely to any one of the criteria. This is most notably the case for interventions that are catastrophically costly, for which insurance is the appropriate mechanism for sharing risk and thus making the services available. The decision tree includes (but does not fully answer) two questions about insurance. One is whether a contributory insurance scheme is feasible--one in which there is explicit coverage of serv- ices in return for a premium that may be uniform, or related to capacity to pay. (The distinction whether one contributes or not is important, because public subsidy of health care for the poor is also a form of insurance, but one to which they do not contribute. If the poor are covered for catastrophically costly interventions, their premiums must in effect be subsidized.) When insurance is feasible and appropriate, there is a further question whether it should be public or private or some of each. This is a complicated question not settled by appeal to any combination of these criteria--but it is evident that both horizontal and vertical equity are relevant to the decision. Private insurance is more likely than publicly financed care to violate one or both of these principles, by discriminating between people who have the same health problem but differ in their other health conditions, their age, capacity to pay, or other characteristics, or by making it more difficult to get coverage for more severe health problems, if the treatments for them are particularly expensive. These failings go far to explain why in most rich countries a large share of insurance is publicly financed, whether directly from the budget or through employment-related social security schemes [2]. Fourth, it turns out that the right way to use the criterion of cost- effectiveness is not once-and-for-all, as with some of the other criteria. Several different paths through the decision tree lead to the question whether a health care intervention's results are worth what it costs. One goes directly from the classification of a service as a public good, and is the only 184 · Health Economics in Development other criterion needed in that case. Another begins with the conclusion that a service has significant externalities, and that private demand for it is inad- equate to secure all the potential social gains--the case of a partly public good. The question of cost-effectiveness arises again when considering what services to subsidize for the poor, although here the criterion is less clear-cut, as the experience of the Oregon Plan illustrates. It does not become irrele- vant, though, because the people paying for the subsidy are almost certain to want some kind of value for money. And for similar reasons, if a service is catastrophically expensive, and it is not feasible to finance it through ordi- nary contributory insurance so that some separate and public funding must be found if it is to be provided, it is again relevant to ask whether the service is worth what it costs. This is the typical situation of a public program of re- insurance, superimposed on a large number of small private or semi-private insurance schemes which cannot afford the risks of a few extraordinarily costly interventions. Cost-effectiveness, in other words, is decisive in only one circumstance but important in several other situations. To sum up: a fully thought-out decision of which health services to spend public money on and for whom, requires looking at all nine of the criteria considered, treating them in the proper sequence and taking account of whether they are consistent or in conflict. Public funds should be spent on public and semi-public goods when those are cost-effective and demand for them is inadequate, on cost-effective interventions which disproportion- ately benefit the poor, and on catastrophically costly care, when contribu- tory insurance will not work effectively or there are good reasons to finance insurance publicly. Interventions which do not pass these tests either are not worth paying for at all, or they can be left to regulated private markets to finance because the costs are bearable without insurance, or private con- tributory insurance is feasible. References [1] Musgrave RA. The Theory of Public Finance, A Study in Public Economy. New York: McGraw-Hill, 1959. [2] Musgrove P. Public and Private Roles in Health: Theory and Financing Patterns. World Bank Discussion Paper No. 339. Washington, DC: The World Bank, 1996. [3] Hammer JS. Prices and protocols in public health care. World Bank Econ Rev 1997; 11. Criteria for Public Spending on Health Care · 185 [4] Jamison DT, Mosley WH, Measham AR, Bobadilla JL, editors. Disease Control Priorities in Developing Countries. Oxford: Oxford University Press, 1993. [5] World Bank. World Development Report: Investing in Health: Figure 3.2. Washington, DC: The World Bank, 1993. [6] Gwatkin DR, Guillot M. The burden of disease among the global poor: current situation, future trends and implications for research and policy. Washington, DC: The World Bank, 1998. Prepared for the Global Forum on Health Research. [7] Beaton GH, Martorell R, L'Abbé KA, Edmonston B, McCabe G, Ross AC, Harvey B. Effectiveness of Vitamin A Supplementation in the Control of Young Child Morbidity and Mortality in Developing Countries: Summary Report. Toronto: University of Toronto, 1993. [8] Bobadilla JL, Cowley P, Musgrove P, Saxenian H. Design, content and financ- ing of an essential national package of health services. In: Murray CJL, Lopez A, editors. Global Comparative Assessments in the Health Sector: Disease Burden, Expenditures and Intervention Packages: 51, Tables 3.2 and 3.3. Geneva: World Health Organization, 1994. [9] Musgrove P. Cost-effectiveness and the socialization of health care. Health Policy 1996:32:14. [10] Okun AM. Equity and Efficiency: the Big Tradeoff. Washington, DC: The Brookings Institution, 1975. [11] Murray CJL. Rethinking DALYs. In [16], 1. [12] Goerdt A, Koplan JP, Robine JM, Thuriaux MC, Van Ginneken KK. Non- fatal health outcomes. In [16] 99. [13] U.S. Social Security Administration. Annual Statistical Supplement, 1996, to the Social Security Bulletin. Washington, DC: Social Security Administration, 1996. [14] Eddy DM. Oregon's methods: did cost-effectiveness fail? J Am Med Assoc 1996:266. [15] Hadorn DC. Setting health care priorities in Oregon: cost-effectiveness meets the Rule of Rescue. J Am Med Assoc 1996:265. [16] Murray CJL, Lopez AD, editors. The Global Burden of Disease, Vol. 1. Boston, MA: Harvard School of Public Health, 1996, on behalf of the World Health Organization and the World Bank. CHAPTER 10 Cost-Effectiveness and the Socialization of Health Care Acknowledgments In revising the paper I have benefited from extensive comments by Peter Berman and Howard Barnum. Introduction: The Nature of the Problem The World Bank's 1993 World Development Report--Investing in Health [1] treats cost-effectiveness as the principal criterion for choosing which health interventions governments should pay for. The Report recognizes the public-finance arguments that government must finance some interven- tions if they are to be provided adequately or at all, because private markets will under-provide public goods and goods with substantial positive exter- nalities. However, even interventions which meet this requirement must still pass some test such as cost-effectiveness--a public good should not be financed simply because it is public. If it is accepted that the object of health expenditure is to improve health, rather than some more general concept such as welfare, and what is financed is a collection of interventions, then choosing to finance them in decreasing order of cost-effectiveness seems to be the way to maximize the health gain from any particular level of spend- ing. "Value for money" implies cost-effectiveness, so long as "value" is measured in health gains. Reprinted from Health Policy 32, 1995, pp. 113­123, Copyright 1995, with permis- sion from Elsevier Science. 187 188 · Health Economics in Development There is an important technical objection to this simple view, which is that either the cost or the effectiveness of any one intervention may depend on what other interventions are offered at the same time. In that case the ranking by cost-effectiveness is not stable and one cannot just proceed down a list until the budget is exhausted or the next best intervention does not seem worthwhile. Taking account of interactions is difficult, but sometimes interventions can be "packaged" to exploit joint costs or synergies. Since a small number of interventions can deal with a large share of the disease burden, it is not necessary to consider all possible combinations: a cost- effective minimum package can be defined to absorb all health spending in poor countries and serve as the nucleus of a larger set of interventions in higher-income countries [2]. Less is known about cost-effectiveness, and it becomes harder to package interventions, as health expenditure rises, because the variety of interventions increases greatly. A more serious difficulty is that while costs are, in principle, objective, the effectiveness of an intervention is always partly subjective. The quan- tification of Disability-Adjusted Life Years (DALYs), Quality-Adjusted Life Years (QALYs), Healthy Life Years or any other measurement of gains in health status depends on a long vector of elements. Some of these are objec- tive, such as rates of mortality or incidence and prevalence of morbidity or disability, and the ages at which diseases strike. Many other elements require subjective valuation. For example, the cal- culation used in the Bank's Report required choosing numbers for the discount rate; the relative value of life at different ages; life expectancy; and the relative disabilities associated with each of many non-fatal conditions [3]. If people differ in their evaluation of these numbers, each one may have his own ranking of cost-effectiveness, but the rankings will not agree enough for there to be a unique relation between cost and health improve- ment associated with any intervention, even for people of the same age and the same severity of health problems. To say that people have different evaluations of effectiveness is equiva- lent to saving they derive different utilities from the same intervention because they differ in their views of the utilities associated with the prior state of ill health or health risk, the state of health after intervention, the disutility caused by the intervention itself, or some combination of these. If the objective is to maximize welfare rather than health, then the situation is still more complicated because people can also differ in their subjective evaluation of non-health benefits from consumption of other goods and services, in their aversion to risk, in income or in other factors. Cost-Effectiveness and the Socialization of Health Care · 189 This problem raises two related questions. First, should individual behavior in seeking and paying for health care be expected to be consistent with cost-effectiveness? Second, does socializing health care, whether through private insurance or through public financing, make collective behavior different in this respect from individual decisions? This paper argues that the more decisions about health interventions are socialized, the more appropriate it is to use cost-effectiveness as a criterion for health care spending. It does not presume to show that no other criterion does, or should, matter in making choices about health care, but only that the rela- tive importance of cost-effectiveness increases with the degree of socializa- tion. This tendency is reinforced when the socialization is compulsory, because people must contribute through taxes rather than voluntary insur- ance payments. That makes cost-effectiveness a particularly legitimate criterion for government decisions. The argument does not depend on the particular measure of health status, of which there are many [4, 5], which is used to derive the effectiveness of interventions. Cost-Effectiveness and Individual Choice In trying to answer the first question posed above, it is easy to invent exam- ples in which people would not be expected to choose the more cost effec- tive of two alternatives. Suppose, for example, that two drugs are available to treat a given health problem which poses a 30 percent chance of dying. One drug costs $2 and reduces the risk of death to 25 percent; the other costs $5 and reduces the risk to 20 percent. Each application of the first drug can be thought of as saving one-twentieth of a life, or one life for each 20 patients who receive it, while the other saves one-tenth of a life, or one life for every ten patients. The first drug is more cost-effective, because the cost is 40 percent as high while the health gain is half as large as with the second drug: the cost per life saved is $40 instead of $50 [6]. However, any- one who had $5 to spend would buy the second drug, unless the additional utility from spending $3 on something else outweighed the extra five per- cent reduction in the risk of dying. Choosing the more cost-effective drug can be interpreted to mean that the individual values his or her life at less than $60, or perhaps that unless the extra money is spent on food, he or she will die anyway. In this example, the individual cannot spend $4 on two doses of the cheaper medication, and improve survival chances by the same amount as 190 · Health Economics in Development with one $5 dose of the costlier one. Spending $4 will, however, buy the same total survival gain for two patients, with $1 left over. The difference between the individual and the social decision turns on this possibility of treating more than one patient and on the indivisibility of individual bene- fits. The claim that the first drug is more cost-effective also depends on adding gains in survival probability linearly across individuals, or assuming that one-tenth of a life for one person is worth exactly the same as one- twentieth of a life for each of two people. In a much more complex model [7], an individual with an additive, mul- tiperiod utility function faces a health problem that reduces the probability of surviving into the next period and therefore the expected utility in that period, where utility depends on disposable income or income less medical expenditure. The utility-maximizing decision about how much to spend on health care this period is consistent with cost effectiveness, and does not depend on what will be spent in the next period if the consumer survives. The same conclusion holds when the model is complicated by discounting future periods, by allowing for nonfatal conditions and for expenditures which improve health status without affecting survival, and by admitting two or more possible interventions with different costs and different effects on survival and quality of life. Moreover, the optimal cost-effectiveness ratio is the same for both interventions, so that at the margin either of them buys the same amount of expected healthy life per dollar. This is possible only if--in contrast to the previous example--the patient can buy any amount of either intervention, up to the limit of income, with the prices of the inter- ventions proportional to the marginal health gains. The optimal relation between prices and health improvements suggests that the overall cost-effectiveness of health expenditure should depend on what interventions cost the consumer, from which it is commonly assumed that prices can be used to "rationalize" demand: raising the price of health care should increase value for money. This could happen because con- sumers were buying so much of all interventions that the marginal health gain was very small, and a uniform reduction in utilization would save money but have little effect on their health. Alternatively, consumers might preferentially reduce purchases of interventions yielding less health gain per unit expenditure. (If the prices in the two-drug example discussed previ- ously were $200 and $500 instead of $2 and $5, the patient might choose the more cost-effective drug despite the lower health gain, because the $300 difference could buy much more compensating utility than $3 could.) How- ever, if interventions do not produce steadily declining marginal health Cost-Effectiveness and the Socialization of Health Care · 191 improvement but require some minimum expenditure in order to produce any gain at all, there is a risk that consumers would cut back or stop using more cost-effective services, causing substantial health losses. What consumers will actually do depends not only on prices and on the true relation between utilization and health improvements but on their knowledge of that relation. In both the formal models just discussed, the consumer does not know for certain whether he or she will become sick or injured or die, but does know the survival probabilities, or the expected gains in health status, with and without each kind of health care. Real con- sumers are usually less well informed, because they do not try to find out, or because physicians mislead them, or simply because no one knows. With- out that information it is impossible to buy care optimally, and so actual purchases cannot be expected to be consistent with cost effectiveness. Incomplete information is a pervasive reason why individuals' spending on health care might not be efficient even according to their own evaluations and preferences, leading them to prefer a collective solution in which lack of information is less damaging [8]. Add to this problem the variation in people's utilities, attitudes toward risk and other factors discussed above, and there is little reason to expect a collection of consumers to rank health care interventions by a single standard of value for money, and spend accordingly. There are numerous empirical estimates of how utilization responds to price changes or differences [9], but very little evidence of how this affects cost-effectiveness. The RAND experiment of health insurance deductibles and co-payments in the United States found that for most consumers, higher prices reduced demand but did not much affect its composition, and had little effect on health status [10]. Thus there was no evidence of prefer- ential reduction in the use of less cost-effective services-- "cost sharing reduces appropriate and inappropriate care in about the same way"--but some evidence that average cost-effectiveness did increase. However, poor people, and especially poor people needing preventive care, tended to reduce utilization in ways that substantially threatened their health, so that their use of health services became less cost-effective. This result may be more relevant for poor countries than the average findings of slightly improved cost effectiveness. One reason why individual consumers may not select health care accord- ing to cost-effectiveness is that they seldom choose between two interven- tions for entirely different health problems, although anyone financing health care for a large number of people must make such choices, if only 192 · Health Economics in Development implicitly. People typically suffer, and decide what to do about, one health problem at a time, and they have a clearer idea of the cost of doing nothing and the gains from health care for that problem than they have about other problems that might occur in the future. Ignorance about tomorrow's prob- lems compared to today's makes it hard to make cost-effective multiperiod choices over different health care needs. Another source of difficulty is that not spending on care today may make only a small or uncertain contribution to one's ability to buy health care tomorrow. (The model described earlier assumes away such connections by supposing that future income and ability to pay for interventions is independent of today's health problems and expenditures [7].) The solution to the difficulty of paying for care needed in the future, of course, is insurance. Socializing Decisions through Insurance For any given health care intervention, insurance pools the financial risk across all the purchasers who are covered by insurance for that intervention. "Pure" insurance would eliminate out-of-pocket payments entirely by pro- viding complete financial coverage. Most private insurance and some public insurance, in contrast, provides such zero-cost coverage only after the insured has spent a certain amount in the form of co-payments; thus it lowers the price for interventions up to some level of expenditure and lowers it all the way to zero only for expenditures beyond that limit. There may also be a floor up to which a deductible applies, with no effect on the price to the con- sumer. Thus the effect of insurance on consumption of health care depends in part on whether the insured are paying all, part or none of the cost of serv- ices. This in turn affects the administrative cost of providing the insurance. Does socializing health care financing via insurance favor cost-effective- ness? The substantial literature on the effects of insurance on behavior and on welfare does not address this question but usually compares expenditure outcomes under insurance with the results of out-of-pocket purchases by individual consumers. It also concentrates on the subsidy to insurance which occurs when employers' contributions to buying insurance are excluded from corporate income and not taxed as income to workers; that issue is not considered here. At first glance the answer to the question about cost-effectiveness and insurance would seem to be a simple no: insured individuals either face lower prices, or no longer have to consider costs at all, so they can consume Cost-Effectiveness and the Socialization of Health Care · 193 health care without asking whether the health gain is worth the full cost. In fact, people may buy insurance partly to protect themselves from the need to consider cost, at a time of urgency and stress. Even if all the insured are spending in the range where prices (co-payments) are between zero and 100 percent of costs, and even if they continue to equate marginal costs to mar- ginal benefits according to their individual utilities, they should be expected to buy services of lower marginal health gain and thus reduce cost- effectiveness on average. The moral hazard intrinsic to insurance [12] is almost the antithesis of cost-effective behavior. To the extent that moral hazard leads people to take less good care of their health--particularly, to spend less on preventive measures not covered by insurance because they will not have to pay for curative treatment that is covered--insurance even appears to undermine cost-effectiveness in preventive and public health measures. More curative interventions may be required, at higher cost, to achieve the same health outcome [13]. The conclusion that insurance is inimical to cost-effectiveness because of moral hazard is modified when account is taken of adverse selection, the second kind of market failure characteristic of the health insurance market. Adverse selection arises because different people know they have different health status and anticipate different needs for health care, and this infor- mation is not available to insurers, who must charge the same premium to all. As a potential consumer of services, each insured person seeks to maxi- mize the expected health gain from care for those problems he or she has or expects to suffer. As a purchaser, each one seeks to minimize costs. The insurance industry in effect allows buyers and users to bargain with one another so as to balance costs against health gains, with high-risk customers wanting to buy more insurance than low-risk customers. If low-risk con- sumers predominate in the market, they can force high-risk consumers to buy less insurance than they would like, offsetting some of the tendency of insurance to stimulate spending without regard for cost-effectiveness; the alternative for high-risk consumers is to pay the full cost of their anticipated health care instead of being subsidized by healthier and less costly customers [13]. Simulations based on the RAND experiment mentioned earlier show that more generous insurance plans with higher limits on out-of-pocket spending are squeezed out of the market, compared to plans which impose more cost on the insured [14]. Because it is difficult empirically to distinguish the effects of adverse selection from those of moral hazard, and 194 · Health Economics in Development because so much insurance is determined by employers (or by the state) and therefore limits adverse selection, it is not clear how much consumption is affected, and there is no specific evidence about cost-effectiveness. These findings suggest that insurance is less inimical to cost-effectiveness than would follow from the reduction in out-of-pocket costs alone, but they do not indicate that purchases under insurance would be any more cost- effective than without it. The expectation of lower cost-effectiveness with insurance still follows if any consumer can buy any health service in any amount at reduced prices, with declining marginal health returns as more is spent, at all levels of consumption. In fact, which services the insured person consumes depends on which services are covered, and insurance plans typically exclude certain services altogether or impose limitations on how often they can be used or how much can be spent on them. If everyone had the same evaluation of the rel- ative cost-effectiveness of interventions, insurance plans would reflect that choice, and plans would be more cost-effective on average as they were cheaper and covered fewer services. It is only because potential customers differ in their utilities, risk aversion and incomes that insurance does not automatically favor cost-effectiveness. Even so, any insurance plan which is actually bought implicitly averages together the cost-effectiveness evalua- tions of all the people who freely buy that plan. This does not mean that every purchaser of insurance thinks about the cost-effectiveness of each intervention under the plan, because plans generally apply the same broad rules about deductibles, co-payments, ceilings and quantitative limits to a whole range of similar services. It does mean that the people who end up covered by a particular plan either have no choice, or are in rough agree- ment about the relation between the expected health effectiveness of the coverage and what it costs. This implies that a competitive third-party insurance industry may lead to relatively cost-effective coverage within each of a number of policies or groups of insured, as people with the same tastes and desire for cost- effective coverage buy the same insurance. Similarly, when workers have a choice of employers but no choice about the insurance provided by each employer, each employer's workforce can be expected to be homogeneous with respect to the kind and amount of insurance wanted [15], also leading to local cost-effectiveness. Competition among insurers or employers will not, however, necessarily tend toward cost-effectiveness of interventions across all groups. Some groups will buy coverage for interventions of relatively low cost-effectiveness, while others will do without coverage for Cost-Effectiveness and the Socialization of Health Care · 195 interventions that are more cost-effective. Limiting consumer choice, as by requiring everyone to have the same plan or at least the same minimum plan, can therefore improve overall cost-effectiveness. The inclusion or exclusion of specific services is extremely important, because many health interventions are of an all-or-nothing character, or require a substantial minimum expenditure to yield any benefits at all. If there are eventual declining returns to further spending, these begin only at relatively high levels of expenditure. When people are relieved from wor- rying (so much) about the cost of care, they are free to concentrate on its effectiveness. Insurance can be expected to have little effect on the utiliza- tion of inexpensive services which people would buy anyway, but it stimu- lates the purchase of lumpy, expensive interventions, particularly if the need for these is hard to predict [16]. Since the cost and the cost-effectiveness of interventions are not highly correlated--there are more and less effective services at all cost levels [17]-- stimulating utilization of expensive services does not necessarily reduce value for money. People who are still paying part of the cost should become more likely in particular to consume care which is both costly and cost- effective. This may offset much of the tendency to expand utilization of services which are inexpensive but not very effective. When expenditure reaches the out-of-pocket limit, cost-effectiveness might be expected to fall again, because all costly interventions become attractive if they provide any health improvement at all. This will be true unless the costs in pain, incon- venience or time lost--which, unlike financial costs, cannot be shifted to third parties--are too high. In summary, the expectation that insurance will undermine cost effectiveness by stimulating consumption of services seems to depend on two conditions. First, individuals react to lowered prices by expanding utilization at the margin, rather than by including expensive services which may provide large health gains but are too costly to buy out of pocket. Second, co-payments are so low that the tendency to equate mar- ginal costs and benefits leads to over-utilization independently of price. By allowing purchases of costly but cost-effective services, and still providing an incentive to weigh marginal benefits against some cost to the consumer, insurance can promote cost-effectiveness even though it may raise total costs. This will be true even if the selection of which interventions to cover is independent of cost-effectiveness. Of course, an insurance plan can be used to impose cost-effectiveness by covering services, or setting co-payments, in proportion to their health gains per dollar, but in a competitive environment such a plan will not be 196 · Health Economics in Development bought by all customers. In general, any uniform plan will not be Pareto optimal, in the sense of improving health or reducing costs for everyone compared to their individual choices [13]. Of all uniform plans, however, one based on cost-effectiveness appears to be the most efficient [7]. Cost- effectiveness is an appropriate criterion for designing insurance provisions, but its scope is limited by the heterogeneity of customers and the degree of competition among insurers. More Complete Socialization: Second-Party Insurance Under third-party insurance, physicians can provide, and possibly even cre- ate demand for, services of low cost-effectiveness because--like their patients--they are free to consider only the health gains and not the costs. This can happen even when medical professionals are scrupulous about not recommending or performing services with no expected health gain. How- ever, if the physicians are also the insurers, they have to consider both the numerator and the denominator of the cost-effectiveness ratio. Such second-party insurance is more fully socialized than third-party insurance because the patient and the doctor cannot pass any costs on to someone else. Insurance such as that provided through health maintenance organizations should therefore lead to a more cost-effective set of interven- tions than would occur with the same patients and the same doctors under third-party insurance. This may happen because coverage of less cost- effective services is explicitly restricted, or, when such services are covered because they are performed less often, at physicians' discretion, or the more cost-effective of two alternatives for treatment is used. It is hard to estimate how much this occurs, because of differences in age, health status, education, attitudes toward medical interventions or other differences between those insured under second- and third-party coverage. The comparison is particularly difficult because of self-selection into one or the other kind of insurance: neither group is a random sample of the popu- lation [11]. And because outcomes are often uncertain, even health mainte- nance organizations sometimes spend heavily on procedures that turn out to be cost-ineffective. Nonetheless, the available evidence shows that second-party insurance controls cost better without sacrificing health gains [18], which suggests that it is more cost-effective. Certainly the criterion of cost-effectiveness is more clearly embodied in the way that care is paid for, than with third-party insurance. Cost-Effectiveness and the Socialization of Health Care · 197 Government Insurance and Taxpayer Support If greater socialization of private health care expenditure favors cost- effectiveness, as argued above, then what happens under the most socialized way of paying for care, namely, public financing? It helps to split this ques- tion into three parts. The first concerns what governments can do, the sec- ond what they actually do, and the third the legitimacy of cost-effectiveness as a criterion for public spending. Governments can make health spending more cost-effective, just by lim- iting public finance to the right interventions. The resulting cost effective- ness of what is paid for publicly will be partly offset by private purchases of less cost-effective interventions, but the effect of public finance can still be to improve the average value for money of health care. Any private insur- ance could, in principle, do the same thing--offer only services that are highly cost-effective--but it might lose customers as a result. The advantage of government in this respect is that beneficiaries cannot choose a different insurer without paying twice, once in taxes and once in premiums. This kind of coercion, like that imposed by employer financed private insurance, may be unattractive to each consumer individually and yet acceptable when everyone knows that everyone else is subjected to the same control. When publicly-financed services are universally available and sufficiently attractive that people actually use them, governments also can eliminate or reduce adverse selection--although, as indicated above, it is not clear how much that interferes with aggregate cost-effectiveness in private insurance. Of course, when only part of the population uses publicly-funded services, adverse selection can actually be increased, with the highest-risk population depending on public finance. This in turn may reinforce any tendency to limit services to those which are more cost-effective. When governments fail to improve the cost-effectiveness of health care, it is because they do not know which interventions are cost-effective; or provide all services indiscriminately; or cover too few people; or waste resources in delivery; or provide such poor quality that people willingly spend their own money on care elsewhere. For all these reasons, govern- ments in poor countries are often rightly accused of spending their health money badly compared to what could be achieved with an appropriate pack- age of care. Nonetheless, they often do no worse than the private sector; and when they finance public health measures and interventions like immu- nizations which private markets cannot or do not provide adequately, 198 · Health Economics in Development governments are more cost-effective. This accounts for a large part of the extraordinary gains in life expectancy and in reduction of morbidity and dis- ability in recent decades. The share of public finance in health care spending generally rises as countries become richer, and a higher public share is somewhat associated with lower overall expenditure relative to national income, or with better health, or both. Even without examining the composition of spending by intervention, there is empirical evidence that greater public control of expenditure promotes value for money [8]. This presumably results from a concentration of public spending on some highly cost-effective interven- tions which private markets usually do not provide so thoroughly; it does not follow that public spending is intrinsically more efficient, only that it is better at covering some of the interventions that provide the most health gains per dollar. How far this interferes with Pareto optimality and con- sumers' freedom to follow their own notions of utility from health services depends on the range of services that are publicly financed and on whether there are differential subsidies according to the cost-effectiveness of the service. Finally, there is the question of whether it is appropriate for govern- ments to base financing decisions on cost-effectiveness, or at least more appropriate than for individuals or private markets. Public insurance or direct public expenditure on health care typically differs from private insur- ance in providing subsidies not only from the healthy to the sick but also from the rich to the poor. The "purchasers" are taxpayers, who know that they are paying more than the average value of the services they can expect to receive. This makes them different from subscribers to an insurance plan, who know that the lucky will end up paying for the care of the unlucky, but who also know that a priori everyone is paying (approximately) a fair share. Subsidizing health care is vastly more complicated than subsidizing food, but it still helps to compare the two cases. In both instances, taxpayers are willing to pay for what they think others need or deserve, but not for just anything they want. Thus food stamps cannot legally be used to buy alco- hol or cigarettes; they can be spent on caviar, but then the buyer suffers for his choice in lower food intake. Cost-effectiveness offers a way to make a roughly parallel distinction in health care, with health gain corresponding to nutritional content in the case of food subsidies. Governments cannot simply subsidize all demand for health care, because that leads to both inequity and runaway costs. They need criteria for what to finance, and cost-effectiveness is a relatively transparent basis for decision. The fact that any objective criterion is based partly on subjective Cost-Effectiveness and the Socialization of Health Care · 199 evaluations of health burdens and gains does not invalidate that argument: what it means is that for cost-effectiveness to be understood and accepted as a basis for public finance, there has to be some consensus on those subjec- tive choices. With private insurance, in contrast, it is only necessary to get agreement among all the people who buy a particular policy. In fact, it may be that only government can promote the kind of public debate and understanding which can lead to consensus on those evaluations over the whole population, as the development of the "Oregon Plan" in the United States demonstrates [19]. However, this experience also illustrates that the public may not accept cost-effectiveness as the only or even the chief criterion for what services to finance. When the original list developed in Oregon turned out to give higher priority to some services with very small health gains but also very low costs than to some expensive but life- saving interventions, the cost criterion was dropped and the list restructured on the basis of effectiveness or health gains alone [20]. Even this contributes much more to health than a random choice with no regard for either out- comes or costs. Since cost-effectiveness is partly subjective, it is consistent with a greater priority for lifesaving interventions, if the disability weights for minor health losses are all revised downward so that death becomes more important relative to many diseases and injuries. The same ranking results if health losses are not added linearly across individuals, as cost- effectiveness implies, but priority is given to larger individual losses, ten healthy life years lost by one person being treated as worse than the loss of a year each by ten people. Public finance of health care involves coercion, which means that it can enforce pooling solutions and eliminate the effects of adverse selection globally, whereas competitive private insurance can do so only over smaller and more homogeneous groups. This makes it easier to apply cost-effec- tiveness as the criterion of what to pay for. This tendency is reinforced by the explicit subsidy involved in having taxpayers finance care for poorer non-contributors: as with other subsidies in kind, individual preferences can be over-ridden in favor of what those paying think the beneficiaries need. Effectiveness certainly corresponds to that notion, and cost-effectiveness may also be both politically and ethically a reasonable criterion for public expenditure. Much of the evidence used here comes from rich countries, and particu- larly from the United States. That country is an outlier even among OECD countries [8], in its reliance on employer-financed private insurance and its limitation of public expenditure to the poor, the elderly and a few other 200 · Health Economics in Development groups or types of spending. It is therefore legitimate to ask whether the argument developed here, that socialization of health care spending makes cost-effectiveness theoretically more appropriate as a criterion and easier to approximate in practice, is relevant to the issues of health care reform in poor countries. There seem to be two strong reasons why it is. First, while there is undoubtedly scope for more socialization of health care spending in poor countries through private insurance, the danger from both moral hazard and adverse selection is arguably much greater than in richer countries. Such private insurance as exists or could easily be introduced is confined to the well-off minority, who are healthier on average than the poor majority, and who are particularly likely to obtain public subsidies enabling their insurance to provide costly, but not very cost-effective, services. These risks also exist for direct public financing, but can be attenuated if financing cov- ers most of the population and therefore must be more limited as to the interventions it can cover. The argument that public financing offers poten- tial efficiency gains over private insurance in such settings of poverty and inequality applies a fortiori to out-of-pocket expenditure, where the hetero- geneity of people's conditions and utilities give no reason to think that spending will be particularly cost-effective. Second, the poorer a country is, the sicker its population is likely to be and the smaller the range of interventions that can be financed for the bulk of the population. These circumstances make it easier to be sure what inter- ventions are more cost-effective, and to compose an essential package of care which may be all that can be offered to the poor, but from which every- one can benefit [21]. Because many of these interventions are relatively cheap as well as quite cost-effective, people are not likely to buy private insurance for them--and such insurance is impossible for the public health measures with large externalities. People's information is also likely to be very incomplete, even regarding basic, life-protecting habits and procedures [1]. The welfare loss from not respecting individual preferences is therefore likely to be smaller, and the gains from socializing health care decisions larger, than in richer countries. References [1] World Bank, World Development Report 1993: Investing in Health. Oxford University Press, New York, NY, 1993. Cost-Effectiveness and the Socialization of Health Care · 201 [2] Bobadilla, J.L., Cowley, P., Musgrove, P., and Saxenian, H., "Methods and Data Used to Design the Minimum Package of Health Services." Background Paper to the World Development Report 1993: Investing in Health. World Bank, Washington, DC, 1994. [3] Murray, C.J.L., "Quantifying the Burden of Disease: The Technical Basis for Disability Adjusted Life Years." Bulletin of the World Health Organization, 72 (1994) 429­445. [4] Brooks, R.G, "The Development and Construction of Health Status. Measures: An Overview of the Literature." The Swedish Institute for Health Economics, Report 1986.-4, Lund, Sweden, 1986. [5] Torrance, G.W., "Measurement of Health State Utilities for Economic Appraisal: A Review." Journal of Health Economics, 5 (1986) 1­30. [6] Hammer, J.S., "The Economics of Malaria Control." World Bank Research Observer, 8 (1993) 1­22. [7] Garber, A.M. and Phelps, C.E., "Economic Foundations of Cost-Effectiveness Analysis." NBER Working Paper No. 4164, National Bureau of Economic Research, Cambridge, MA, 1992. [8] Barr, N., "Economic Theory and the Welfare State: A Survey and Interpretation." Journal of Economic Literature, 30 (1992) 741­803. [9] Griffin, C.C., "User Charges for Health Care in Principle and Practice." Economic Development Institute Seminar Paper No. 37, World Bank, Washington, DC, 1988. [10] Newhouse, J.P., and the Insurance Experiment Group, Free for All? Lessons from the RAND Health Insurance Experiment Harvard University Press, Cambridge, MA, 1993. [11] Pauly, M. V., "Taxation, Health Insurance and Market Failure in the Medical Economy" Journal of Economic Literature, 24 (1986) 629­675. [12] Arrow, K., "Uncertainty and the Welfare Economics of Medical Care." American Economic Review, 53 (1963) 941­973. [13] Pauly, M. V., "Overinsurance and Public Provision of Insurance: The Roles of Moral Hazard and Adverse Selection." Quarterly Journal of Economics, 88 (1974) 44­62. [14] Marquis, M.S., "Adverse Selection with a Multiple Choice Among Health Insurance Plans: A Simulation Analysis." Journal of Health Economics, 11(1992) 129­151. [15] Goldstein, G.S. and Pauly, M.V., "Group Health Insurance as a Local Public Good." In R. Rosett (Ed.), The Role of Health Insurance in the Health Services Sector, National Bureau of Economic Research, New York, NY, 1976, pp. 73­110. [16] Phelps, C.E., "Tax Policy, Health Insurance and Health Care." In J. Meyer (Ed.), Market Reforms in Health Care. American Enterprise Institute, Washington, DC, 1983, pp. 198­224. 202 · Health Economics in Development [17] Jamison, D.T., Mosley, W.H., Measham, A.R. and Bobadilla, J.L., (Eds.), Disease Control Priorities in Developing Countries. Oxford University Press, New York, NY, 1993. [18] Enthoven, A.C., Theory and Practice of Managed Competition in Health Care Finance. North Holland, New York, NY, 1988. [19] Oregon Health Services Commission, "Prioritization of Health Services: A Report to the Governor and Legislature, State of Oregon." Portland, OR, 1991. [20] Hadorn, D.C., "Setting Health Care Priorities in Oregon: Cost-Effectiveness Meets the Rule of Rescue." Journal of the American Medical Association, 265 (1991) 2218­2225. CHAPTER 11 Is the Eradication of Polio in the Western Hemisphere Economically Justified? Acknowledgments The central arguments in this article were developed in a previous draft (July 1986) that benefited from discussions with Ciro de Quadros, Marjorie Pollock, William P. McGreevey, Phillip Nieburg, T. Stephen Jones, and Alfred Thieme. None of them bears responsibility for the ideas presented here or even necessarily agrees with those ideas. I am further indebted to Luis Locay, Warren Sanderson, Dieter Zschock, and Dennis Young for helpful comments when the draft was presented (in May 1987) to a seminar at Averell Harriman College for Policy and Management at the State Uni- versity of New York in Stony Brook, New York, USA. Introduction Since the eradication of smallpox in the 1970s, no other disease has been eliminated from the world by vaccination. Advances in mass immunization campaigns using oral vaccine have successfully interrupted the transmis- sion of wild poliomyelitis virus in many countries, however, and have sharply reduced its incidence in many others (1,2). The benefits of polio immunization appear from some studies to outweigh its costs (3). And the Reprinted, with permission, from Bulletin of the Pan American Sanitary Bureau 22(1), 1988. 203 204 · Health Economics in Development cost-effectiveness of mass campaigns relative to other means of reaching the susceptible population has been established, at least in some circum- stances (4). It therefore seems possible, by a suitable intensification of such efforts, to eradicate polio--if not all over the world, then at least in the Western Hemisphere--within the next few years. In view of the success of the PAHO/WHO Expanded Program on Immunization (EPI) in the Americas since its inception in 1977, in April 1985 (5) PAHO recommended that its Member Governments support a five-year, US $46 million campaign to eliminate polio entirely from the Americas, after which it would be relatively easy to deal with whatever cases might be imported. The Member Governments ratified this proposal in September 1985 (6); and since then PAHO has been developing the cam- paign's detailed strategy and obtaining financial commitments from private, bilateral, and multilateral donor agencies. To satisfy some of these agencies' requirements, a cost-benefit analysis was prepared. This article describes the assumptions and findings of that analysis, which indicate the eradication of polio is economically justified, and discusses some of their implications. Its concluding section considers the terms according to which the eradication of polio can be deemed an alternative to curative care. Assumptions The analysis that follows attempts to answer one specific question: Is the cost of eradicating polio, through the program adopted by PAHO, justified through the medical costs saved by not having to treat or rehabilitate polio victims? This estimate of the benefits from polio eradication takes no account of the gains from reduced pain and suffering, from the greater eco- nomic productivity of individuals who would otherwise be paralyzed and rendered unproductive, or from the reduction in other vaccine-preventable diseases that can be expected to result from a successful campaign against polio. If eradication is economically justified by reduced medical costs alone, then there is no doubt that it is still more justified when account is taken of other benefits. The logic of the argument is as follows. For each of the five years of the eradication campaign, and then for each of 10 years thereafter, estimates are made of the following: the number of cases of paralytic polio that would be prevented; the cost of treating and rehabilitating that number of polio The Eradication of Polio in the Western Hemisphere · 205 victims; the cost of the eradication effort; and the net benefit (in terms of reduced medical expenses minus the cost of eradication). These net bene- fits are then discounted at 12% per year, meaning that $1.00 saved next year is worth only $0.88 saved today, etc.1 (This is the discount rate used by the Inter-American Development Bank for project evaluation, and is chosen because the Bank is helping to finance the eradication campaign.) Because the campaign is superimposed upon continuing national efforts to control polio through the EPI, two cost-benefit calculations are made. One compares the total cost--US $74 million in national effort and $46 million from international donor agencies over the first five years, plus $10 million per year in national resources thereafter--with the total cases and costs that could be expected to occur in the absence of any substantial effort to control polio. The other calculation finds costs derived from the current estimated incidence of polio and compares these to the cost of the resources being sought from donors in order to ensure polio eradication by eliminating the relatively small numbers of cases that still persist after nearly a decade of the EPI. The first calculation compares total costs to total benefits (in terms of reduced medical expense), while the second com- pares marginal or incremental costs to the marginal benefits of going from the present case incidence to eradication. In both cases it is assumed that every polio victim would receive treatment, so that the comparison is really between the cost of preventing polio and the cost of treating all those who would otherwise get the disease. This assumption that all victims receive treatment is relaxed later, so that no benefit is attributed to cases not actu- ally treated. A number of additional assumptions underlying the calculations deserve further explanation. These assumptions are as follows: 1. The background or "natural" level of polio incidence is derived from the situation existing before the EPI began, when about 3,000 cases of paral- ysis and 350 deaths were reported annually in the Americas. It is recog- nized that before the EPI started, polio was greatly underreported (5), perhaps by a factor of five.2 If this estimate is accurate, then the true pre- EPI incidence would have been about 15,000 cases per year. This should probably be regarded as an upper bound. 2. The EPI helped to reduce the hemispheric total of reported polio cases to about 500 a year in 1984 and 1985. Assuming no change in the degree of underreporting, this means the actual incidence would have been 206 · Health Economics in Development about 2,500 cases annually. Of course, the improved surveillance that accompanied the EPI might also have reduced the underreporting signi- ficantly, so that the true incidence might have been lower, say on the order of 1,500 cases per year. It is also possible that expecting the level to remain at 2,500 cases per year for the near future without eradication is being over-optimistic. In the absence of an eradication campaign, national efforts might not be able to keep the incidence that low. For one thing, polio fluctuates cyclically, and the 1984­1985 level of cases appears to represent a cyclical trough from which a slight rebound could be expected. For another, vaccination coverage could actually decline because of financial difficulties and a false sense of confidence about the extent of control in the absence of a reliable surveillance system. The example of Jamaica illustrates this latter risk. After over five years of reporting zero cases, levels of coverage declined; an outbreak then occurred in 1982 that produced over 50 cases. The cost of controlling the epidemic and treating the victims has been estimated at more than ten times the cost needed over the preceding five years to prevent the outbreak (7). There were substantial polio increases in Brazil and Colombia from 1985 to 1986 (8), although these were partly offset by declines in Mexico, Haiti, and Peru. The buildup of a pool of susceptibles and any decline in vaccination coverage would have the same results in other countries. For this reason it is assumed that 3,000 cases a year remain to be eliminated, rather than the 2,500 figure that would result from 500 observed cases with 80% underreporting. This 3,000 figure should also probably be regarded as an upper bound, the lower bound being about half as high. The external funds to be utilized in the eradication project will go to ensure that a surveillance system is built up and that supervisory systems are in place--so as to guarantee continued high levels of coverage and eventual eradication of the wild poliovirus. Without these additional resources, it may be very difficult for the countries involved to organize the needed sur- veillance systems, and it is to be feared that prevailing levels of coverage will decline for lack of supervisory systems. 3. The cost of treating a polio victim has been estimated from a 1982 study conducted in Brazil (9). The expenses included were those of treatment during the acute phase of the disease (US $880 on the average, within a The Eradication of Polio in the Western Hemisphere · 207 range of US $350 to US $2,800 in different hospitals), together with those of surgery, rehabilitation, and subsequent therapy. Rehabilitation sometimes extends over several years, so rehabilitation costs must be discounted. In the Brazilian study, discounting was done at 6% per year and was applied over 10 years; for purposes of the present cal- culations, the results reported have been adjusted to reflect the discount rate of 12% used here. With that adjustment, the average cost of the surgery, rehabilitation, and therapy phase is estimated at $4,949. Hence, the total estimated cost of treating a polio case is $5,829. The combination of such a high individual treatment cost and a large number of unreported cases means, of course, that the estimated total cost of treating all polio victims would be quite high. In that sense, cost and case estimates could bias the results in favor of the eradication campaign. How- ever, any such bias is offset by excluding from consideration all of the other costs associated with paralytic polio. Furthermore, the calculations based on these high estimates are modified later in the analysis in order to deter- mine whether eradication would still be justified with fewer cases or lower treatment costs. 4. The five-year eradication campaign could not expect to bring the inci- dence of polio down to zero in the first year. Instead, as surveillance and coverage expand, the campaign is expected to achieve zero cases by the fifth year. It is assumed for purposes of simplicity that the decline in cases is linear over the five years of the campaign. This means that net bene- fits increase from the first to the fifth year, increase sharply in the sixth year (when spending drops back to the maintenance level of US $10 mil- lion per year), and thereafter remain constant, apart from discounting. At the end of the campaign, some 15,000 cases per year are being prevented. If there are currently about 3,000 cases annually, the difference of 12,000 cases that do not occur can be attributed to the current level of control exerted through the EPI. The eradication campaign is assumed to pre- vent another 1,000 cases in the first year (for a total of 13,000) and an additional 500 cases in each succeeding year. 5. Benefits and costs are calculated for 10 years beyond the end of the eradi- cation campaign, which carries the calculation to the end of the century. Net benefits continue to accrue beyond that point if eradication is main- tained, but discounting makes their present value quite small ($1.00 is worth only $0.18 fifteen years from now). 208 · Health Economics in Development Costs and Benefits When All Polio Victims Are Treated The foregoing assumptions lead to the costs and benefits shown in Table 11.1. When total costs (including national efforts as well as donor contri- butions to eradication) are compared to total savings (assuming all cases are treated), there is a net present benefit, after discounting, of US $217.2 mil- lion in the first five years, and the eradication campaign is economically justified in each of the first five years, well before full eradication is achieved. This simply reflects the fact that the current level of EPI cover- age is economically justified by the potential savings in treatment costs on the basis of the assumptions made here. During the 10 years following eradication, these calculations indicate a further discounted net saving of US $264.2 million. Because of discounting, this is much less than the estimated sum of undiscounted savings over 10 years, which is US $774.4 million. (Savings in each year of the decade would be US $77.4 million, but discounting would reduce their present value to only US $41.5 million in the first year and even smaller values in each subsequent year.) Discounting would also reduce the present value of anticipated savings during the five-year eradication effort from US $288.0 million to US $217.2 million. During the whole fifteen-year interval, the present value of net savings is estimated at US $481.4 million. Whether the prevention of 220,000 cases of paralytic polio would be worth this much (or more, or less) to the poten- tial victims and their families is not considered. This estimate is so high because it is very expensive to treat even one polio victim. However, the conclusion that eradication is justified does not depend on this cost being as high as US $5,829. In fact, if the treatment cost were only US $1,728 the net total discounted savings over the five-year eradication campaign would be zero, and the effort would still pay for itself over the next 10 years. (Net savings in each year would be US $15.9 mil- lion, for a ten-year discounted total benefit of US $54.3 million.) If the campaign had the entire 15 years to pay for itself, the cost of treatment could be as low as US $1,207. Alternatively, the incidence of polio could be much lower than is assumed in Table 11.1. That is, assuming treatment costs of US $5,829 per patient, the number of cases could be reduced to about 3,100 and eradica- tion could still be justified. Given that the current level of polio control by vaccination is much cheaper than treatment of all the cases that would otherwise occur, it may 15 AL 40.0 46.0 62.1 ALL YEARS, TOT 220.0 220.0 481.4 233.2 187.2 1,282.4 1,062.4 Ensuing AL an 0 onths.m YEARS 6-15, ---- TOT 30.0 55.2 150.0 874.4 100.0 774.4 264.2 174.9 174.9 six and 6- for PER b c c -- 15.0 87.4 10.0 77.4 -- 3.0 0 -- YEARS 15, 17.5 17.5 ANNUM annually Campaign AIGN AL ear -- 6.9 70.0 10.0 58.3 46.0 12.3 12% YEARS, TOT 408.0 120.0 288.0 217.2 CAMP at Five-Y 0.601 3.0 9.2 8.3 5.0 15.0 87.4 24.0 63.4 38.1 17.5 discounting shown. to Successful are a 0.673 2.5 9.2 5.4 3.6 totals AIGN 14.5 84.5 24.0 60.5 40.7 14.6 During CAMP corresponds ten-year eatedrT TION 345 one 15. 0.753 2.0 9.2 2.5 1.9 the 14.0 81.6 24.0 57.6 43.4 11.7 are year year Eradication ERADICA for in only, OF Polio ictimsV year 2 0.844 1.5 8.7 9.2 0.5 0.4 to YEARS 13.5 78.7 24.0 54.7 46.1 factor 0.194 the to with Polio year six all thus from 1 0.945 1.0 5.8 9.2 3.4 3.2 year 13.0 75.8 24.0 51.8 48.9 in factor Associated Assuming mid-years, at 0.536 discount Benefits US$): US$) US$) from Period, of US$): the and of of marginal of calculated in varies total savings savings are Costs a millions cases (thousands) treatment (millions value (thousands) (millions value factor versus (in of versus millions cases in treatment variations Maintenance factor eradication benefit) (in of in factors 11.1 of eradication of maintenance Saving present discounted costs benefit) maintenance costs of prevented expenses or (net of prevented expenses or saving (net present discounted ear of discount Number Savings Cost Net Net ableT en-YT Discount otalT benefits Donor benefits Number Savings Cost Net Net Discount This Because a b c 209 210 · Health Economics in Development still be asked whether the donor contribution of US $46 million for the eradication campaign would pay for itself in terms of reduced incidence and associated lower treatment costs. The amount requested from donor agen- cies is 38% of the total cost of eradication, but it would be used to eliminate only 20% of the pre-EPI level of incidence (3,000 cases per year out of 15,000), the other 80% being controlled by national efforts costing US $74 million during the five years of the campaign. Consequently, as the second part of Table 11.1 shows, the donor contri- bution exceeds the anticipated saving (again assuming treatment of all polio victims) during each of the campaign's first two years. This is followed by positive net benefits as eradication is achieved, for a total net benefit of US $6.9 million during the five years of the campaign. There would also be a positive net benefit of US $17.5 million in each subsequent year, while eradication would presumably be maintained by national efforts without further donor financial assistance. Even with discounting, over the next decade this latter benefit would amount to a further US $55.2 million. The result is an estimated net discounted positive benefit of US $62.1 million over the entire fifteen-year period. This calculation naturally depends more for its positive value upon the assumed high cost of curative treatment. For the campaign to break even in five years (showing zero net discounted savings from the donors' contribu- tions), treatment could not cost less than US $4,874. That would result in total discounted savings of US $46.1 million over the ensuing decade. These savings (over all 15 years) would still be positive at any treatment cost higher than US $2,106. Therefore, roughly speaking, the eradication of polio appears justified if treatment costs at least US $2,000, purely on the grounds of reducing the total discounted costs of treatment plus prevention over a fifteen-year period. In other words, while the donor contribution directed at eradication pays off more slowly than the level of polio control already achieved, it is still an economically justified investment compared to the cost of treating everyone who would otherwise get polio. This remains true despite the need to devote resources to activities other than immunization (surveillance and laboratory work) that are necessary in order to ensure that eradication is achieved and maintained. Moreover, as in the previous calculation, no account is taken of other benefits anticipated from this investment--such as an increased capa- city to control other diseases and consequent further medical savings. In sum, there is no reason to suppose that the current level of polio con- trol has already absorbed all the potential benefit, leaving nothing more to The Eradication of Polio in the Western Hemisphere · 211 be gained from complete eradication of the disease. Rather, making an addi- tional effort to eliminate polio entirely appears justified. This conclusion seems invalid only if the cost of treating a polio victim is much less than that assumed here, or if the number of victims treated is far smaller. The next section considers the second of these possibilities. Costs and Benefits When Treatment is Incomplete The apparent economic justification for eradicating polio contradicts the findings of a cost-benefit analysis of polio vaccination in Brazil covering the mass immunization campaigns begun in 1980 (10). On the assumption that such campaigns would end in 1983, and that thereafter the normal, pre- campaign rate of vaccination would be enough to keep polio from reap- pearing before 1990, it was concluded that the mass campaigns did not justify their costs (US $30 million), and that no more than US $43.4 million per year should be spent to maintain the pre-1980 level of vaccination coverage. This study made several assumptions that differ from those reflected in Table 11.1. For one thing, the discount rate was taken to be 18% rather than 12%, which made future benefits less valuable. For another, the fixed costs of treating children in the acute stage of the disease were assumed to be zero, because of supposed excess facility and staff resources at pediatric hospitals. This amounts to supposing that any such resources would not be used for other medical care and would not be released--in other words, that ministries of health would maintain superfluous staff and facilities. (The costs of the subsequent rehabilitation and surgery, which were acknowl- edged to require specialized personnel and facilities, were not regarded as zero.) The most important difference, however, is that savings were calcu- lated only for the estimated number of actual curative treatments, rather than for the number of victims who could benefit from such care but did not always receive it. Primarily for this reason, the mass vaccination cam- paign appeared to be justified in Northeast Brazil, where the incidence of polio was relatively high, but not in the rest of the country. Relating the costs of immunization to actual rather than potential costs of treatment in this way raises two important issues. The first concerns the appropriate way to deal with those victims who get polio but receive no medical care. These people are entirely ignored if only actual spending is considered, but of course they account for much of the potential benefit of 212 · Health Economics in Development immunization if any kind of price is put on pain and suffering (10). The sec- ond issue concerns marginal costs and benefits. Once polio has been brought partly under control by immunization, the remaining gain from greater cov- erage may be small. However, the cost of obtaining greater coverage is likely to be high, since the current level has to be maintained while immunization is extended to the rest of the population. This makes polio very different from smallpox, which could be combated by concentrating only on those areas still reporting the disease (only surveillance activities, not vaccination, were needed in areas where smallpox had already been eradicated). As a result of this problem of increasing marginal cost and decreasing marginal gain in the case of polio, it may never seem justified to finish the job. The calculations in Table 11.1 indicate, however, that complete eradi- cation is justified for polio if the extra expense of donors' contributions is compared to the extra gain made possible thereby. Among the benefits from complete rather than almost-complete eradication are the prevention of later outbreaks like the previously mentioned one in Jamaica. These can be expensive to control, but because of their uncertain magnitude and fre- quency no attempt has been made to estimate the present discounted value of the costs they represent. It should also be noted that because of uncertainty in the estimates of costs and incidences it is impossible to determine either the marginal point where preventive efforts cease to be justified or the maximum vaccination coverage that pays for itself. Both calculations in Table 11.1 assume that treatment would be provided to everyone who actually contracted the disease. Most of the estimated benefits, however, are only potential savings that greatly exceed realizable savings attainable through actual reduction in treatment expenditures. Therefore, the next task is to see whether those realizable savings, by them- selves, are enough to pay for the cost of eradication, without attributing any benefit to cases where people are affected by polio but receive no medical care. This requires estimating the number of cases that are or would be treated. Before introduction of the EPI, the number of cases treated was roughly equal to the number reported, in part because some countries reported only those cases actually treated. (This accounts in large part for the very high level of underreporting.) In the absence of control measures, the number of cases treated would be at least as large as it was a decade ago. Allowing for some improvement in coverage or expansion of treatment, and recognizing that in many pre-EPI years there were more than 4,000 The Eradication of Polio in the Western Hemisphere · 213 reported cases of polio, it seems reasonable to take 4,000 cases per year as the background or "normal" level of treatment that would occur in the absence of immunization. As in the calculations reported above, it cannot be supposed that the eradication campaign would immediately eliminate the need to treat those cases. Instead, it is supposed that in the first year of the campaign there would be savings from 2,000 fewer treatments, and that this number would rise to 4,000 cases over the five-year period. This esti- mate, which appears in the first line of Table 11.2, shows that over the entire fifteen-year period some 55,000 fewer treatments would be required. Following this assumption, net savings are of course much smaller than they would be if all polio victims were treated. Savings remain negative throughout the five years of the campaign and turn positive thereafter. The result is a total net discounted benefit of ­US $27.3 million during the eradication campaign, followed by a positive net benefit, after discounting, of US $45.4 million during the next decade. Total net benefits during the entire fifteen-year period are estimated at US $18.1 million. This means that eradication of polio would pay for itself by reducing the medical costs of treating those victims who actually are or probably would be treated. Hence, in order to justify an eradication campaign, it is not nece- ssary to attribute any benefits to people who probably would not receive treatment. The magnitude of the net discounted benefit is drastically reduced (from US $481.4 to US $18.1 million), but it continues to be pos- itive. However, because the number of treatments is much reduced, the cost per treatment could not fall appreciably without turning savings negative; specifically, the minimum cost would be $5,097. Assuming that some polio victims are not treated has exactly the same effect on estimated savings as assuming that fewer people get polio in the first place. Ethically, of course, the two situations are very different; and the total benefits are also different once pain and suffering are taken into account. Nevertheless, for the purpose of this analysis the two are identical. Thus, the calculations in Table 11.2 can be interpreted as meaning that polio eradication would be justified if there were only 4,000 cases annually in the absence of vaccination with all cases being treated--in which case only about 1,000 cases would remain to be prevented by the eradication campaign. This implies that the results do not depend critically on the assumed high incidence of unreported polio; and so, as noted earlier, the level could be as low as 3,100 cases per year. The profile derived in Table 11.2 of immunization expenditures and savings ascribed to reduced treatment costs is displayed graphically in 15 AL 0.6 ALL 55.0 18.1 14.0 81.6 46.0 35.6 YEARS, TOT 320.7 220.0 100.7 Maintenance AL 0 ear 11.1 YEARS 6-15, TOT 40.0 45.4 10.0 58.3 58.3 18.3 233.2 100.0 133.2 en-YT ableT 6- and in PER 4.0 a 1.0 5.8 0 5.8 a YEARS 15, 23.3 10.0 13.3 ANNUM -- -- Shown Campaign as AIGN AL 4.0 YEARS, TOT 15.0 87.5 32.5 27.3 23.3 46.0 22.7 17.8 ear 120.0 Same CAMP Five-Y the is 4.0 0.7 0.4 1.0 5.8 9.2 3.4 2.0 shown. 23.3 24.0 are Factor Successful a totals AIGN 3.5 3.6 2.5 0.9 5.2 9.2 4.0 2.6 20.4 24.0 Discount CAMP During ten-year The TION the 3.0 6.5 4.9 0.8 4.7 9.2 4.5 3.4 17.5 24.0 eated.rT ERADICA only, Eradication OF year are to Polio YEARS 2345 2.5 9.4 7.9 0.7 4.1 9.2 5.1 4.3 14.6 24.0 with ictimsV year from Polio 1 2.0 0.6 3.5 9.2 5.7 5.4 all 11.7 24.0 12.3 11.6 factor Associated of discount Benefits Fraction a US$): US$) US$): US$) the of of of of and in only total savings marginal savings Costs millions treatments (thousands) treatment (millions value versus millions treatments (thousands) treatment (millions value versus (in of in eradication (in of in variations eradication 11.2 Assuming of benefit) benefit) of maintenance Saving present costs discounted costs of maintenance saving present discounted prevented expenses or (net of prevented expenses or (net of Number Savings Cost Net Net ableT otalT benefits benefits Number Savings Cost Net Net Donor Period, Because a 214 The Eradication of Polio in the Western Hemisphere · 215 Figure 11.1. The upper panel of that figure shows the un-discounted pro- file while the lower one includes the effect of discounting. As a result of dis- counting, the area of net gain is shrunken compared to the area of net loss. Figure 11.1 Costs and Benefits of Polio Eradication, Assuming Treatment of Only Some Victims or Reduced Numbers of Cases (from Data in Table 11.2) Costs of savings, in millions of U.S. dollars A. Without discounting Savings in treatment 24.0 (Net loss) (Net gain) Immunization costs 11.7 Eradication phase Maintenance phase (first 5 years) (subsequent years) 0 1 2 3 4 5 6 7 8 9 10 Years B. Discounting at 12% per year 22.7 Immunization costs (Net loss) Savings in treatment 11.1 (Net gain) 0 1 2 3 4 5 6 7 8 9 10 Years 216 · Health Economics in Development Table 11.2 also compares marginal (donor) costs and marginal benefits in the manner of Table 11.1, assuming only a small number of cases treated. Here it is supposed that the reduction in treatments never exceeds 1,000 cases per year, starting from an approximate reduction of 600 cases the first year. This calculation shows a total discounted net benefit of only US $0.6 million over the entire 15-year period. Based on reduced medical costs alone, the donors' contribution almost exactly pays for itself, assuming treatment costs of US $5,829 per case. In summary, these estimates indicate that the eradication of poliomyelitis is a justifiable investment, even without making any allowance for benefits other than those due to realizable reductions in expenditures to treat victims of the disease. Indeed, the cost of treating even a small fraction of those who need treatment is large enough to pay for the total prevention of polio. In other words, the eradication of polio would actually put money in the cof- fers of the Ministry of Health, or whoever now pays to treat polio victims. It is important, however, to sound a note of caution. This projected result depends on there being enough current expenditure on treatment. It would no longer hold, for example, if the level of treatment were only one- fourth lower than that assumed in Table 11.2. This process of justifying an eradication campaign by its effect in reducing public expenditure depends on there being sufficiently high public expenditure to start with; and so the process can lead to effects that are clearly perverse. In our case, literally applied, and giving no allowance for non-monetary benefits in terms of reduced pain and suffering, it implies that the eradication of polio would be justified after spending millions of dollars over many years to treat polio vic- tims, but would not be justified as an alternative to such a treatment expenditure. That is, eradication would be more justified the later it came, after increasingly large sums of money were spent for treatment. Immediate Versus Delayed Eradication To see how such justification of immunization, in terms of reduced costs alone, could lead to a delay in vaccination efforts, consider two hypotheti- cal regions (A and B) with 15,000 cases of paralytic polio per year (the esti- mated pre-EPI level in Latin America and the Caribbean). Suppose that immunization has not begun in either region, and suppose further that in neither case are victims of the disease initially being treated. The costs of treating a case (US $5,829), conducting a five-year eradication campaign The Eradication of Polio in the Western Hemisphere · 217 (US $120 million), and maintaining eradication thereafter (US $10 million per year) are assumed to be the same as in the previous analysis. In Region A, efforts are made to start treating victims for purely ethical reasons, treatment being extended to 1,000 patients the first year and 1,000 additional patients per year thereafter. At the end of five years, someone performs a cost-benefit estimate of the sort presented above and discovers that it would be cheaper to immunize people. Over the next five years, immunization is gradually extended to enough of the population to inter- rupt the transmission of wild poliovirus, and eradication is achieved. There- after, immunization of infants is maintained, and while treatment continues for the victims accumulated during the whole ten-year period (five without immunization and five after immunization began), no new patients are admitted for treatment in the eleventh and subsequent years. Assuming lin- ear treatment and immunization trends, as shown in Table 11.3, 112,500 people get polio during the decade of whom 41,500 are treated and 71,000 receive no treatment. In Region B, nobody worries about cost-benefit analysis of this sort. Immunization is begun immediately, rather than waiting for five years. Treatment of victims begins at the same time and is extended at the same rate as in Region A, except that because of immunization, treatment never rises beyond 4,000 cases per year and falls to zero in the sixth year. Over the ten-year period only 37,500 people get polio, of whom 11,500 receive treat- ment and 26,000 do not. This latter figure is only 37% of the number of untreated victims accumulated in Region A. From year 11 onward both regions are identical, in that they have no new polio cases and spend US $10 million each per year to maintain eradication. Any comparison of the two regions need therefore consider only the first 10 years. What do costs look like in the two cases? Region A spends a total of US $241.9 million on treatment and US $120.0 million on immunization over the decade, for an undiscounted total cost of US $361.9 million. Region B spends only $67.0 million on treatment (just 28% of what Region A spends), but--since five years of maintenance are included after the five years of eradication--it spends US $170.0 million on immunization, US $50.0 million more than Region A. Region B's undiscounted total expendi- ture is therefore US $237.0 million, or 66% as much as Region A's. Discounting expenditures at 12% per year has more effect on the costs in Region A, because spending there reaches its peak later, in year eight. This is due to initial postponement of the eradication campaign and also to the relatively slow expansion of treatment that is assumed; costs would and AL TEN- YEAR TOT 41.5 71.0 37.5 11.5 26.0 67.0 112.5 241.9 120.0 361.9 183.6 170.0 237.0 162.7 Immediate 01 1.5 1.5 0 8.7 3.4 24.0 32.7 11.1 10.0 10.0 One 4.5 4.5 0 3.8 26.2 24.0 50.2 19.2 10.0 10.0 7.5 7.5 0 4.3 43.7 24.0 67.7 28.9 10.0 10.0 Campaigns, 7.0 3.5 4.8 10.5 40.9 24.0 64.8 31.0 10.0 10.0 Eradication 6.0 7.5 00000 00000 00000 5.4 13.5 35.0 24.0 59.0 31.6 10.0 10.0 Polio YEARS 5.0 1.5 1.5 000000 8.7 15.0 10.0 29.1 29.1 17.5 24.0 32.7 19.7 Hypothetical 11.2 4.0 4.5 4.0 0.5 15.0 11.0 23.3 23.3 15.7 23.3 24.0 47.3 31.8 woT and of 3.0 7.5 3.0 4.5 11.1 15.0 12.0 17.5 17.5 13.2 17.5 24.0 41.5 31.3 Results ablesT 2.0 9.9 2.0 8.5 15.0 13.0 11.7 11.7 10.5 11.7 24.0 35.7 30.1 and in as 123456789 1.0 5.8 00000 5.8 5.5 1.0 5.8 Costs earY 15.0 14.0 13.5 12.5 24.0 29.8 28.2 ear per en-YT of 12% US$) US$) at (thousands) US$) (thousands) US$) (thousands) (thousands) (million campaign): (million campaign): cases treated (thousands) (million cases treated (thousands) (million Comparison Discounted A polio cases (delayed ictimsV otalT polio cases ictimsV otalT of of (immediate of of 11.3 A treatment immunization B treatment immunization of of Delayed, Cost of of Cost ableT Region Number Number Untreated Cost Cost otalT Discounted Region Number Number Untreated Cost Cost otalT Discounted One 218 The Eradication of Polio in the Western Hemisphere · 219 be shifted toward the early years if treatment were extended more rapidly. In Region B, total expenditure reaches its peak in year four, being higher than in Region A during each of the first five years. As a conse- quence, the discounted total costs are US $183.6 million in Region A and US $162.7 million in Region B, so that the Region B costs are 89% as high as those in Region A. At the end of 10 years, neither region has any new polio cases. However, Region B is clearly better off. It has spent US $20.9 million less after dis- counting (US $124.9 million less without discounting); it has 30,000 fewer treated polio victims (who suffer some damage from the disease despite treatment); and it has 45,000 fewer untreated, paralyzed victims. Thus, making an immediate effort to eradicate the disease pays off both in reduced health damage and in lower total treatment and prevention costs. If either the cost of treating a polio victim or the number of victims were lower, the monetary saving in Region B compared to Region A would of course be smaller, but it would always be positive. If one assumed that eradication could only be justified by saving actual (not potential) expenditure on treat- ment, however, Region A pursued the right course by not starting immuni- zation until the costs of treatment had become relatively high. Concluding Remarks What accounts for this perverse result? Part of it is due to discounting future costs and benefits. When the assumed number of treatments is reduced from 15,000 per year (Table 11.1) to 4,000 per year (Table 11.2), the un-discounted net savings fall from US $1,062.4 million to US $100.7 million. (This is much more than the approximately 15:4 reduction in treatment savings, because the costs of immunization are independent of treatment levels). Discounting means that net savings are reduced much more than ten-fold, because savings increase through time; thus, net savings of US $481.4 million become only US $18.1 million, a 24-fold reduction. What this means is that the higher the discount rate, the higher the num- ber of current treatments necessary to justify the cost of eradication. If, as in Region A, immunization is delayed while treatment costs increase, the effect of discounting is to delay eradication still more. It might seem that the answer to this problem is not to discount the future, but instead to base decisions on un-discounted costs and benefits. The logic of discounting, after all, supposes that a given individual, who is 220 · Health Economics in Development the same person today and tomorrow, values tomorrow less than today (11). But the children who will suffer paralytic polio in the future, if the disease is not eradicated, have for the most part not been born yet. Discounting their future therefore means valuing them less than those are already here, which is very different from making inter-temporal choices for a given person. However, to abandon discounting means being willing to wait forever, provided that eventually benefits outweigh costs. The resources required to eradicate polio could be applied to other uses, including medical uses, which might pay off more quickly. So even though discounting the future raises an awkward ethical question, there is no escaping the need to give priority to the present, at least so long as the benefits considered in the two periods can be compared. The whole question of whether eradicating polio is worth the cost would not even arise if the private market for immunization worked properly. No parent wants to see his child paralyzed, and the cost of immunization is less than the expected cost of treatment per un-immunized child. Therefore, every parent should be more than willing to pay to have his child protected.3 If this does not happen, the fault lies with some combination of poverty and ignorance. It is true that public expenditure to eradicate the disease takes resources away from competing private uses. But requiring that such expen- diture "pay its way" amounts to supposing that the alternative private expen- ditures would be equally justified--which seems questionable in a world where private demand has not yet caused all susceptible children to be immu- nized. And if the rationality of private spending is to be doubted, then it is not clear why public spending must produce positive discounted net benefits. But the most important reason why the eradication of polio may not appear to be economically justified (as in Region A) does not arise from dis- tinctions between the present and the future or between public and private expenditure. It arises from the different way that curative treatment and preventive activities are judged. The "justification" for immunization is that it costs less than treating polio victims. If the aim is to minimize the expen- diture required to avoid paralysis or death from polio, then eradicating the disease is clearly preferable to continued curative treatment. But if the aim is to reduce public expenditure on health, then immunization appears to be justified only if curative spending is high enough. In general, the foregoing account assumes that some level of treatment will be provided, with or without economic justification, and then applies an economic test to see whether prevention should replace treatment as the The Eradication of Polio in the Western Hemisphere · 221 way to deal with polio. Why should this be considered the right test of an eradication campaign's merit? After all, treating polio victims does not save money for the government; and if the aim were simply to reduce expenditure, then curative care could not be justified either. But if "it is unacceptable, given the technology presently available, that any child in this hemisphere should suffer paralytic poliomyelitis" (5, p. ii), then the eradi- cation of polio is not only ethically justified but also economically sound. Notes 1. If benefits in year t are designated Bt, while costs incurred in that year are Ct, Bt Ct is the net benefit. The corresponding discounted net benefit is (Bt Ct)/ (1 i)t, where i is the interest or discount rate used. The present value of this stream of net benefits (positive or negative) is the sum of these terms over all the years of a project, or in the present case through the first 15 years, after which net benefits are positive but, because of discounting, are quite small. 2. It is not easy to estimate polio underreporting, although surveys of residual lameness provide a basis (see 1, Session III, Section A). The assumption that before the EPI only about 20% of the polio cases were reported in Latin America and the Caribbean has been suggested by Ciro de Quadros and Marjorie Pollock as a rea- sonable estimate. 3. Once coverage by vaccination is almost complete a parent might consider that his unvaccinated child was adequately protected by the screen of vaccinated chil- dren, so that there would be no further gain from the child's immunization. This argument would apply, if ever, only when coverage was complete enough so that the risk of infection was essentially zero; it would not hold at the typically quite incom- plete levels of coverage found in Latin America. Even at higher levels of coverage, this argument would make sense only if the cost of having the child vaccinated were high compared to the benefit, or the risk of paralysis from the vaccine itself were substantial. References [1] Horstmann, D.M., T.C. Quinn, and F.C. Robbins (eds.). International Symposium on Poliomyelitis Control. Review of Infectious Disease 6 (Suppl 2), 1984. [2] Sabin, A.B. Oral poliovirus vaccine: History of its development and use and the current challenge to eliminate poliomyelitis from the world. Journal of Infectious Disease 151:420­436, 1985. 222 · Health Economics in Development [3] Willems, J.S., and C.R. Sanders. Cost-effectiveness and cost-benefit analysis of vaccines. Journal of Infectious Disease 144:486­493, 1981. [4] Creese, A.L. Cost effectiveness of alternative strategies of poliomyelitis immu- nization in Brazil. Review of Infectious Disease 6 (Suppl 2): S404­407, 1984. [5] Pan American Health Organization, Expanded Program on Immunization in the Americas: Progress Report, PAHO document CE 95/15, 11 April 1985. Washington, D.C., 1985. [6] Pan American Health Organization, PAHO member countries endorse polio eradication resolution. EPI Newsletter vol. 7. no. 5. October 1985. [7] Ashley, D., and R. Bernal. Poliomyelitis in Jamaica: immunization policies and socioeconomic implications. World Health Forum 6:265-267, 1985. [8] Pan American Health Organization. Polio in the Americas: Weeks 1­53, 1986. EPI Newsletter vol. 8. no. 6, 1986. [9] Ministry of Health, Brazil, Memória sobre estimativa de custos dos casos de poliomielite no Brasil em 1982. Brasília. 1984. [10] Garlow, D.C. Mass Vaccination to Combat Polio: A Cost-Benefit Analysis for Brazil. Mimeographed document. Instituto de Pesquisas e Estudos Econômicos. Universidade Federal de Rio Grande do Sul, 1983. [11] Prest. A.R., and R. Turvey, Cost-Benefit Analysis: A Survey. In: American Economic Association and Royal Economic Society. Surveys of Economic Theory (vol. 3). St. Martin's Press. New York. 1967. CHAPTER 12 Cost-Benefit Analysis of a Regional Vaccination System Acknowledgments I am grateful to my former colleagues Robert Knouss, Francisco López Antuñano, Gabriel Schmuñís, Francisco Pinheiro, Ciro de Quadros, Alberto Pellegrini Filho, Victor Escutia, and Mario González Pacheco, as well as Drs. Guillermo Soberón, Akira Homma, and Armando Isibasi, for guidance and for the estimates on which this study is based. Any error, how- ever, is the sole responsibility of the author. Introduction Although mass vaccination of vulnerable populations has been quite suc- cessful at reducing communicable disease morbidity and mortality, we still confront serious public health problems arising from diseases for which no adequate vaccines exist (1,2). Either a vaccine has not been developed, there having been only partial progress to date; or the existing vaccines are of lim- ited effectiveness; or they are restricted to certain pathogenic serotypes and therefore would not protect the populations exposed to other serotypes; or they are too costly for mass application. In view of this situation, the gov- ernments participating in the World Children's Conference held in September 1990 proposed various measures including development of an "infant vaccine" for safe early protection against a variety of diseases (3). Reprinted, with permission, from "Cost-Benefit Analysis of a Regional System for Vaccination Against Pneumonia, Meningitis Type B and Typhoid Fever" Bulletin of the Pan American Health Organization 26 (2), 1992. 223 224 · Health Economics in Development Independently, in early 1989 the Pan American Health Organization began examining the prospects for undertaking a regional effort to perfect and disseminate certain vaccines. These vaccines would be directed against a lim- ited number of diseases of particular interest to PAHO's Member Countries because of the high morbidity and mortality they caused or because of large expenditures needed to treat their victims. This initiative, named the Regional Vaccine System (Sistema Regional de Vacunas­ SIREVA), was seen as includ- ing the phase of epidemiological research in the participating countries, basic research to develop the new vaccines, clinical and field trials, and construction and operation of a pilot production plant to support these other phases. Once a vaccine had been found effective, safe, and affordable, production on a commercial scale would begin, possibly under arrangements with state or private laboratories; and mass vaccination of children would commence, perhaps through an extension of the Expanded Program on Immunization (EPI). The three diseases of bacterial origin considered targets of the effort are pneumococcal pneumonia, meningococcal meningitis caused by Group B Neisseria meningitidis, and typhoid fever. SIREVA's original design included three options designated "A," "B," and "C." Option A dealt only with meningitis and typhoid fever, while options B and C included development of vaccines against all three target diseases. However, option B included only 10 participating countries with 10 collaborating national laboratories, while option C included 17 coun- tries--thus envisioning vaccination of a larger population and implying greater development costs for SIREVA directed at identifying serotypes and testing vaccines in all 17 countries. All three options called for two princi- pal laboratories or vaccine centers to be involved in the project, one asso- ciated with the National Institute of Public Health of Mexico and the other with the Oswaldo Cruz Foundation in Brazil, these being the centers with the greatest experience and technical sophistication in the Region. On analyzing the costs and benefits the three options, it was concluded that the pneumonia vaccine should definitely be included in the system and that three vaccines could be applied in the countries of the Region where the selected diseases are now of major importance. However, it was also decided that the system would be developed in 11 countries, since a network of 10 national laboratories would be sufficient for the epidemiological work and the clinical and field trials. This option became the final version of the proposal (4). The only variation still being considered deals with the number of people who would be vaccinated against meningitis, the cost-benefit calculations being repeated for two possible scales of operation that differ by a factor of Cost-Benefit Analysis of a Regional Vaccination System · 225 two. It should be noted that the limit placed on the number of countries ini- tially involved would not limit later administration of the vaccines in other countries of the Americas, or even in other regions, where the cost of vacci- nation might be justified by the benefits. The purpose of the present analysis is to estimate and compare SIREVA's costs and benefits, and to establish under what circumstances the benefits would justify the costs, and thus justify establishment of the system. With respect to costs, it is necessary to distinguish between two ele- ments: expenditures for vaccine development (including field trials and adap- tation to different epidemiological conditions) and expenditures for vaccine administration (that is, for vaccinating the population). Beyond that, how- ever, it is not the purpose of this presentation to distinguish between differ- ent kinds of costs. Therefore, the account that follows makes no attempt to discuss the composition of the costs attributable to SIREVA (that is, their distribution among basic research, clinical trials, pilot production, etc.); only their distribution over time will be considered. Similarly, we will not consider the biologic, chemical, and epidemiological aspects of the target diseases and prospective vaccines. Rather, the information used in this analysis deals only with the number of people that would be vaccinated and the numbers that would become sick or die if unvaccinated; the costs of implementing SIREVA, vaccinating the population, and treating patients; and possible additional benefits attributable to vaccination. The last part of the analysis estimates the sensitivity of the results to changes in the para- meters utilized, this type of estimate being especially crucial when neither the costs nor the benefits are known but must be estimated with varying degrees of precision. The costs attributable to SIREVA as such (the vaccine development costs cited above) have been estimated for a period of 10 years under the assump- tion that, although the system could continue to function for many more years, the expenditures in the eleventh and following years would be dedi- cated to the development of new vaccines not contemplated in the initial plans. Therefore, it would not be correct to attribute or charge expenses of those future years to the first three diseases, and vaccination against them would have to justify only the expenses of SIREVA's first decade. These expenses, presented by year in Table 12.1, have been estimated at US $115.3 million in constant dollars. Any cost-benefit analysis is based on what are called present values of cost and benefit flows over time, these flows being discounted according to how far in the future they occur. This procedure requires the selection of a discount rate (r), which is conceptually equivalent to an interest rate. A 226 · Health Economics in Development cost t years in the future is then estimated by dividing the present cost (C) by (1 r) for each of t 1 years (5), the discounted cost being C (1 r)(1 r)(1 r) . . . C (1 r)(t 1) C (1 r)( t 1) For example, if an item's present cost (C) were $1,000 and the discount rate were 10% (0.10), then its discounted cost five years in the future (t 5) would be $683, calculated as follows: Year 5 discounted cost C (1 r)( t 1) $1,000 (1 0.10)( 5 1) $1,000 (1.10)( 4) $1,000 1.1(4) $683 Of course, the total discounted cost (C*) over t years would be the sum of the costs in each of the years considered (from here on we shall use an asterisk [*] to designate a discounted sum). And so, if we let the letters CS stand for the SIREVA costs (for vaccine development), the total SIREVA costs appearing in Table 12.1, discounted over the decade, can be expressed by the formula CS* SUM CS(t) (1 r)( t 1) , Table 12.1 Cost of SIREVA, by Year, in Constant US$a COST OF SIREVA (CS), IN YEAR MILLIONS OF US$ 1 5.41 2 26.36 3 11.23 4 10.33 5 10.53 6 10.48 7 10.43 8 10.43 9 10.06 10 10.06 Total (CS, not discounted) 115.32 Total (CS*, discounted) 80.31 aThe analysis only attributes the costs of the first 10 years of SIREVA to the development of vaccines against meningitis, typhoid fever, and pneumonia because it is expected that initiation of vaccination with all three vaccines will occur in the first decade. Cost-Benefit Analysis of a Regional Vaccination System · 227 where CS(t) is the undiscounted cost in year t, (1 r)( t 1)is the discount factor, and SUM indicates the sum of the costs of all the years in question. The same method can be used to calculate discounted benefits. Box 12.1 lists the multiplicative discount factors in the form (1 r)( t 1) for a discount rate of 10% per year (r 0.10) from Year 1 of SIREVA (when the factor is equal to 1.0) to Year 30, which is the furthest horizon consid- ered in this analysis and for which the factor decreases to only 0.063. This means that one dollar of costs or benefit that only occurs in Year 30 would have a present value of $0.063; and conversely, $0.063 invested today at a rate of interest of 10% per year would have a value of $1.00 after 30 years. Box 12.1 Discount Factors, by Year, for a Rate (r) of 10% per Year YEAR FACTOR COMMENTS 1 1.0000 Initiation of SIREVA 2 0.9091 3 0.8264 4 0.7513 5 0.6830 6 0.6209 7 0.5645 Start of vaccination against typhoid fever 8 0.5132 9 0.4665 Start of vaccination against meningitis 10 0.4241 Start of vaccination against pneumonia 11 0.3855 12 0.3505 13 0.3186 14 0.2897 15 0.2633 16 0.2394 From year 16 on, the numbers of vaccinations do not vary. 17 0.2176 18 0.1978 19 0.1798 20 0.1635 Sum for years 16 to 20 0.9981 Sum for years 1 to 20 9.3647 21 0.1486 22 0.1351 23 0.1228 24 0.1117 25 0.1015 26 0.0923 27 0.0839 28 0.0763 29 0.0693 30 0.0630 Sum for years 21 to 30 1.0045 Sum for years 1 to 30 10.3692 228 · Health Economics in Development The last line of Table 12.1 illustrates the effect of discounting the future costs of SIREVA at 10% per year over the course of a decade, which reduces the undiscounted figure (US$115.3 million) to a discounted cost of US $80.3 million. The use of other discount rates would obviously give other totals. How- ever, it has been judged that any reasonable rate would fall between 8% and 15% per year, and that changes introduced by using such rates as extreme values would be a good deal smaller than possible changes introduced by uncertainties regarding the value of other variables. For example, the cost of vaccinating one individual is not yet known, but it is conceivable that it could vary by a factor of 10, while a discount rate of 8% would not differ from one of 15% by as much a factor of two. Moreover, changes in the dis- count rate only affect the relative weights of costs and benefits occurring in the same year. Therefore, the question of whether the benefits justify the costs is not as sensitive to variations in the discount rate as it is to variations in the costs or benefits taken separately. Anticipated Effects of SIREVA According to the projections for SIREVA, it will be possible to begin vacci- nation against typhoid fever in Year 7 of the system's operation. Vaccination against meningitis would begin in Year 9, and vaccination against pneumo- nia in Year 10. In all three cases it is anticipated that vaccination will com- mence at a high rate, so as to reduce the number of susceptible individuals in the existing population. Later, the number of vaccinations carried out would be reduced to focus on newborns at relatively higher risk, though possible fluctuations could be occasioned by future outbreaks. In the case of pneumonia, only one year of high coverage is foreseen. This high coverage phase would extend over five years for the other two diseases, and in the case of typhoid fever there would be a period of inter- mediate coverage followed by a second reduction in the coverage rate after another four years. Table 12.2 shows estimates of the numbers of people who would be vac- cinated against each of the three target diseases in any given year. However, the number of individuals vaccinated does not correspond to the number immunized, because development of vaccines that are 100% effective is not anticipated; rather, the estimates of disease cases and deaths prevented (see Table 12.3) are based on the assumption that the vaccines will be 90% Cost-Benefit Analysis of a Regional Vaccination System · 229 Table 12.2 The Projected Numbers of Individuals to be Vaccinated, by Target Disease and Year NUMBER VACCINATED (NUM), IN MILLIONS, AGAINST: YEAR MENINGITIS TYPHOID FEVER PNEUMONIA TOTAL 7 65 65.0 8 65 65.0 9 39.0­78 65 104.0­143.0 10 39.0­78 65 39.0 143.0­182.0 11 39.0­78 65 19.5 123.5­162.5 12 39.0­78 39 19.5 97.5­136.5 13 39.0­78 39 19.5 97.5­136.5 14 19.5­39 39 19.5 78.0­97.5 15 19.5­39 39 19.5 78.0­97.5 16 19.5­39 26 19.5 65.0­84.5 17 19.5­39 26 19.5 65.0­84.5 --- --- --- --- --- --- --- --- --- --- 30 19.5­39 26 19.5 65.0­84.5 Discounted total number vaccinated (NUM*), in millions: 20-year horizon 106­212 227 67 400­506 30- year horizon 126­251 253 87 466­561 Table 12.3 Total Discounted Numbers of Disease Cases Prevented, Deaths Prevented with Treatment of all Cases, and Deaths Prevented without Treatment of any Cases, over 20-Year and 30-Year Horizons 20-YEAR HORIZON 30-YEAR HORIZON Cases prevented (PREC*): Meningitis 9,547­19,094 11,313­22,626 Typhoid fever 305,897 341,075 Pneumonia 16,835 21,700 Deaths prevented, with treatment of all cases (PREDT*): Meningitis 477-955 565­1,130 Typhoid fever 3,059 4,411 Pneumonia 1,684 2,170 Total 5,220­5,698 7,146­7,711 Deaths prevented, without treatment of any cases (PRED*): Meningitis 4,774­9,547 5,657­11,314 Typhoid fever 30,590 34,108 Pneumonia 5,051 6,511 Total 40,415­45,188 46,276­51,933 230 · Health Economics in Development effective. It should also be noted that the Table 12.2 data refer to individu- als rather than to vaccine doses, since a series of two or more doses per person may be needed to complete the vaccination and achieve 90% immu- nization. The cost-benefit analysis also assumes that the cost of vaccinating one individual will be independent of the number of people vaccinated. The lat- ter number will always be large enough (at least 19.5 million per year) to benefit from possible economies of scale. Similarly, it is assumed the bene- fit obtained per individual vaccinated will be constant and independent of how many others receive the vaccine. Among other things, this implies that the chance of one unvaccinated individual becoming ill does not depend on the number of individuals immunized; that is, a possible "collective immu- nity" effect is not taken into account (6). As a consequence of these assumptions, the total discounted costs and benefits can be found by totaling and discounting the number of people vac- cinated and later applying to that discounted sum the costs and benefits per person. More explicitly, if the cost of vaccinating someone against disease `i' is designated VAC(i), then by definition the total cost (CST) of vaccinating some number (NUM) of people against disease `i' in some future year t is as follows: CST (i, t) VAC(i) NUM(i, t), and discounting and totaling both sides of the equation yields CST*(i) SUM CST (i, t) (1 r)( t 1) SUM VAC(i) NUM (i, t) (1 r)( t 1) VAC(i) SUM NUM (i, t) (1 r)( t 1) VAC(i) NUM*(i) The same type of calculation can be applied to the number of disease cases prevented (applying as a multiplicative factor the probability that a vaccinated individual would have acquired the disease if he or she had not been vaccinated); to the number of deaths prevented (successively applying the probability that an individual with the disease died of it, whether or not the effects of curative treatment on the probability of survival are consi- dered); and to the total benefit obtained from vaccination (using the indivi- dual or unit benefit as the multiplicative factor). In relating all these calculations to a discounted sum of individuals, one is not saying that an individual vaccinated 15 years from now is worth less Cost-Benefit Analysis of a Regional Vaccination System · 231 than one vaccinated before that, but only that the economic value of the cost and the associated benefit are less today because they occur further in the future. This discounting and totaling of individuals instead of monetary sums is nothing more than a valuable mathematical simplification. Variables employed in making relevant calculations and the formulas used to discount and total them are summarized in Box 12.2, which also contains a glossary of all of the terms utilized in the analysis. The discounted and totaled numbers of individuals vaccinated, desig- nated NUM*, are shown at the bottom of Table 12.2 by disease for two dif- ferent horizons 20 years and 30 years. In contrast to the SIREVA costs for vaccine development (CS), which end in 10 years, the costs of administer- ing the vaccines (CST) never end, so long as the disease is controlled but the pathogens are not eradicated. Therefore, for the purpose of this anal- ysis, it is necessary to choose a final year. It seems reasonable to think that if SIREVA could be justified, this justification would probably occur within 20 years, a period that would include more than a decade of application of each vaccine. Beyond 20 years, any protection becomes very speculative; in particular, it is not known what might happen to the risks of acquiring a disease or the benefits of being protected. Solely to illustrate a longer horizon, the calcul- ations have been repeated for a 30-year period. As will be seen, this extension of the period does not significantly affect the system's net estimated worth. As the Table 12.2 projection shows, during the 20 years following SIREVA's initiation the equivalent, in terms of present value, of between 106 million and 212 million people would be vaccinated against pneumonia. Overall, in discounted terms, between 400 million and 506 million people would be vaccinated during the period, the actual figure depending on the extent of vaccination against meningitis. This total could refer to the dis- counted equivalent of 400­506 million individuals; or it could involve fewer individuals, some of them being vaccinated against two or even three of the target diseases. On extending the horizon to 30 years these values increase, but much less than proportionately to the number of additional years of vaccination because the discount factors (see Annex 1) give little weight to the years fur- thest away. In terms of present values, the entire third decade has the same value as only the last five years of the second decade, which in turn are only worth the same as the first year by itself. To go from the number of people vaccinated to the number of disease cases prevented it is necessary to multiply by the effectiveness of the vaccine 232 · Health Economics in Development Box 12.2 Glossary of Symbols, Variables, and their Relationships--in their Approximate Order of Appearance in the Text SYMBOL OR VARIABLE DEFINITION t 1,2,3 . . . Years since initiation of SIREVA i 1,2, or 3 Disease NUM Number of vaccinations administered; equal to the number of indi- viduals vaccinated if every individual is vaccinated against only one disease. NUM refers to the number of complete vaccinations, not to the number of doses, if vaccination requires the application of two or more doses. PREC Number of cases of a disease prevented by the vaccination program BEN Total benefit obtained by prevention of disease cases C Total cost CS Cost of SIREVA (for vaccine development) CST Cost other than for SIREVA (vaccine manufacture, distribution, and administration) Note: The variables C, CST, NUM, PREC, and BEN are classified by disease (i) and year (t). The variable CS is classified solely by year; CS(i) does not exist. By definition, C(t) CS(t) CST(t). SUM Indicates the summation of a variable over a series of years t (up to 20 or 30 years in the calculations) r Discount rate for future years (0.1 or 10% in the calculations) * Indicates the discounted sum of a variable; for example, C* SUM C(t) (1 r)( t 1) and BEN* SUM BEN(t) (1 r)( t 1) Note: The variables C, CS, CST, NUM, PREC, and BEN are all transformed into C*, CS*, . . . ., by discounted summation. For all except CS, the sum can be obtained for a single disease (i) or for all three diseases taken together. VAC(i) Unit cost of vaccinating one individual against one disease (i). Thus CST(i,t) VAC(i) NUM (i,t), and CST*(i) VAC(I) NUM*(i). VAC Summing for all three diseases gives the average implied maxi- mum cost of vaccination. It is calculated as follows: VAC (BENT* CS*)/NUM*. EFV(i) Effectiveness of the vaccine (i). (In the calculations it is always assumed that EFV equals 0.9 or 90%.) SUF(i) The probability of a person not vaccinated against target disease (i) acquiring that disease. Thus PREC(i,t) SUF(i) EFV(i) NUM(i,t), which gives PREC*(i) SUF(i) EFV(i) NUM*(i). MOR(i) The probability that an individual with disease (i) will die if not treated. MORT(i) The probability that an individual with disease (i) will die iftreated. PRED(i,t) The number of deaths prevented by vaccination, assuming those ill would receive no treatment. PRED(i,t) MOR(i) PREC(i,t), and thus PRED*(i) MOR(i) PREC*(i). The corresponding totals for all of the target diseases taken together are PRED(t) and PRED*. Cost-Benefit Analysis of a Regional Vaccination System · 233 PREDT(i,t) The number of deaths prevented by vaccination, assuming those ill would receive treatment. PREDT(i,t) MORT(i) PREC(i,t), and thus PREDT*(i) MORT(i) PREC*(i). The corresponding totals for all of the target diseases taken together are PREDT(t) and PREDT*. UTU(i) Unit benefit or utility of prevention--the benefit derived from pre- venting one case of disease (i). Thus the benefit of vaccinating one individual is UTU(i) SUF(i) EFV(i). BEN(i,t) The benefit derived from vaccinating NUM(I,t) individuals, so that BEN (i,t) UTU(i) SUF(i) EFV(i) NUM(i,t) UTU(i) PREC(i,t); and hence BEN*(i) UTU(i) PREC*(i). The corresponding totals for all of the target diseases taken together are BEN(t) and BEN*. Note: The net unit benefit (benefit minus cost) of vaccinating one individual is UTU(i) SUF(i) EFV(i) ­ VAC(i), and the net total benefit is [UTU(i) SUF(i) EFV(i) ­ VAC(i)] NUM(i,t) BEN(i,t) - CST(i,t). The same relationship is valid for NUM*(i), BEN*(i), and CST*(i). BTR(i) Unit cost of treatment--the cost of adequately treating one case of disease (i). BENT The benefit derived solely from not having to treat disease cases. Note that BENT BEN because the former does not include all of the benefits; hence BENT(i,t) BTR(i) PREC(i,t), and BENT*(i) BTR(i) PREC*(i). The corresponding totals for all of the target dis- eases taken together are BENT(t) and BENT*. D Delay (in years) between vaccination and the hypothetical onset of disease had the vaccinated individual not been vaccinated. The correct adjustment to the benefit derived from the prevention of one case can be calculated as follows: ADJD (l r)D (which determines whether the person is really immunized) and then by the chance that the person would become sick if unimmunized. The effective- ness of all three vaccines is estimated at 90%, while the incidences of the three diseases (per 100,000 population at risk) are estimated at 10 (with a maximum of up to 50) for meningitis, 150 for typhoid fever, and 28 for pneumonia. These figures give the likelihood of preventing a disease case by vaccinating one person a probability of 9, 135, and 25 chances per 100,000 respectively. (Only the lower estimated incidence is used for meningitis, because this reduces the benefits without affecting the costs; and if SIREVA is justified under these circumstances, it would be even more justified if the disease incidence were higher.) The probabilities of preventing a case are shown in Table 12.4, which will be discussed later, while Table 12.3 indicates the estimated numbers of cases that would be prevented, by disease, for horizons of 20 and 30 years. By far the greatest disease prevention occurs with regard to typhoid fever, 234 · Health Economics in Development Table 12.4 The Estimated Cost of Treatment (in Constant US$), Probability of Preventing One Case, and Implied Maximum Cost of Vaccination,a by Disease, Independent of the Number of Vaccinations MENINGITIS TYPHOID FEVER PNEUMONIA Probability of preventing one case, 9.0 10 5 1.35 10 3 2.5 10 4 SUF(i) EFV (i) Unit cost of treatment, BTR(i) $3,000 $584 $6,306 Implied maximum cost of one vaccination, $0.27 $0.79 $1.58 SUF(i) EFV(i) BTR(i) aThe implied maximum cost of vaccination is the value such that the cost of vaccinating one person compensates exactly for the probable cost of having to treat that individual for the disease. It is calculated by multiplying the unit cost of treatment by the probability of preventing one case. For this calculation the fixed cost of developing vaccines against the target diseases is not considered. because of its high incidence, the discounted number of cases to be prevented totaling over 300,000. For meningitis and pneumonia the esti- mated figures are lower by an order of magnitude, ranging from 10,000 to 23,000 for meningitis and from 17,000 to 22,000 for pneumonia. In some cases a person who acquires the disease will die. The likelihood of this varies greatly, depending on the disease and whether the victim does or does not receive adequate and timely treatment. With such treatment almost no one dies of typhoid fever, since the death rate is estimated at no more than 1%; and even without treatment that rate rises to only 10%. Regarding pneumonia, it is estimated that the lethality is 3% with treatment and 10% without treatment, while for meningitis the corresponding rates are estimated at 5% and 50% (4). Therefore, the numbers of deaths pre- vented are not proportional to the numbers of cases prevented--the risk of death depending on the particular disease involved and also varying by a factor as great as 10, depending on whether one assumes that each patient does or does not receive appropriate treatment. Table 12.3 shows the estimated numbers of deaths that vaccination would prevent, by type of disease, and also shows the total number of deaths preventable by SIREVA. Showing these latter totals is appropriate; for although it would be incorrect to total the numbers of cases of diseases that are very different with respect to severity and danger, it is legitimate to total the resulting deaths. Depending on the horizon selected, the totals range from 40,000 to 52,000 deaths prevented if no treatment is assumed, and from 5,000 to 8,000 if it is assumed that every victim receives appro- priate care. Cost-Benefit Analysis of a Regional Vaccination System · 235 Economic Benefit: Treatment Cost Saved The benefits obtained by preventing one case of a disease include some that are difficult to quantify or evaluate economically, such as reduction of the patient's pain and suffering. Other benefits, although possibly less important, are easier to evaluate in economic terms; among these is the treatment cost saved as a result of not having to care for the patient. Clearly, attributing this monetary benefit to vaccination depends on an assumption that the victim would receive the treatment if he were to contract the disease. This benefit is received by the person or institution that otherwise would have to pay the treatment cost, whether the paying party is the patient or not. Thus, one way to compare costs with benefits is to relate the cost of vac- cinating one person with the expected cost of treating that individual, con- sidering these procedures as alternatives. Obviously, this last assumption is more reasonable when the treatment results in a complete cure, without permanent injury to the patient. When the patient dies despite the treat- ment (which is possible with all diseases and occurs in up to 10% of pneu- monia cases) or is left with significant sequelae (as can easily occur with meningitis), treatment is a very incomplete substitute for prevention. Of course, one must compare the cost of vaccination with the expected or probable cost of care, because not all vaccinated individuals would become sick if unvaccinated. The comparison depends, therefore, on the likelihood that the vaccine would prevent a case of the disease. This likeli- hood, as already noted, is shown on the first line of Table 12.4; it is calcu- lated by multiplying the effectiveness of the vaccine (EFV) against disease (i) by the probability of suffering the disease, SUF(i). This is the same logic that has been used to justify eradication of poliomyelitis, through the sav- ings in treatment costs that would result from vaccinating virtually the entire population at risk (7). The second line in Table 12.4 shows the average cost of providing a patient with correct and timely care, designated BTR(i). This is estimated to range from less than $600 in the case of typhoid fever to more than $6,000 in the case of pneumonia. Taken together, the likelihood of preventing a case and the cost of treating that case determine a hypotheti- cal cost of prevention where the prevention and treatment costs would equal one another, and so the net saving from vaccination would be zero. This cost can be viewed as the implied maximum cost of vaccination in the sense that at any lower cost the vaccination would be less costly than the treatment. 236 · Health Economics in Development As can be seen on the third line of the table, this latter cost, SUF(i) EFV(i) BTR(i), is calculated by multiplying the probability of preventing one case by the cost of treating one case. And the cost of preventing one case is the cost of one vaccination, VAC(i), divided by the probability of pre- venting one case with one vaccination or VAC(i) [SUF(i) EFV(i)]. On relating this expression to the cost of treatment, BTR(i), one sees that where VAC(i) SUF(i) EFV(i) BTR(i) there is exact equality between vaccination and treatment costs; and like- wise, where VAC(i) SUF(i) EFV(i) BTR(i) vaccination offers a net economic benefit. In monetary terms (see Table 12.4), the corresponding values range from $0.27 per vaccination in the case of meningitis to $1.58 in the case of pneumonia. Regarding typhoid fever, it should be noted that the low cost of treating one case of this disease is par- tially balanced by its high incidence in the population, so that the implied maximum cost of vaccination against typhoid fever is $0.79, or almost three times the implied maximum cost in the case of meningitis. As has been said, until the vaccines are developed and administered on a mass scale, there can be no exact picture of vaccination cost. The interpre- tation of the calculations in Table 12.4 is that vaccination will be justified-- through savings in treatment costs, without considering other benefits--as long as it costs no more than $0.27 per vaccination against meningitis, etc. If, for example, it were feasible to vaccinate at a unit cost of $0.10--which would cover not only the cost of the vaccine but also the cost of distributing and administering it to the population, then clearly vaccination would be highly worthwhile. In contrast, if the unit cost were $1.00, only vaccination against pneumonia would appear to be justified, assuming no other benefits were considered. Even if one compares only vaccination and treatment, however, the cal- culation presented in Table 12.4 is still incomplete because the vaccines involved do not yet exist and have to be developed. This implies that the benefits, in the form of saved medical costs, would have to cover not only the costs of vaccination (costs "outside SIREVA" or CST), but also the vac- cine development costs "within SIREVA" (CS). In addition, SIREVA is Cost-Benefit Analysis of a Regional Vaccination System · 237 attempting to develop three vaccines, without the total cost of the program being attributed to one or another of these products. The first of these conditions implies that the maximum cost allowed for vaccination is going to be less than that shown in Table 12.4, since the ben- efits must also cover SIREVA's vaccine development costs. The second con- dition (of co-production or inseparability) implies that judging the worth of each vaccine individually makes no sense, because it will be necessary to judge the entire system with respect to the average cost of vaccination against the three target diseases. This matter can be summarized as follows: For there to be a net benefit after considering both types of costs, it is necessary that BENT* CS* CST*, where BENT is the total benefit in saved treatment costs--the cost of treat- ing one individual times the number of cases prevented. (Table 12.3 indi- cates the number of cases prevented, and Table 12.4 shows the unit costs of treatment.) As noted above, CST* VAC NUM*, where VAC denotes the average cost of vaccination within the time interval involved. Both BENT* and CST* must be assumed for all the target dis- eases, and both must also be discounted and summed over time. Then, in order for the net benefit condition to be met, it is necessary that BENT* CS* VAC NUM*. Table 12.5 presents the corresponding calculations. Starting with the values of CS* and NUM* from Tables 12.1 and 12.2, respectively, Table 12.5 pro- ceeds to list values for the three ingredients of BENT*, these being the saved costs of treating each disease, and the BENT* totals for 20-year and 30-year horizons. The last entries show the 20-year and 30-year values of BENT* CS* and of VAC. The figures shown indicate that over a period of 20 years some US $80.3 million, at present value, would be spent developing the three vaccines, which would be administered to the discounted equivalent of 400­506 mil- lion individuals. The disease cases prevented would represent an estimated saving of $29-$57 million for meningitis, $179 million for typhoid fever, and $106 million for pneumonia. The total benefit would amount to $313­$342 million before subtracting the costs of SIREVA itself, yielding a net benefit of $233­$262 million. After dividing this amount by the total 238 · Health Economics in Development Table 12.5 The Average Implied Maximum Cost of Vaccination (in Constant US$) Derived from the Number of Individuals Vaccinated, Treatment Costs, and the Cost of SIREVA, for 20-Year and 30-Year Horizons 20-YEAR HORIZON 30-YEAR HORIZON Costs of SIREVA (CS*), in US$ millions 80.3 80.3 Total number of vaccinations (NUM*), in millions 400­506 465­591 Saving on treatment costs (BENT*), by disease, in US$ millions: Meningitis 29­57 35­68 Typhoid Fever 179 199 Pneumonia 106 137 Total, three diseases 313­342 370­404 Net saving, after deducting SIREVA's cost (BENT* CS*), in US$ millions 233­262 290­324 Average implied maximum cost of one vaccination­VAC (net savings divided by the number of vaccinations), in US$ 0.52­0.58 0.55­0.62 number to be vaccinated, it can be concluded that SIREVA is justified with respect to the treatment costs saved as long as the population could be vac- cinated for no more than $0.52­$0.58 each. If one individual were to receive all three vaccines, the permitted cost would rise to $1.56­$1.74. Applying these same calculations to the 30-year horizon yields an average permitted cost that is greater by a few cents because the costs of SIREVA would be distributed over more years of vaccination, and so their relative weight in the total costs would be less. Overall Benefits from SIREVA and Vaccination Costs The exercise in the previous section fixes a value on the benefit derived from SIREVA, equating it to saved medical treatment costs, and on this basis estimates the maximum cost of vaccination that would be compatible with a net positive benefit. This procedure can be reversed by first fixing a value on the cost of vaccination and then deriving from it an estimate of the minimum benefits could be of any type, without being limited to the treatment costs saved. The worth of preventing a death, the value of eco- nomic production saved by preventing death or illness, the reduction of physical and emotional suffering, and other benefits could be included. In this regard, since the disease cases and deaths prevented are the most Cost-Benefit Analysis of a Regional Vaccination System · 239 quantifiable results of vaccinating the population, it seems natural to esti- mate minimum benefits in terms of these concepts. The condition that must be satisfied is BEN* CS* CST* C*, where C* is the total costs and BEN* is the discounted sum of all the ben- efits (including BENT*, the benefit of not having to treat those who would become ill). If we then designate the benefit or "utility" per case prevented as UTU, and the number of cases prevented as PREC*, we see that for case prevention alone to satisfy the condition it will be necessary for UTU PREC* C*, or equivalently, UTU C* PREC*. Similarly, if we designate the benefit per death prevented as UTUD, we see that for mortality prevention alone to satisfy the condition it will be nec- essary for UTUD C* PRED*, where PRED* is the number of deaths that would occur in the absence of vaccination and treatment. The corresponding calculations appear in Table 12.6. They are limited to the 20-year horizon, since it was determined (in Table 12.5) that exten- sion of the horizon to 30 years does not significantly affect the results. Start- ing with the cost of SIREVA (CS*), numbers of cases prevented (PREC*), and numbers of deaths prevented (PRED*) that appear in Tables 12.1 and 12.3, Table 12.6 derives the other component of the total cost--the cost of vaccination--from the total number vaccinated (see Table 12.2), using a unit cost (VAC) first of $1 and then of $10. The resulting total cost (C*) is then used to calculate the benefits per case prevented (UTU). It turns out that the first VAC cost ($1) yields values quite close to those that would permit justification of SIREVA solely on the basis of medical treatment costs saved, while the second VAC cost ($10) yields values so high that the benefits per case prevented would have to be substantially greater. The results of attributing the entire benefit to the prevention of death (UTUD) are shown in the last line of the table. Naturally, the benefit 240 · Health Economics in Development Table 12.6 Implied Minimum Benefit per Case Prevented and per Death Prevented as a Function of the Cost of Vaccination, without Considering Patient Treatment (20-Year Horizon) COST OF VACCINATION (VAC) US $1.00 US$10.00 Total number of cases prevented (PREC*), in thousands 332­342 332­342 Total number of deaths prevented (PRED*), in thousands 40­45 40­45 Cost of SIREVA (CS*), in US$ millions 80.3 80.3 Cost of vaccination (CST*), in US$ millions 400­506 4,000­5,060 Total Cost (C*), in US$ millions 480­586 4,080­5,140 Minimum benefit per case prevented (C*/PREC*), in US$ thousandsa 1.4­1.7 12.3­15.0 Minimum benefit per death prevented (C*/PRED*), in US$ thousandsb 12.0­13.0 102.0­114.2 aThe values for the minimum benefits do not change significantly on extending the horizon to 30 years. bThe minimum benefit per death prevented does not attribute any benefit to preventing nonfatal disease cases. involved would have to be much greater. Given that, on the average, approximately 10% of the untreated cases would terminate in death, the minimum benefit per death prevented would have to be some 10 times greater than the minimum benefit per disease case prevented. For example, at a cost of vaccinating one person for $1.00, SIREVA is justified so long as an average benefit per case prevented of between $1,400 and $1,700 is obtained. This is based on the estimate that a total of $480 to $586 million at present value will be spent in order to prevent a total of some 332,000­342,000 cases of the three diseases. The minimum necessary benefit per case rises to $12,000 if the cost of vaccination is fixed at $10; it does not rise in the same proportion as the unit cost of vaccination because the actual expenditures of SIREVA are not affected. It should be noted, however, that when the cost per vaccination is $1.00 or greater, these fixed costs of developing the vaccines are of relatively little importance compared to what would have to be spent applying them. Therefore, justifying the vaccination is almost equivalent to justifying SIREVA, if there is no other way to develop the vaccines that is less costly than the system proposed. The final calculations (on the last line of Table 12.6) are somewhat arti- ficial, since they attribute benefits only to the prevention of death. This Cost-Benefit Analysis of a Regional Vaccination System · 241 establishes a kind of "maximum of the minimum" for the necessary benefit justifying SIREVA--at levels on the order of $12,000 in the first instance and $100,000 in the second. It should be noted, however, that benefits from cases prevented and deaths prevented can be combined. That is, it is appropriate to compensate for the costs of SIREVA through any combination of benefits per death prevented and benefits per non-mortal case prevented that satisfies the relationship [UTU (PREC* PRED*)] (UTUD PRED*) C*, where UTU, the benefit per case prevented, would be substantially less than UTUD, the benefit per death prevented. The expression (PREC* PRED*) refers to the number of individuals who would become sick but not die if they were not vaccinated. Both the calculations in Table 12.5 and those in Table 12.6 implicitly assume that the benefit associated with vaccination occurs immediately, simultaneously with vaccination. This assumption is justified if the target disease would probably attack an individual within a short time or never, as is typically true of the diseases targeted by the Expanded Program on Immunization, which affect primarily children (although those diseases can appear several years later than the normal age of immunization). If, on the other hand, a large proportion of those affected will typically become ill many years after vaccination--and if the vaccine retains its effectiveness for many years, so that it is not necessary to repeat the vaccination frequently-- the calculations that were just presented can prove optimistic or overly favorable because they do not consider the interval between the moment of vaccination and the probable moment of becoming ill. The greater this interval, the longer the benefits are delayed relative to the costs, and the greater they have to be to compensate for this delay. The way to adjust for the possible optimistic bias is to estimate the average inter- val between vaccination and illness in years (D) and then to discount the benefits with respect to the costs by the factor (l r) D , utilizing the same discount rate(r) applied elsewhere. By way of example, Table 12.7 shows the sizes of adjustments associated with several different intervals of delay. Thus, if the benefits were delayed an average of 10 years, they would have only 38.55% of the value they would have if they appeared immediately. The rest of the table shows the impact of these adjustments on parameters calculated in Table 12.5 (the implied maximum cost of vaccination) and Table 12.6 (the implied 242 · Health Economics in Development Table 12.7 The Effects of Adjusting for the Delay between Vaccination and Disease Onset upon the Implied Maximum Cost of Vaccination and upon the Implied Minimum Benefits of Preventing a Disease Case, in Constant US$ DELAY (D), IN YEARS 0 5 7 10 15 Adjustment factor (AJUD) 1.000 0.7513 0.5132 0.3855 0.2394 Effects on the implied maximum cost of vaccination (original value multiplied by the adjustment factor), in US$: By disease, without counting the cost of SIREVA: Meningitis 0.27 0.20 0.14 0.10 0.06 Typhoid fever 0.79 0.59 0.41 0.30 0.19 Pneumonia 1.58 1.19 0.81 0.61 0.38 Average for SIREVA (all costs for three diseases): Minimum 0.52 0.39 0.27 0.20 0.12 Maximum 0.58 0.44 0.30 0.22 0.14 Effects on the implied minimum benefit per case or death prevented (original value divid- ed by the adjustment factor) in US$ thousands: Per case prevented; vaccination cost US$1.00: Minimum 1.4 1.9 2.7 3.6 5.8 Maximum 1.7 2.3 3.3 4.4 7.1 Per death prevented; vaccination cost US$1.00 Minimum 12.0 16.0 23.4 31.1 50.1 Maximum 13.0 17.3 25.8 33.7 54.3 minimum benefit per case prevented). As can be seen, a relatively short delay such as five years does not greatly affect the results; but longer delays such as 15 years produce much stronger effects--resulting in multiplication or division of the benefits or costs by a factor of four or more. Conditions Justifying SIREVA and Sensitivity of the Results It has not been possible to carry out a closed and precise cost-benefit analy- sis for the proposed system at this time because its exact costs are not known and there is no consensus on how to evaluate its benefits. Therefore, the Cost-Benefit Analysis of a Regional Vaccination System · 243 analysis presented in the above sections is based on the relationships between these unknown elements, rather than upon definitive values assigned to them. For every level of benefit per disease case prevented, there is a corresponding maximum value for vaccination cost that still leaves a positive net benefit. And conversely, each unit cost of vaccination estab- lishes a minimum for the total benefit of preventing one case (or one death) compatible with net benefit from the system. The corresponding calcula- tion of these two ways of presenting the relationship, shown in Tables 12.5 and 12.6, can be considered the essence of the present analysis. In general terms, the calculations allow one to conclude that SIREVA would be justified by its benefits if it were possible to develop the vaccines at the costs estimated for the different options and later to administer them to the population at a unit cost of half a dollar or less. At this level of expense, the system could generate sufficient treatment cost savings to com- pensate for the entire cost of developing and administering the vaccines. Even if it were assumed that in the absence of SIREVA not all the disease victims would receive adequate and timely treatment, the system would still be justified if benefits per disease case prevented were found to have a min- imum average value between $1,000 and $2,000. Part of these benefits would derive from prevention of deaths; and if it were estimated that it would be worth spending somewhat more than $10,000 on the average to avoid one death, this benefit alone would justify the proposed expenditures. How sensitive are these results to variations in the different parameters considered in the analysis? If a small change in one of them causes the sys- tem to stop appearing viable, then the proposal would be risky, given the great uncertainty in the estimated values. The analysis has taken into account all of the following factors: the cost of SIREVA itself (development of the vaccines), the cost of vaccinating one individual against one disease, the cost of treating one case of a disease, the number of individuals vacci- nated, the effectiveness of the vaccine, the incidence of the target diseases, their lethality with and without treatment, the discount rate, and the possi- ble delay between a person's age at vaccination and age at disease onset. For some of these factors, where less is known or it is possible to anticipate a large variation, an explicit sensitivity analysis has been made. For other ele- ments the probable variation in the factor and the consequences for the results have been discussed briefly. To terminate this analysis, the sensitiv- ity of the conclusions to the elements mentioned are discussed below. In general, there is no reason for hesitation in exploring the possibilities of changes on the order of 10% or 20%; the concern is whether one ought to 244 · Health Economics in Development anticipate variations of an order of magnitude or so in the system's esti- mated yield. Actual cost of SIREVA. The importance of these costs depends on whether they are large or small relative to the total vaccination cost. If they are small, they can vary considerably without greatly affecting the total cost. For example, at the maximum cost calculated for which vaccination is justified in terms of medical costs saved, the costs of SIREVA itself are one-third or less of the total costs, so that they could be underestimated by 50% and still not have a great effect upon the system's yield. Aside from an increase in SIREVA's cost, the relative importance of this element would be greater if the unit cost of vaccination were less than esti- mated. In that case, however, the reduced cost of administering the vaccines would compensate for a large increase in the cost of developing them. For example, consider the calculation in Table 12.5 and assume that the element CS* (the cost of SIREVA) were doubled. Then CS* would be $160.6 mil- lion, but the system would still be justified for any vaccination cost VAC less than $0.36. Vaccination cost. As has been seen, this element is crucial; and if one calcu- lates benefits only in terms of medical expense saved, this imposes a clear maximum value upon vaccination cost that is at the level of $0.50. Increas- ing the unit vaccination cost to $1.00 requires greater total benefits; and if the vaccination cost were as high as $10.00, the saving in treatment cost by itself would be far too small to justify the system. Hence, everything depends on whether vaccination is achieved at a reasonable cost, and the proposal assumes that result. To achieve such a result, it may be necessary to incorporate the new vaccines into the EPI; that way the logistic costs would be minimal, and little more would have to be spent beyond that needed to cover the costs of manufacturing the vaccines. Treatment cost. It is assumed that this element is relatively well known, so that its possible variations need not be taken into account. In any case, if the scheme of analysis utilized in Table 12.6 is adopted to consider the total value of the benefits per case prevented, this treatment cost variable becomes less important--because it then constitutes only one component of the benefits, and perhaps not the greatest of them. Number of individuals vaccinated. This factor is crucial for the simple reason that the costs of developing the vaccines must be offset by administering them to a large enough number of people. If it were not for this fixed Cost-Benefit Analysis of a Regional Vaccination System · 245 development cost, the calculations in Table 12.4 could be applied directly; SIREVA's justification would be independent of the scale of operation; and the average cost of vaccination could be as high as $0.78. Comparing these calculations with the values listed in Table 12.5 shows how the need to com- pensate for the system's fixed costs affects the results. Both the maximum cost of vaccination and the implied minimum benefit vary directly with changes in the number of individuals covered by SIREVA. It is assumed, however, that the estimates of this latter number would not be in error by more than a small percentage. Vaccine effectiveness. This factor cannot vary much because a vaccine would not be administered if it were not at least 70% or 80% effective. Therefore, vaccine effectiveness cannot affect the results very much. It would only be important if after expending millions of dollars on SIREVA, the effort failed and effective vaccines were not obtained; the entire proposal is based upon confidence that this will not occur. Disease incidences. The estimates of these parameters are very low, the maxi- mum value used being 150 cases per 100,000 inhabitants for typhoid fever. Any increase would only make the system more viable; and so the only con- sideration should be whether the incidences of the target diseases have been overestimated. Changes in the probability of getting sick affect the benefits the same way that changes in the number of people vaccinated do, but with- out affecting the cost--unless one could, at lower risk, vaccinate fewer peo- ple. The epidemiological studies constituting part of the system's develop- ment will help to define these risks better and so to adjust, if necessary, the projected extent of mass vaccination. Disease lethality. This factor cannot vary much, even admitting that it is not known exactly. In any case, it is important only if one desires to attrib- ute a specific benefit to the prevention of death, for there would clearly be great benefit in preventing each of the target diseases even if no one died of them. Discount rate. As has already been discussed, this element cannot vary by more than a factor of two, and its influence affects the distribution of costs and benefits over time without affecting their comparison in a given year. Therefore, the results of the analysis are not considered very sensitive to the rate selected. Delay between vaccination and prevented illness. As Table 12.7 shows, this fac- tor becomes a matter of concern if it is necessary to assume a delay of more 246 · Health Economics in Development than about half a decade. If the disease presents risks over the entire human life-span, part of the benefit is left unperceived in terms of present value. Even though the probable impact of such a delay would only divide the ben- efits in half, this circumstance would require an average benefit twice as large, or a cost of vaccination half as large, as those projected. It is clear that the justification, or lack thereof, of a project such as the one being analyzed depends upon how all of these elements are evaluated, and upon the values assigned to prevention of disease and death--values outside the purely economic realm. The present analysis only attempts to trace a dividing line between the possible combinations of factors, known or estimated, that show whether or not SIREVA would be viable in the sense of producing benefits that more than compensate for its costs of develop- ment and application, within a reasonable span of time. References 1. Institute of Medicine, National Academy of Sciences. New vaccine develop- ment: establishing priorities (vols 1 and 2). Washington, DC: National Academy Press; 1986. 2. World Health Organization. Research and development in the field of vac- cines; progress report by the Director General. Executive Board, 87th Session, Geneva: 21 November 1990. (Document EB87/6). 3. World Health Organization, UNICEF, and UNDP. Declaration of New York: the children's vaccine initiative. New York: 10 September 1990. 4. Organización Panamericana de la Salud. SIREVA: estudio de factibilidad, sis- tema regional de vacunas para los países de América Latina y el Caribe. Unpublished document. Washington, DC: 1990. 5. Drummond MF, Stoddard GL, Torrance GW. Methods for the economic eval- uation of health care programs. Oxford: Oxford University Press; 1986. 6. May RM. Ecology and population biology. In: Warren KS, Mahmoud A, eds. Tropical and geographic medicine, 2nd ed. New York: McGraw-Hill; 1990. (Chapter 19). 7. Musgrove P. Is polio eradication in the Americas economically justified? Bull Pan Am Health Org. 1988; 22(1): 1­16. CHAPTER 13 Cost-Effective Malaria Control in Brazil Introduction Although the Malaria Eradication Program of the Ministry of Health in Brazil had succeeded by the late 1970s in freeing the majority of the coun- try from malaria transmission, it was unable to contain the rapid spread of the disease in the Amazon Basin. By June 1984, that region, including nine of the country's 26 states, accounted for 97% of all reported malaria cases, with high fatality rates. Between 1977 and 1988 the coefficient of mortality (deaths per 100,000 population) in the Amazon more than quadrupled. The enormous extent of the region, the substantial and hard-to-trace migration into and within it, and the existence of numerous transient and dispersed set- tlements, rendered ineffective the traditional eradication strategy based on active case detection for treatment and eliminating the vector through wide- spread use of insecticides. At the end of 1983, there were 280,000 reported cases of the disease, but with the number of people infected rising by 40,000 every year, incidence reached almost half a million four years later. In 1986 the Government of Brazil requested World Bank technical and financial assistance to develop an Amazon Basin Malaria Control Project (known as PCMAM from its Portuguese title) to support the national pro- gram. The project was expected initially to be conducted over four years and cost US $200 million; it became effective in September 1989 and closed in June 1996, with a final cost of US $133.7 million, of which US $72.9 mil- lion was financed by a Bank loan (The World Bank, 1996). The project was originally aimed at getting the malaria outbreak in the Amazon Basin under Co-authored with Dariush Akhavan, Alexandre Abrantes and Renato d'A. Gusmão. Reprinted from Social Science and Medicine, Vol. 49, No. 3, Copyright 1999, with per- mission from Elsevier Science. 247 248 · Health Economics in Development control, preventing the spread into uninfected areas and strengthening institutional capacity. Like the program which it supported, the project con- sisted of vector control (application of insecticides in dwellings and fogging of high-risk communities), entomological surveillance, treatment, special efforts for disease control in indigenous areas and information, education and communication (IEC). This paper provides an evaluation of the project during the interval of almost seven years corresponding to the World Bank loan. (There is no dis- tinction between the Bank project and the pre-existing government pro- gram, so this is not an evaluation of the marginal contribution of the Bank- financed project.) A first estimate of the results of the project is provided by an extensive evaluation by one of the authors (Akhavan, 1996); some of these initial findings have been published in a World Bank evaluation (The World Bank, 1996) as well as in a government summary publicizing the results of this and a parallel project for the control of three other endemic diseases in northeastern Brazil (National Health Foundation, 1996). A more detailed state-by-state analysis was conducted later (Akhavan, 1997). The present study provides projections of three key variables in the absence of the project: the incidence, severity (proportion of falciparum) and lethality (case fatality rate) from malaria in the Amazon Basin during the period 1989­1996, and summarizes the estimated savings in lives, morbidity (cases) and disability-adjusted life years (DALYs) from malaria control. Old and New Strategies for Fighting Malaria Prior to 1991, malaria control was the responsibility of a semi-autonomous federal agency, the Superintendency for Public Health Campaigns (Super- intendência para Campanhas de Saúde Pública, SUCAM), which carried out nation-wide malaria and endemic disease control campaigns through a workforce of 40,000. The agency was noted for its strong staff and line organization and had an excellent record in sustaining endemic disease con- trol programs in remote areas, under very difficult conditions. Municipali- ties and States had no responsibility for endemic disease control. Hospitals and outpatient clinics affiliated with the National Health System (known from 1990 on as the Sistema Único de Saúde, SUS) did not treat malaria patients, who were routinely referred to SUCAM facilities or staff. In 1991, SUCAM was extinguished and its functions transferred to a new agency, the National Health Foundation (Fundação Nacional de Saúde, Cost-Effective Malaria Control in Brazil · 249 NHF), which went through a period of political and organizational turbu- lence, largely as the result of rapid decentralization (Brazilian Institute of Municipal Administration 1996a, b). This created something of a vacuum in the field, seriously undermined staff morale and jeopardized the opera- tional capacity that had characterized SUCAM. These difficulties added to the natural complexity of controlling malaria by the traditional strategy of vector control everywhere that ecological conditions favored transmission and were only overcome starting in 1993. Mortality from malaria had begun to fall in 1989, the year the project went into effect, and over the next three years fell to only half the peak level registered in 1988. Nonetheless, by 1992 it became apparent that this approach was inadequate, because the number of cases continued to grow. Beginning in late 1992, therefore, the Brazilian control program was reori- ented in line with the new Global Malaria Control Strategy (World Health Organization, 1993a) emphasizing disease management, which replaced an earlier effort (World Health Organization, 1978) to define control strate- gies following the failure of eradication efforts. The Pan American Health Organization (PAHO) collaborated with the NHF in this first large-scale implementation of the new approach (World Health Organization, 1993b). A key element of the change was to stratify the population by risk and con- centrate control activities accordingly (Pan American Health Organization, 1991). The project was refocused, with greater emphasis on early diagnosis and immediate and intensive treatment of patients, while the use of pesti- cides was more closely targeted to municipalities and communities with high malaria incidence. The first of these changes aimed to prevent human deaths rather than kill mosquitoes, while the second emphasized killing those mosquitoes most likely to carry malaria, particularly the form, P. falciparum, which causes nearly all deaths (Miller and Warrell, 1990). This targeting drew on epi- demiological studies in the 1980s (Sawyer and Sawyer, 1987; Cruz Marques, 1988), which had shown that malaria was concentrated in relatively few municipalities, often characterized by new and inaccessible agricultural set- tlements or wildcat gold mining areas (garimpos), where insecticide spraying or fogging is typically ineffective (Najera et al., 1993, Table 13.2). In 1985, two states, Rondônia and Pará, accounted for over 73% of all cases in the region and were the source of infection of most cases identified in the rest of the country. In 1986, only 22 out of the 458 municipalities of the Ama- zon Basin accounted for 60% of reported cases of malaria. However, resources were not being allocated according to incidence: some 70% of the 250 · Health Economics in Development government program's resources were being used in areas with only 3% of cases. It is much easier to misallocate preventive efforts than treatment; in consequence, the cost-effectiveness of prevention is likely to vary much more, and may reflect more waste, than that of case treatment. This pattern shows up when cost-effectiveness is estimated separately for each of the nine Amazon Basin states (Akhavan, 1997). The project developed new diagnosis and treatment protocols and car- ried out an extensive training program for health care professionals in SUS hospitals and ambulatory clinics, which also received supplies of antimalar- ials. As a result, parasitoscopic diagnosis efficiency improved by 20%; the number of hospital admissions for malaria, which had increased from about 10,000 in 1984 to 20,000 in 1988, rose to more than 50,000 in 1992 and 1994 before declining slightly; and the better care led to a 55% fall in the estimated overall case fatality rate. One paradoxical effect of the changed strategy was initially to increase the apparent number of municipalities in every category of risk of infection, as Table 13.1 shows. In 1992, the Annual Parasitological Index (API), an indicator of the probability of contracting malaria, was available for only 80% of the 654 municipalities considered to have the ecological potential for transmission. Improvements in the NHF information system brought coverage up to 98% by 1993. To improve reporting and allow for greater focus on the highest-risk areas, the number of posts equipped for micro- scopic diagnosis of malaria was expanded from 405 in 1992 to 1,095 at the end of the project in 1996. At the same time that more municipalities appeared to be at high risk, the concentration of disease control efforts greatly reduced the number of Table 13.1 Population (Millions) and Number of Municipalities by Risk of Malaria Transmission, Based on API, 1988­1995 RISK CATEGORY HIGH RISK (API 50) MODERATE RISK (50 API 10) LOW RISK (10 API 1) YEAR POPULATION MUNICIPALITIES POPULATION MUNICIPALITIES POPULATION MUNICIPALITIES 1988 21.78 26.41 16.82 1989 22.79 25.19 17.26 1990 23.85 25.79 17.70 1991 21.66 24.75 16.56 1992 22.09 63 25.24 103 16.88 188 1993 5.40 86 13.08 127 56.83 192 1994 5.86 105 4.41 113 10.33 174 1995 3.15 126 6.06 111 4.91 159 Cost-Effective Malaria Control in Brazil · 251 people at moderate or high risk (API 10.0 positive blood slides per 1,000 population), by reducing the risk in the more populous municipalities. Focusing surveillance in those areas meant that between 1992 and 1995 coverage by prompt diagnosis expanded from 11 to 34% of the population at greatest risk, among whom the share of positive blood slides went from 24 to 42%. Increases in coverage and in positivity rates as a result of this concentration of effort were also observed for the populations at moderate, low and no risk of transmission. Collection of blood slides is more cost- effective when rates are high than when they are low and collection is main- tained only to monitor eradication efforts (Najera et al., 1993). The change in treatment strategy meant giving each suspected malaria case presenting to any level of the program (any facility, from a health post to a hospital) a complete chloroquine treatment (25 mg/kg of body weight). This is still regarded as efficacious for P. vivax infections and partly effec- tive in reducing the clinical symptoms of P. falciparum disease. Mefloquine, a synthetic antimalarial which allows effective early treatment of P. falci- parum even in ambulatory settings, was also licensed and distributed widely in SUS. As a result of earlier and more aggressive treatment, 20% fewer fal- ciparum cases required hospitalization in 1995 than in 1992. Mixed infec- tions, patients who had both falciparum and vivax, were classified and treated as falciparum because of its greater severity. Directly Observed Results: Malaria Cases, Severity and Program Expenditures Figure 13.1 shows how the malaria epidemic began to come under control. The upper panel relates the total number of blood slides positive for malaria to the total expenditure on malaria control by SUCAM and subsequently by NHF; the lower panel shows the same information, but for P. falciparum only. These are all directly observed variables, involving no estimations or assumptions. The numbers in Figure 13.1 differ slightly from those used to estimate the total health benefits and the total costs of the program, which do involve assumptions. First, they refer to the entire country rather than just the Amazon Basin, but as indicated earlier, nearly all cases occur in that area, as do more than 90% of expenditures, so the difference is small. Sec- ond, the number of positive slides differs slightly from that of new cases, because two or more slides may be taken to confirm a diagnosis; the differ- ence is small, probably of the order of 3%. Third, salaries of staff are not 252 · Health Economics in Development Figure 13.1 Cases of Malaria and Total Budget for Malaria, 1975­96 Total Budget in US$1,000 $100,000 1985 $90,000 $80,000 1987 1992 $70,000 1986 1983 $60,000 1996 1995 $50,000 1993 1991 1978 1982 $40,000 1975 1979 $30,000 1984 1994 19761977 1988 1990 1980 $20,000 1981 $10,000 1989 $0 100,000 200,000 300,000 400,000 500,000 600,000 700,000 Cases of Malaria Total Budget in US$1,000 $100,000 1985 $90,000 $80,000 1992 1987 $70,000 1996 1995 $60,000 1986 1983 1993 $50,000 1991 1978 1982 $40,000 1979 1984 $30,000 1975 1980 1976 1977 1994 1990 1988 $20,000 1981 $10,000 1989 $0 50,000 100,000 150,000 200,000 250,000 300,000 Cases of Falciparum Cost-Effective Malaria Control in Brazil · 253 attributed to the malaria control program, so expenditures in Figure 13.1 reflect only other recurrent spending. However, salaries are small compared to those other costs, which do include the transportation and per diem expenses of field workers. Finally, expenditures are in current US dollars, without adjustment for US inflation or discounting to present value as is done when adding up total project costs. For year-by-year comparisons the second adjustment is irrelevant and that for price changes in the US is small. What Figure 13.1 principally shows is that the epidemic expanded in every year from 1975 to 1989, irrespective of the level of SUCAM's expen- ditures on control and also of whether spending rose or fell from one year to another. Large annual increases in expenditure do seem to have slowed the expansion, but never enough to reverse the epidemic; in any case, the association is not robust. In 1987­1989, the program all but collapsed, with expenditure falling from US $78 million to just over US $10 million. With the implementation of PCMAM and the infusion of World Bank funds, the budget quickly returned to almost the 1987 level and the number of cases declined slightly in 1990 and then increased and stabilized in 1991­1992. There was a sharp reduction in cases in 1993, despite a decreased budget, and after a rebound in 1994, stabilization and further decline in 1995­1996, when the number of positive slides fell back to the level of a decade earlier at the same or lower expenditure. The pattern is very similar for cases of P. falciparum alone, through 1988; from 1989, the falciparum epidemic was better contained than that of malaria in general, with little or no rebound in 1991­1992 and then a return to the level of 1983. This appears to reflect the concentration of vector control in areas of high falciparum incidence and the consequent reduction of severity since the project adopted the revised control strategy. Methods: Estimating Illness, Lives and Disability-Adjusted Life Years Saved The malaria control program produced health gains partly by preventing cases, some of which would have ended in death while the rest produced only short-term morbidity, and partly by treating cases, particularly by pre- venting deaths from P. falciparum infection. The estimate of health benefits from vector control begins with the projected incidence of cases of malaria and proceeds through the expected severity (share of falciparum in total cases) and lethality (case fatality rate) to derive the losses in deaths and mor- bidity that would have occurred in the absence of the program. 254 · Health Economics in Development To facilitate comparison with other diseases which also cause both death and disability, the savings in deaths and episodes of illness are then con- verted to estimates of DALYs saved, using the same disability weights for non-fatal conditions, the annual discount rate and the age weights as in the WHO-World Bank estimates of the global burden of disease (World Bank, 1993; Murray, 1994, 1996). On a scale in which perfect health is 1.0 and death is zero, the health loss due to vivax illness is assigned a weight of 0.22 and the disability from non-fatal falciparum is given a weight of 0.375, reflecting the greater severity of falciparum morbidity. The future is dis- counted at a constant annual rate of 3%, which matters greatly for death (the number of discounted years lost to premature mortality is much less than the number of calendar years or potential years of life lost) but has no effect on short-term morbidity. Finally, the value of life at each age, or the loss from illness or death at that age, follows an age-weighting function which starts at zero, rises rapidly to a maximum of 1.525 at age 25, and then declines slowly and almost exponentially toward zero. An extensive expla- nation of these subject parameter values (Murray, 1996) and a sensitivity analysis of the effects of variation in the discount rate and the age weights (Murray et al., 1994) have been published and are not reviewed here. Between 1980 and 1988 the observed coefficient of incidence (new cases per 100,000 population) rose steadily from 1,311 to 3,461, or from a little more than 1% of the population being sick with malaria each year, to almost 3.5%. Since this increase was nearly linear, it was assumed that the same constant rate of increase would prevail in 1989­1996 as in 1980­1988, reaching a level of 5,611 or more than 5.5% of the population suffering an attack of malaria in one year. This is still well below what expert malariolo- gists consider to be saturation levels of incidence, and there is no evidence in the Brazilian data of a cyclic or other non-linear pattern. When incidence is projected separately for each state in the Amazon Basin, the sum of pro- jections rises very rapidly in 1988­1991 before leveling off at a coefficient of more than 7,000 per 100,000 population, more than double the assump- tion used here for the region as a whole (Akhavan, 1997). In contrast to the projection, the observed coefficient declined in most years and never exceeded 3.5% of the population. Applying the difference between observed and projected incidence to the population of the Amazon Basin gives the number of cases estimated to have been prevented each year. This is the major source of uncertainty in the analysis; where treatment is concerned, the only uncertainty is how many cases may have occurred and gone untreated, but that does not affect the cost-effectiveness of the Cost-Effective Malaria Control in Brazil · 255 program. For 1996, only half the cases treated or estimated to have been prevented are counted, because the project ended in June of that year and the calculation of costs runs through only half the year. The observed severity, or share of malaria cases due to falciparum, reached a peak of 55% in 1986 and stayed in the range 0.53­0.55 from 1984 to 1987. It is assumed that in the absence of the program, severity would have remained at that peak; again, this number appears to be below the sat- uration equilibrium level. Falciparum parasites become available to infect vectors in only about 5.5 days and the incubation period before symptoms begin is 12­14 days; for vivax parasites these intervals are, respectively, 8 and 11­12 days. In consequence of the shorter time before transmission is possible and the longer time when a person is infected but asymptomatic, the basic reproductive rate of falciparum is at least double that of vivax, which means that severity may saturate at around 60­65%. Finally, 10% of those sick with falciparum are assumed to die within a short period if not treated, while the other 90% of sufferers from falciparum and all those infected with P. vivax are assumed to be sick for 4 months (0.333 years) before recovering fully. This period is so short that discounting the future at 3% makes almost no difference. Age-weighting, however, makes a substantial difference, because most deaths occur in adolescence or early adulthood, when the weights used to calculate DALYs are near the maximum value of 1.525. The result is that the discounted and age-weighted interval of illness is 0.433 years, longer than the calendar interval; being sick at those ages is worse than average, for any level of disability. Some DALYs may also be lost in the future, because vivax malaria can cause relapses (Miller and Warrell, 1990). Since little is known about their frequency or severity, except that they tend to become briefer and less severe, this possi- ble contribution to morbidity is ignored here and discussed briefly later. The probability of dying from falciparum malaria in the absence of any treatment is the third key parameter in the estimate of deaths prevented by the control measures, and therefore in the cost-effectiveness of those meas- ures, and requires some discussion. Beneson (1997, p. 350) states that "prompt treatment is essential, even in mild cases, since irreversible com- plications may appear suddenly; case-fatality rates among untreated chil- dren and non-immune adults exceed 10% by a considerable margin". There are few other published estimates of this rate, so several expert malariolo- gists were consulted (Mangabeira da Silva, 1996; Campbell, 1997; Collins, 1997; Hoffman, 1997); all agreed that 10% is a conservative assumption, 256 · Health Economics in Development their own estimates ranging as high as 30­45%. Case-fatality rates in the range of 25 to 45% have been observed in Africa, but greater malnutrition and higher incidence among children, who are particularly susceptible, may make malaria more lethal there than in Brazil. Death in 10% of untreated falciparum cases appears unlikely to overstate the disease risks or the cost- effectiveness of control measures in the Amazon Basin. Applying the disability weights (0.375 for falciparum and 0.22 for vivax) to the interval of illness leads to a loss of 0.162 DALYs per non-fatal, untreated case of falciparum and 0.095 per case of vivax. The total health gain from preventing this temporary disability is some 226,000 DALYs. Although only 5.5% of malaria sufferers would die (10% of the 55% infected with falciparum), they would account for a much greater health loss. The average age at death from malaria during this period was 14 years, at which age life expectancy under the life table assumptions employed (Mur- ray, 1994) is still 66 years. Age-weighting and discounting reduce this to 36.27 DALYs lost per death. This value is not very sensitive to the assump- tion that life expectancy is as high as 66 years, because both discounting and age-weighting greatly reduce the importance of years far in the future. (Varying the assumption about life expectancy would, of course, affect other measures of health gain such as potential years of life lost.) The total health gain from preventing deaths through vector control is about 3.65 million DALYs, and the total gain from prevention of cases, including reduced morbidity, is 3.88 million DALYs, as shown in the bottom line of Table 13.3. Both because of mortality and because morbidity is more severe, nearly all this gain comes from controlling P. falciparum; control of P. vivax accounts for less than 2% of the total. Estimating the DALY gains from treatment is simpler, because only the observed incidence and severity matter; these determine the number of suf- ferers from P. falciparum who, if not treated, would have a 10% chance of dying of the disease. Lethality with treatment is much lower; it reportedly declined from 0.72 to 0.40% between 1980 and 1988 and continued to fall to 0.15% in 1996. However, it is suspected that overall mortality is sub- stantially under-reported (Akhavan, 1996): some people die without getting treatment and some die even after receiving ambulatory care and these deaths may not be registered. The only study of under-reporting of hospi- tal deaths, undertaken in the state of Rondônia in 1985 (Fiusa Lima and da Silva, 1988), found that actual deaths were 38.8% higher than reported. To take account of this, it is conservatively assumed that on average through- out the period, 0.78% of treated falciparum cases would have died; this Cost-Effective Malaria Control in Brazil · 257 applies the 1985 under-reporting estimate to the average reported lethality in 1980­1988 (1.388 times the average of 0.40 and 0.72 is 0.78), with no allowance for subsequent decline. Lives saved by treatment are therefore the number of people sick with fal- ciparum, discounted to present value, times the difference between untreated and treated lethality, 9.22% (10.0 0.78). Savings from reduced morbidity are similar to those gained by prevention, with the difference that it is assumed the typical patient would still suffer for 12.5 days, 10 before seeking treatment and 2.5 days from the start of treatment to the cessation of symp- toms. (This is a very conservative assumption; if malaria victims received care within a few days, the health gain from treatment would be substantially larger.) This means that the interval of health gain drops from 0.333 to 0.299 years, or 0.388 years when discounted and age-weighted. This does not apply to those falciparum sufferers saved from death, because they would probably have died quickly rather than being sick for four months. Tables 13.2 and 13.3 show how the estimates of health gains were con- structed, following the assumptions described earlier. The total number of malaria cases which occurred despite the control efforts was 4.1 million, or 3.7 million in 1996 present value. The total number of cases prevented is estimated as 1.97 million, equivalent in present value to 1.83 million cases. Expressing the estimates in present value terms takes account of their occur- rence over a span of almost eight years, so as to match the discounted cost estimates. The assumptions about severity and lethality imply that 101,000 more people would have died of malaria had it not been for the preventive component of the project and another 1.7 million would have been sick. The effectiveness of concentrating vector control in areas where P. falci- parum was most prevalent shows in the substantial reduction of reported fal- ciparum cases, while the number of P. vivax cases fluctuated with no trend. Under the assumption about lethality described above, treatment resulted in saving almost 130,000 lives and in reducing the morbidity caused by 1.27 million cases of falciparum and 2.19 million cases of vivax. All together, some 5.5 million people benefited from the control program, either as patients or because they did not get malaria. Table 13.3 summarizes the sav- ings in lives, cases and DALYs, according to the type of malaria and whether the gains came from prevention or treatment. The total gain measured in DALYs was 8.97 million: 5.08 million from treatment or 31% higher than the gain of 3.88 million DALYs from prevention. Fully 93% of this health gain derived from preventing deaths, with almost no difference in the share between prevention and treatment. and AX VIV LUEAV 238,494 253,316 276,647 292,048 363,709 338,018 351,800 157,729 1,060,505 1,211,256 2,271,761 Sick Prevented PRESENT ARUM 63,871 1.00 2,271,761 Cases ALCIPF 215,037 204,030 183,384 211,083 157,128 181,454 192,884 813,534 595,337 1,408,871 Sick AX 0.90 1,267,983 VIV TED 293,318 302,473 320,710 328,703 397,435 358,603 362,354 157,729 1989­1996. 1,245,204 1,276,121 2,521,325 Dead REPOR 129,897 0.10 0.0078 0.0922 CASES ARUM Malaria, 63,871 ALCIPF 264,469 243,622 212,592 237,576 171,698 192,505 198,671 958,259 626,745 1,585,004 reatingT Sick and 32,856 91,726 vivax PRESENT ALUEV 156,833 186,766 330,147 333,787 399,349 299,208 468,181 1,362,491 1,830,672 0.45 823,803 1.00 823,803 and Preventing CASES 40,409 by PREVENTED 109,526 181,813 210,207 360,760 354,115 411,329 299,208 541,955 1,425,412 1,967,367 Sick falciparum 0.90 SavedsY TION 0.55 1,006,870 906,183 16.04 16.40 16.78 17.12 17.48 17.84 18.20 18.56 POPULA (MILLIONS) DAL Morbidity Dead and and 0.10 100,687 252 668 1085 1228 2064 1985 2260 3224 Lives Lives DIFFERENCE in VED Morbidity) Cases, INCIDENCE/100,000 Savings 3478 3331 3183 3308 2741 3089 3082 2388 eatment)rT OF OBSER Prevented Prevented and from (Reduced 1989­92 1993­96 1989­96 Estimated ted Cases Cases Untreated COEFFICIENT 3730 3999 4268 4536 4805 5074 5343 5611 if eatedrT (Gain Lives Cases PROJECTED of of if in in 13.2 Repor Strategy Strategy: Period: ableT YEAR 1989 1990 1991 1992 1993 1994 1995 1996 otalsT Old New Entire Severity Number Outcome Lethality Difference Saving Saving Cases 258 Savings AL VERAGE)A of TOT (OR 230,584 5,269,550 (0.314) (3.07) (1.91) (1.62) 8,967,107 L Conversion FATA AX VIV NON- CASES 2,271,581 0.22 0.299 0.388 0.085 193,084 TMENT TREA L 1989­1996. FATA 5,089,574 FROM NON- CASES 1,267,983 0.375 0.299 0.388 0.146 185,126 Malaria, VINGS ARUM SA treatment: ALCIPF reatingT from THS DEA 129,897 1.00 66.19 36.27 36.27 4,711,364 otalT and L FATA Preventing AX VIV NON- CASES 823,803 0.22 0.333 0.433 0.095 78,475 by L Saved Ys earsY PREVENTION FATA 3,877,533 DAL Life FROM NON- CASES 906,183 0.375 0.333 0.433 0.162 147,141 ARUM and VINGS prevention: SA ALCIPF Lives from THS DEA 100,687 1.00 66.19 36.27 36.27 3,651,917 otalT Cases, Disability-Adjusted to X age- Estimated Cases years and lives cases weight case in weight duration) Ys 13.3 and in in per DAL Ys Lives ableT Saving Saving Disability Duration Unadjusted Discounted weighted DAL (disability adjusted otalT in 259 260 · Health Economics in Development This concentration reflects the fact that on average, a beneficiary of the control program saved only three years of life, or less than two DALYs, but someone who would otherwise have died gained 66 years of life and 36 DALYs. In contrast to the results for preventing cases, where falciparum was nearly twice as important as vivax, the gains from reducing morbidity through treatment are almost equally divided between falciparum and vivax: this reflects the success of the program in reducing severity when it would otherwise have remained constant. Estimating Program Costs and Cost-Effectiveness The estimated health gains just described were derived from both preven- tion and treatment, so the costs attributable to malaria control include both kinds of expenditures. Spending on prevention (vector control) was partly through the World Bank project (PCMAM), all of which occurred in the Amazon Basin, and partly through the NHF malaria control program, which operated in the whole country; it is estimated that 92% of those expenditures occurred in the Amazon. Capital investment and non-salary recurrent expenditures (insecticides, travel costs and per diem for field workers) of the NHF program were directly recorded and charged to the program. Labor costs are not budgeted to the control program; they were estimated from the number and type of personnel needed for vector control operations and salaries for each level of worker. Expenditure on treatment was partly for hospitalization and partly for ambulatory care and includes the cost of diagnosis. The number of hospi- talizations rose from 7,000 or less before 1987, to over 20,000 in 1988 and to a peak of over 53,000 in 1992. As treatment improved, the average length of stay fell abruptly from around 5.5 days before 1992, to about 4.4 days from that year on. The number of ambulatory treatments rose from around 30,000 in 1984­1985 to around 500,000 in 1992­1995. The cost of ambu- latory care was estimated by assuming that all reported but non-hospitalized cases were treated, and applying to each patient the costs of four ambula- tory visits, two blood tests and the required medications, equal to US $18.70. `Costs' of both hospital and ambulatory care are what the federal health system (SUS) paid for treatments, and understate true costs to the extent that they do not cover salaries of staff of the Ministry of Health or NHF, or the capital cost of existing hospitals and clinics. (New invest- ments are included in the cost estimates.) SUS tariffs are believed to be Cost-Effective Malaria Control in Brazil · 261 significantly above or below true costs for many interventions, but the fact that it was possible to expand treatment rapidly in the program suggests that for malaria, federal payments were adequate to cover at least all recurrent costs. Since the more severe hospitalized cases were almost entirely due to falciparum, ambulatory patients were assumed to be infected with falciparum in only 44% of cases and with vivax in the rest. However, the cost of med- ication is only 8% of the cost of ambulatory treatment, so differences in which drugs are administered, reflecting differences in the distribution of the two kinds of malaria, have little effect on total costs. All financial information was initially reported in Brazilian currency and underwent three transformations to make the numbers comparable in pres- ent value. These are (i) conversion from Brazilian currency to US dollars in the current year, using the average of the official buying and selling rates for the dollar in that year; (ii) adjusting to 1996 US dollars by the US GDP deflator for the current year; and (iii) discounting to 1996 present value at 3% per year. These adjustments remove, so far as possible, the effect of inflation in the US and the much more rapid inflation (prior to 1995) in Brazil, and express all values in constant US dollars of 1996 purchasing power. The discount factor (3%) is the same used to discount numbers of cases prevented or occurring, and therefore to discount the sums through times of lives and DALYs saved. For recurrent expenditures (largely or entirely consumed in the same year as purchased), only these adjustments are necessary. Capital expendi- tures require two further adjustments. The first is to incorporate deprecia- tion over the useful life of the capital. This is assumed to be constant (lin- ear), with different lifetimes for the three categories; 25 years for buildings, 10 years for vehicles and 5 years for equipment. The cost to the program in a given year is then the value of the capital investment divided by the use- ful life. Capital entirely depreciated during the project has its cost entirely attributed to the program; capital with a useful life extending beyond 1996 is charged to the program only for 1996 and prior years. The second adjust- ment is to take account of the opportunity cost of capital. In a stable eco- nomic environment this would be given by the rate of interest, but the high inflation in Brazil prior to stabilization in 1994 led to very high nominal and even real interest rates. The US Prime Rate was therefore used as a better approximation (although possibly an underestimate) of the true opportunity cost of capital. All these steps are shown in the initial evaluation (Akhavan, 1996). Table 13.4 summarizes the final result: expenditures in discounted dollars of 1996 1989­1996 7,575 3,599 307.1 92.57 16.46 48,820 37,646 78,567 90,534 28,430 62,105 AL 525,620 476,800 398,234 3,773.1 616,155 TOT 714 1989­1995) 1993­1996 1,767 159.0 87.77 17.13 13,859 11,379 42,188 43,387 13,955 29,431 for AL 185,111 171,252 129,064 1,717.6 228,498 TOT 3% at 99 17 813 697 15.7 1996 5,496 4,940 1,089 69.22 3,851 269.7 14.28 22,986 22,173 16,677 27,926 Discounted 1995 82 63,878 4,726 305 4,340 59,152 10,669 48,483 12,238 2,818 42.3 66.57 9,419 518.7 18.16 76,115 US$, 240 348 52.4 1994 1996 4,897 4,309 7,708 5,159 98.42 8,792 498.7 17.63 34,688 29,791 22,083 13,951 48,640 of 267 48.6 1993 3,423 1,123 2,033 4,889 7,369 430.5 17.12 63,559 60,136 18,315 41,821 12,258 100.53 75,817 (Thousands 1989-1992 5,808 2,886 148.0 97.80 15.90 Patients) AL 34,961 26,266 36,379 47,149 14,474 32,674 340,510 305,549 269,171 2,055.4 387,658 y TOT 1989­1996 464 53.5 1992 Ambulator 1,143 8,632 8,768 4,789 89.44 8,521 512.7 16.62 77,210 10,239 66,971 58,203 13,310 90,520 Program, or 634 312 43.2 1991 6,548 5,602 8,191 4,566 7,908 490.1 16.14 53,221 46,673 38,482 12,474 105.72 65,694 Control Hospital of 611 1990 28.5 1,168 9,407 2,987 8,109 517.6 15.67 Malaria 86,636 11,800 10,020 74,836 65,429 11,096 104.98 97,732 the of (Thousands 22.8 1989 6,374 2,863 1,499 2,012 2,132 93.47 8,136 535.0 15.21 10,013 10,269 123,443 117,069 107,057 133,712 Costs reatedT y 13.4 Y cases Cost cases cost Cases TEGOR Buildings ehiclesV costs Equipment Salaries Other Investment No. Unit No. Unit ableT CA Prevention Recurrent eatmentrT Hospital Ambulator otalT and 262 Cost-Effective Malaria Control in Brazil · 263 value. Overall, from 1989 through the end of PCMAM in mid-1996, the malaria control program is estimated to have cost US $616.2 million, divided between US $525.6 million for vector control and related activities and US $90.5 million for treatment. About 30% of treatment cost was for hospitalization and 70% for ambulatory care. Preventive expenditures included US $48.8 million of investment in buildings, vehicles and equip- ment, US $78.6 million in salaries and US $398.2 million for all other oper- ating costs, including the costs of preparing and managing the project, insecticides, fuel, training and staff travel and per diem. Since the program is estimated to have prevented some 1.83 million cases of malaria (in present value), it also eliminated the need to treat that number of episodes of disease. Assuming the same distribution between hospitalization and ambulatory care as in the cases actually treated, the unit cost per patient would have been the same, averaging US $22.19 over the period. This implies a total saving in treatment expenditures of US $41.5 million. (Savings would of course be smaller, if some of the prevented cases would not in fact have been treated had they occurred.) Subtracting that from estimated actual expenditures leaves a net cost of the malaria control program of US $574.6 million, and implies that for every dollar spent on prevention, seven cents were saved in treatment. Comparing the costs to the estimated saving in lives gives overall cost- effectiveness estimates of US $2,672 per life saved, or US $2,492, taking account of savings in treatment costs. These ratios ignore the much smaller health gains from reduced morbidity. If costs and gains are compared for prevention alone, the ratio is higher, because prevention accounted for the bulk of the costs but saved slightly less than half the total lives; the corre- sponding figures are US $5,200 per life saved without taking account of treatment cost savings and US $4,808, counting those savings. In contrast, treatment appears to have cost only about US $700 to save a life. The last three columns of Table 13.5 summarize the results just presented for cases prevented and lives saved, which are based on the most plausible assump- tions discussed in the text and incorporated in Tables 13.2, 13.3 and 13.4. When the savings in morbidity are included and compared to lives saved, using DALYs as the composite measure, the result is an overall cost- effectiveness of US $69 per DALY (US $64, if based upon the cost net of savings in treatment). As with lives saved, the cost is higher for prevention, with ratios of US $136 and US $125, respectively, as savings are or are not taken into account. With treatment, it cost less than US $18 to save one disability-adjusted life year. 264 · Health Economics in Development Discussion Since the project helped to introduce a major change in malaria control strategy halfway through the period analyzed, Table 13.5 presents the cost- effectiveness results separately for 1989­1992 and for 1993­1996, consider- ing only lives saved by preventing or treating cases of falciparum. The same assumptions about incidence are used as for the analysis of the whole period 1989-1996, because the control efforts in 1989­1992 do not appear to have had much effect on incidence. The cost of saving a life by treatment rose 25% between the early and the later years, presumably as a result of more aggressive care (a higher share of patients hospitalized). However, the cost of saving lives by prevention fell dramatically when preventive efforts were more sharply focused against falciparum, from about US $13,000 to between US $2,000 and US $2,500. This improvement in efficiency did not come at the cost of the absolute number of lives saved by preventing cases, which nearly tripled between one period and the other, while expenditures on prevention were cut almost in half. The overall result was to reduce the average cost of preventing a death from nearly US $4,000 to under US $2,000, or as low as US $1,500 when savings in treatment cost are consid- ered. (These savings increased in line with the gain in efficiency of preven- tive efforts.) This analysis validates the new control strategy and illustrates how wasteful the preventive program previously was, which explains why the epidemic had continued to expand. In the absence of the preventive effort, incidence was projected to rise lin- early and rather rapidly, which could lead to an over-estimate of the number of cases prevented and therefore of lives saved by prevention. (This would not, of course, affect the estimate of how many lives were saved by treatment.) By the same token, the assumption of increasing incidence could bias the comparison between the early and later periods; higher projected numbers of cases in the later years might magnify the apparent gains from the revised strategy. In Table 13.6, incidence is subjected to sensitivity analysis, to see how much overall cost-effectiveness, and the contrast between 1989­1992 and 1993­1996, depend on assumptions as to how many cases of malaria would have occurred in the absence of the project. Expenditures on the pro- gram are assumed to be the same as those actually recorded or estimated. Incidence is first assumed to remain constant at the level of 1988, or 3,361 per 100,000 population. So far as the period 1993­1996 is concerned, this is roughly equivalent to assuming that if the control strategy had not changed, the earlier strategy would have maintained incidence constant at about the 1989­92 level, so that any gains in the later period would AL 2,672 2,492 Most TOT 41,515 616,155 574,639 230,584 1,006,870 1,408,871 alues).V 1989­1996 ---- ---- 697 697 TMENT 90,534 90,534 PERIOD, TREA 129,897 1,408,871 Present ENTIRE ---- 5,220 4,808 41,515 525,620 484,105 100,687 PREVENTION 1,006,870 (Discounted AL TOT 1,760 1,517 31,501 228,498 196,997 749,370 595,337 129,827 1993­1996 1993­1996, TMENT ---- ---- 790 790 TEGY 43,387 43,387 54,890 TREA 595,337 STRA 1989­1992,, NEW ---- 2,470 2,050 31,501 74,937 PREVENTION 185,111 153,610 749,370 ext)T AL Falciparum and TOT 3,847 3,748 10,014 387,658 377,644 257,500 813,534 100,757 from 13.3 1989­1992, ---- ---- 629 629 Lives TMENT and TEGY 47,149 47,149 75,007 TREA 813,534 STRA Saving 13.2 OLD of ---- ablesT 10,014 25,750 13,224 12,835 PREVENTION 340,510 330,496 257,500 (from US$) (US$) prevented treated cost US$) cost Cost-Effectiveness prevented US$) cases cases saved total net Assumptions (thousand life on on 13.5 (thousand cost saved per (thousand Based Based ableT CONCEPT Cost Expenditures Net Falciparum Falciparum Lives Cost Plausible 265 ity) Present 1,730 1,503 68,944 525,620 456,676 303,783 Lethal 1989­1996 3,037,830 or 7,500 (Discounted TO,Y Severity in STEEPL 808 576 52,989 185,111 132,122 229,192 RISING 1993­1996 2,291,918 1989­1996 Change and (No INCIDENCE Incidence 4,565 4,351 1989­1992 15,955 74,591 1993­1996 340,510 324,555 745,912 Projected to 1989­1992,, 11,245 48,982 10,731 10,501 1989­1996 525,620 514,375 489,819 Related 3,461 LEVEL, Falciparum 1988 Prevention TA from of 4,170 3,939 ANT 10,262 44,385 1993­1996 185,111 174,849 443,848 Lives CONST Saving of INCIDENCE 983 Cost-Effectiveness 4,597 45,971 74,072 73,858 1989­1992 340,510 339,527 Analysis: Cost-Effectiveness US$) US$) US$) total net cases on on Sensitivity life (US$) 13.6 cost cost cost saved per Based Based (thousand prevented (thousand (thousand prevented saved ableT alues).V CONCEPT Cost Expenditures Net Falciparum Lives Cost 266 Cost-Effective Malaria Control in Brazil · 267 represent the improvement over the previous strategy rather than gains with respect to no program at all. The results of the comparison are that if there had been no danger of an expanding epidemic, only 49,000 lives would have been saved by prevention and the cost of saving a life would have been correspondingly much higher, US $10,731. Since fewer cases prevented would also mean smaller savings in treatment, only US $11.2 million, there is little difference according as total or net cost is used. The most striking result, however, is that since observed incidence hardly changed in 1989­1992, very few lives would have been saved under the old strategy, and the cost per life saved would have been as high as US $74,000. Nearly all the gain would have come under the revised strategy, with a cost per death prevented only slightly higher, around US $5,400, than with ris- ing incidence. The apparent improvement in cost-effectiveness therefore is not an artifact of assuming that the epidemic would have continued to grow. All the indications are that, on the contrary, Brazil faced an explosive epi- demic of malaria and that the reorientation and concentration of prevention was crucial in bringing it under control. Applying the projected incidence in each state separately (Akhavan, 1997) would only strengthen this conclusion. If in the absence of the project the coefficient of incidence had risen steadily to 7,500 (which is slightly slower than the sum of state-level projections), the preventive component of the project would have saved 304,000 lives and the overall cost-effectiveness would have been between US $1,503 and US $1,730 per life saved. Savings in treatment cost might also have been much larger, US $568.9 million, although it is less plausible that all this saving could have been realized, since it would imply that Brazil had the capacity to treat an additional three million cases of falciparum malaria alone. Assuming very rapid growth in cases is also more favorable to the old strategy, which would have prevented 75,000 deaths during 1989­1992 even without reduc- ing observed incidence, bringing the cost per life saved down to about US $4,500. Even so, the revised strategy continues to appear more cost-effective, saving lives through preventing cases for only US $600 to US $800. Of the three crucial parameters in those calculations, only the severity or share of falciparum cases was both readily observable and nearly constant during the interval of the project. There is therefore no apparent reason to evaluate the project using different estimates of severity. The other impor- tant parameter is the case fatality rate, or lethality, for untreated falciparum, since that determines the health gains from both treatment and prevention. As indicated above, that rate was assumed to be 10%, but expert opinion suggests that it could be substantially higher. Table 13.7 shows the outcome