Human Development Network Health, Nutrition, and Population Series Health Policy Research in South Asia Building Capacity for Reform Abdo S. Yazbeck David H. Peters Editors 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 (Stuart Gillespie, Milla McLachlan, and Roger Shrimpton, editors) Health Expenditures, Services, and Outcomes in Africa: Basic Data and Cross- National Comparisons, 1990­1996. 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) Improving Women's Health in Pakistan. 1998 (A. G. Tinker) 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) 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 Policy Research in South Asia Human Development Network Health, Nutrition, and Population Series Health Policy Research in South Asia Building Capacity for Reform Abdo S. Yazbeck David H. Peters EditorsAtsuko AoyamaAtsuko Aoyama THE WORLD BANK Washington, D.C. © 2003 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. First printing August 2003 1 2 3 4 06 05 04 03 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 concerning 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. ISBN 0-8213-5531-7 Library of Congress Cataloging-in-Publication Data Health policy research in South Asia : building capacity for reform / edited by Abdo S. Yazbeck and David H. Peters. p. cm. -- (Health, nutrition, and population series) Includes bibliographical references. ISBN 0-8213-5531-7 1. Medical policy--Research--South Asia. 2. Medical policy--South Asia. 3. Health care reform--South Asia. 4. Public health administration--South Asia. I. Yazbeck, Abdo. II. Peters, David H., 1962­ III. Series. RA395.S628H436 2003 362.1'07'2054--dc21 2003052538 Contents Foreword xvii Sadiq Ahmed, World Bank Acknowledgments xix Acronyms and Abbreviations xxi Section I: Introduction 1 1 Overview 3 Abdo S. Yazbeck, World Bank David H. Peters, Johns Hopkins University 2 A Framework for Health Policy Research in South Asia 23 David H. Peters, Johns Hopkins University Abdo S. Yazbeck, World Bank Section II: Analysis of Inequality 31 3 The Distribution of Public Health Subsidies in India 33 Ajay Mahal, Harvard University v vi · Health Policy Research in South Asia 4 Equity in Financing and Delivery of Health Services in Bangladesh, Nepal, and Sri Lanka 65 Institute of Policy Studies, Sri Lanka 5 Geographic Resource Allocation in Bangladesh 101 Tim Ensor, University of York Atia Hossain, Priti Dave Sen, Liaquat Ali, Shamin Ara Begum, and Hamid Moral, Ministry of Health and Family Welfare, Bangladesh Section III: Expenditure Analysis 129 6 Public Expenditure Review of the Health and Population Sector Program in Bangladesh 131 Tim Ensor, University of York Atia Hossain, Priti Dave Sen, Liaquat Ali, Shamin Ara Begum, and Hamid Moral, Ministry of Health and Family Welfare, Bangladesh 7 Sri Lanka's National Health Accounts: National Health Expenditures 1990­1999 163 Institute of Policy Studies and Ministry of Health, Sri Lanka 8 The Bangladesh Health Facility Efficiency Survey 195 Ravi P. Rannan-Eliya, Institute of Policy Studies, Sri Lanka Aparnaa Somanathan, Institute of Policy Studies, Sri Lanka Section IV: Private Sector Analysis 227 9 Private Health Care Sector in India--Policy Challenges and Options for Partnership 229 V. R. Muraleedharan, Indian Institute of Technology (Madras) Sunil Nandraj, World Health Organization, India Office Contents · vii 10 Private Health Provision in Uttar Pradesh, India 257 S. Chakraborty, Indian Institute of Management, Lucknow, Uttar Pradesh, India 11 Private Primary Care Practitioners in Sri Lanka 279 Ravi P. Rannan-Eliya and Prashanthi Jayawardhane, Institute of Policy Studies, Sri Lanka Leela Karunaratne, Department of Family Medicine and Community Medicine, University of Sri Jayawardanepura, Sri Lanka Section V: Consumer and Provider Perspectives 305 12 Consumer Redress in the Health Sector in India 307 Bejon Misra, Consumer VOICE, Delhi 13 Quality Health Care in Private and Public Health Care Institutions 333 Prasanta Mahapatra, Institute of Health Systems, Hyderabad, Andhra Pradesh, India 14 Voices of Stakeholders in the Health Sector Reform in Bangladesh 369 Nilufar Ahmad, South Asia Social Development and Environment, World Bank Tables 1.1 Selected Health Indicators in South Asian Countries 4 1.2 Inequality in Health Outcomes and Health Sector Outputs in South Asia 8 3.1 Unit Cost Estimates from Facility-Level Data in India 39 3.2 Distribution of Health Care Services in Public Health Facilities by Type of Service, State/Region, and Socioeconomic Status, 1995/96 42 viii · Health Policy Research in South Asia 3.3 Distribution of Use of Services by Public and Private Facilities by Type of Service, State/Region, and Per Capita Expenditure Quintile, 1995/96 44 3.4 Share of Richest 40 Percent of the Population in Total Hospital Charges Paid to Public Health Facilities by State and Region, 1995/96 51 3.5 Distribution of Public Subsidies by State/Region and Socioeconomic Status, 1995/96 55 4.1 General Indicators for Bangladesh, Nepal, and Sri Lanka, 1997 72 4.2 Background Information on National Health Systems of Bangladesh, Nepal, and Sri Lanka 73 4.3 Overview of Survey Data Used in the Study 76 4.4 Infant and Under-Five Mortality Rates by Income Quintile of Equivalent Consumption for Bangladesh, Nepal, and Sri Lanka 78 4.5 Infant and Under-Five Mortality Rates by Education of Mother and Place of Residence for Bangladesh, Nepal, and Sri Lanka 79 4.6 Distribution of the Benefits of Government Expenditures on Personal Medical Services in Bangladesh and Sri Lanka 84 4.7 Distribution of the Benefits of Government Expenditures on Collective Services and on All Services in Bangladesh and Sri Lanka 85 4.8 Variations in Government Expenditures on Health by Region 86 4.9 Distribution of Payments for Health Care for Each Source in Bangladesh and Sri Lanka by Income Decile 90 4.10 Kakwani and Suits Indexes for Different Payment Mechanisms for Health Care in Sri Lanka 91 4.11 Redistributive Impact of Government Taxation and Financing of Health Care Services in Sri Lanka, 1995/96 92 Contents · ix 4.12 Redistributive Impact of Government Taxation and Financing of Health Care Services in Bangladesh, 1996/97 93 5.1 Allocation of Funding to Public Facilities 104 5.2 Typical Budgets for a District Hospital and an Upazila Health Complex 106 5.3 Distribution of Spending on Public Health in an Average District by Type of Facility 107 5.4 Range of Public Health Spending Per Capita across Districts by Division and Presence of a Medical College Hospital (MCH), Fiscal 2000 108 5.5 Inpatient and Outpatient Use of District Hospital (DH) and Medical College Hospital (MCH) Facilities, by Selected District and Proximity of Patient Residence to Facility 119 6.1 Financial Indicators of the HPSP 133 6.2 Public Expenditure Allocation of Benefits by Gender and Type of Service 149 6.3 Average Payments per User by Income Group and ESP Category 152 6.4 Revenue Projections under Different Assumptions 158 7.1 Total Expenditures on Health (TEH), 1990­99 168 7.2 Per Capita Health Expenditures, 1990­99 169 7.3 Total Expenditures on Health at Current Market Prices, 1990­99 169 7.4 Total Expenditures by Function, 1990­99 174 7.5 Relative Share of Funding by Public and Private Expenditures to Selected Functional Categories in 1997 175 7.6 Relative Share of Funding by Central, Provincial, and Local Governments to Selected Functional Categories in 1997 177 7.7 Total Government Expenditures by Function, 1990­99 178 x · Health Policy Research in South Asia 7.8 Total Nongovernment Expenditures by Function, 1990­99 179 7.9 Total National Expenditures by Type of Provider, 1990­99 180 7.10 Total Government Expenditures by Type of Provider, 1990­99 181 7.11 Total Nongovernment Expenditures by Type of Provider, 1990­99 182 7.12 Per Capita Expenditures by Source, 1997 185 7.13 Per Capita Expenditures by Central Government, 1990­99 185 7.14 Per Capita Expenditures by Provincial Council and Local Government, 1990­99 186 7.15 Total Expenditure on Health Per Capita Not Directly Attributable to Any Province, 1990­99 186 7.16 National Health Expenditures in Selected Asia-Pacific and Organisation for Economic Co-operation and Development Countries and Economies 190 7.17 International Comparison of Expenditures by Source 191 8.1 Allocation of Recurrent Costs to Inpatient and Outpatient Services 203 8.2 Distribution of Sampled Facilities in Survey by Type and Division 204 8.3 Key Statistics by Category of Facility 205 8.4 Available Equipment at Facilities 207 8.5 Available Utilities at Facilities 207 8.6 Staffing Indicators and Ratios 208 8.7 Allocation of Staff Time to Inpatient Care 209 8.8 Average Annual Number of Outpatient Services and Tests by Category and Type of Facility 210 8.9 Inpatient Service Statistics 211 8.10 Distribution of Recurrent Costs by Category of Cost 213 8.11 Share of Recurrent Costs Accounted for by Inpatient Use 214 Contents · xi 8.12 Breakdown of Recurrent Costs in Providing Inpatient Services 214 8.13 Breakdown of Recurrent Costs in Providing Outpatient Services 214 8.14 Gross Unit Costs for Inpatient and Outpatient Services 215 8.15 Cost per Bed-Day Occupied by Type of Facility and Division 217 8.16 Cost per Admission by Type of Facility and Division 217 8.17 Cost per Outpatient Visit by Type of Facility 218 8.18 Indicators of Resource Availability at THCs by Division 218 8.19 Comparison of Hospital Services Statistics 220 8.20 Comparison of Hospital Unit Costs as a Percentage of Per Capita Gross National Product 221 8.21 Comparison of Hospital Staffing Indicators 222 10.1 Basic Demographic Indicators for Uttar Pradesh and India 260 10.2 The Burden of Disease in Uttar Pradesh, 1994 261 10.3 Estimated Number of Inpatient Institutions and Beds in the Public and Private Sectors in Uttar Pradesh, 2000 261 10.4 Brief Statistics of Chosen Districts 263 10.5 Number of Samples Taken in Each Study Location 264 10.6 Questionnaire Type, Respondents, and Information Sought 265 10.7 Diagnostic Services Offered by Health Care Institutions and Solo Practitioners 267 10.8 Nature of Concessions Offered to Poor Patients 271 10.9 Service Charges at Private Health Care Institutions 272 10.10 Opinions Expressed by Patients on Exit 273 11.1 Geographical Distribution of General Practitioners, 1984­2000 289 xii · Health Policy Research in South Asia 11.2 Age and Sex Distribution of Private Practice Patients, 2000 292 11.3 Most Frequently Prescribed Drugs in GP Practices in Sri Lanka, Australia, and the United States, 2000 295 12.1 Distribution of Samples Covered in the Hospital Sector 312 12.2 Distribution of Samples Covered in the Consumer Sector 312 12.3 Distribution of Facilities Surveyed by Availability of Patient Redress System 314 12.4 Distribution of Facilities with Some Consumer Redress Systems by Frequency of Review 315 12.5 Distribution of Complaint Boxes/Books: Claimed and Observed 316 12.6 Distribution of Types of Complaints Commonly Received According to Type of Facility 317 12.7 Time Taken to Respond to Complaints in Private and Public Hospitals 318 12.8 Distribution of Pending Cases by Reasons for Claims 319 12.9 Lawyers' Recommendations for Increasing the Efficiency of Consumer Forums 326 13.1 Health System Goals According to a WHO Framework, 2000 334 13.2 Integrated Framework for Assessment of the Quality of Health Care 339 13.3 Quality of Care Received by Ever-Married Women during Their Most Recent Visit to a Private or Public Sector Health Facility in India, 1998/99 343 13.4 Premises Ownership and Scale of Accommodation of Private and Public HCIs 347 13.5 Practice Guidelines, Medical Audits, and Medical Records in HCIs 350 13.6 Patient Exit Interview--Place of Interview and Respondent Characteristics 352 Contents · xiii 13.7 Patient Satisfaction Levels in Private and Public Health Care Institutions 354 13.8 Frequency of "Excellent" or "Very Good" Rating by Patients for Each of the 11 Items in the Quality of Service Questionnaire 356 13.9 Frequency of "Excellent" Rating by Patients for Each of the 11 Items in the Quality of Service Questionnaire 357 13.10 Level of Patient Satisfaction with Respect to Specific Facilities 358 13.11 Patients Responding "Yes" to Questions about Quality of Services by Respective HCIs 359 13.12 Recommendations to Improve Quality of Services in Private HCIs 361 Figures 4.1 Use of Public Outpatient Providers in Bangladesh and Sri Lanka 82 4.2 Use of Public Inpatient Providers in Bangladesh and Sri Lanka 83 4.3 Benefit Incidence of Government Health Expenditures--International Comparisons 87 4.4 Sources of Funding in the National Health Systems of Bangladesh, Nepal, and Sri Lanka 88 5.1 The Relationship between the U.N. Human Development Index and Public Spending Per Capita in Bangladesh, 1996 109 5.2 Distribution of Total Public Spending on Health by Upazila, 1999/2000 110 5.3 Public Spending on Health by Age Group in Bangladesh, 1998 114 5.4 Cumulative Effect of Weighting on Fiscal 2000 Division Allocations for Health in Bangladesh by Weighting Factor 118 xiv · Health Policy Research in South Asia 5.5 Potential Transition of Actual Spending to Needs-Based Health Spending Targets in Bangladesh by Division 122 6.1 Revenue and Development Spending on the HPSP 1995­2000, Original and Revised Budgets 135 6.2 The Distribution of Recurrent Expenditure by Salary and Nonsalary Items 137 6.3 ESP, Non-ESP, and "Super Overhead" Expenditures, 1999/2000 139 6.4 ESP and Non-ESP Spending in the HPSP Assuming Proportionate Overhead Allocation, 1999/2000 140 6.5 The Distribution of Revenue and Development Spending by ESP Component (Salary Spending Allocated According to Work Pattern Analysis)--Provisional Estimates, 1999/2000 142 6.6 The Distance from Equal Per Capita Targets (Division and All Other Districts), 1999/2000 144 6.7 Per Capita Public Health Expenditures by Income of District 146 6.8 Use of Services by Income Quintile 150 6.9 Average Patient Cost per Visit 151 6.10 Composition of Patients' Payments for Public Medical Treatment 153 6.11 Finance Available for Publicly Funded Health Care, 1999­2005 157 7.1 Trends in Total Expenditures on Health (TEH) for Sri Lanka, 1990­99 167 7.2 National Health Expenditures by Source as Percentage of GDP, 1990­99 170 7.3 National Health Expenditure by Source, 1997 171 7.4 Government Spending by Administrative Level, 1990­99 172 7.5 Total Government Expenditures by Function, 1997 175 Contents · xv 7.6 Total Nongovernment Expenditures by Function, 1997 176 7.7 Total Government Expenditures by Provider, 1997 183 7.8 Total Nongovernment Expenditures by Provider, 1997 184 7.9 Total Government Expenditures Per Capita by Province 187 7.10 National Health Expenditures by Source as Percentage of Gross Domestic Product 191 8.1 Performance Indicators for THCs by Division, Bangladesh, 1997 212 10.1 Percentage Distribution of Sources from Which Loans Were Received by Health Care Institutions (HCIs), Diagnostic Centers, and Solo Practitioners 268 10.2 Percentage Distribution of Specific Benefits Received from the Government 269 11.1 Age Distribution of Private General Practitioners, 1972­2000 291 12.1 Percentage of Facilities with an Office Procedure to Deal with Complaints 313 12.2 Percentage of Consumers Satisfied with the Consumer Redress Mechanisms in Pending Cases 321 12.3 Time Taken to Reach Final Judgment 321 12.4 Percentage of Defendants Who Engaged a Lawyer 324 Foreword In many ways the future is bright for South Asia. The past few years have seen amazing achievements in economic outlook, as well as in human development, health, and education. But South Asia contin- ues to face difficult challenges, especially in the areas related to the Millennium Development Goals. Almost half of the world's poor, living at less than $1 a day, live in South Asia. This pervasive poverty contributes to and is exacerbated by low levels of human develop- ment. The picture is of high levels of illiteracy, maternal mortality, and malnutrition. In order for South Asian countries to achieve a brighter future, investments in human capital are needed. Investments, however, will produce the desired results only if the enabling policy environments are in place. In the health sector, an enabling environment would mean equitable and efficient resource allocation policies, strong institutional frameworks, and a clear understanding of the role of the public sector as a steward of health sector activities for both private and public entities. Devising appropriate policies and building enabling institutions can take place only if there is a firm under- standing of the needs of each of the countries and of the constraints faced by the health system and the population. In other words, there is a clear need for health policy research to allow policymakers to design and implement programs. This book focuses on this theme, emphasizing the need for timely and homegrown health policy research to feed into the policy xvii xviii · Health Policy Research in South Asia process. The message is simple: South Asian researchers and institu- tions can and do produce high-quality, state-of-the-art health policy research and are able to influence the policy debate. By collecting and disseminating such a variety of research papers in the health sec- tor, the editors of this book not only help bring together a rich group of health policy research findings for a range of audiences in South Asia, but also more broadly contribute to learning for audiences in other regions facing similar challenges. Sadiq Ahmed Chief Economist and Sector Director Economic Management South Asia Region The World Bank Acknowledgments The editors would first like to thank all the authors who contributed to this volume and repeatedly responded to our requests for infor- mation. We would also like to thank Adam Wagstaff, who helped to initiate this project; this volume benefited from his ideas and sup- port. A number of managers in the South Asia Region of the World Bank supported this volume: Charles Griffin, Human Development Sector Director; Sadiq Ahmed, South Asia Region Chief Economist; Anabela Abreu, HNP Sector Manager for South Asia; and Hugo Diaz-Etchevehere, Acting Sector Manager. Alexander Preker, Chief HNP Economist for the World Bank, provided critical technical support, including managing and chairing a technical review. Within that review, we would like to thank the three reviewers for their very useful suggestions and comments: Peter Berman of the Harvard School of Public Health, Tom Merrick of the World Bank Institute, and Richard Skolnik of the George Washington School of Public Health. Many of the chapters in this volume were the result of Bank- financed sector work, and we would like to thank the World Bank staff that managed these activities and provided valuable technical support to the researcher: J. S. Kang, G. N. V. Ramana, and Rashmi Sharma. We also want to thank the editors and production staff for helping us and the chapter authors improve the quality and readabil- ity of the volume: Cindy Fisher, Mary Fisk, Fiona Mackintosh, and Nicki Marrian. xix Acronyms and Abbreviations ACT Advocacy for Control of Tuberculosis ADP Annual Development Plan AHCPR [United States] Agency for Health Care Policy Research ALOS Average length of stay ANM Auxiliary nurse midwife APHEN Asia Pacific Health Economics Network APHIDB Andhra Pradesh Health Institutions Database APNHAN Asia-Pacific National Health Accounts Network APP Alternative private practitioner APVVP Andhra Pradesh Vaidya Vidhana Parishad ASCI Administrative Staff College of India ATC Anatomic, Therapeutic, Chemical BBS Bangladesh Bureau of Statistics BCC Behavior Change Communications BCG Bacillus Calmette Guerin BEACH Bettering the Evaluation and Care of Health BGFES98 Bangladesh Facility Efficiency Study Phase I BIA Benefit incidence analysis BIMARU Bihar, Madhya Pradesh, Rajasthan, and Uttar Pradesh (India) xxi xxii · Health Policy Research in South Asia BIS Bureau of Indian Standards CB CFS Central Bank of Sri Lanka Consumer Finance Survey CGPSL College of General Practitioners of Sri Lanka CMMU Construction Management and Maintenance Unit COMAC-HSR Concerted Action Programme on Quality Assurance CSD Census and Statistics Department CSQ Client satisfaction questionnaire CT Computerized tomography DALY Disability-adjusted life-year DCH Diploma in Child Health DfID [United Kingdom] Department for International Development DFM Diploma in Family Medicine DG Director general DH District hospital DHS Demographic and health survey DOH Department of Health DOTS Directly Observed Treatment, Short-Course DPT Diphtheria-pertussis-tetanus ECG Electrocardiogram EIP Evidence and Information for Policy EPI Expanded program of immunization ESP Essential Services Package FCGP Fellow of the College of General Practitioners of Sri Lanka FSI Floor space index GDP Gross domestic product GH General hospital GNP Gross national product Acronyms and Abbreviations · xxiii GP General practitioner HCI Health care institution HDI Human Development Index HDS Health and Demographics Survey HEDIS Health Plan and Employer Data Information Set HES Household expenditure survey HEU Health Economics Unit HEU/DI Health Economics Unit/Data International HIPC Heavily indebted poor country HPSP Health and Population Sector Program HWG Health Watch Group ICHA International Classification of Health Accounts ICPC International Classification of Primary Care IEC Information, education, and communications IHSD Institute for Health Sector Development IIPS International Institute for Population Sciences IMC Indian Medical Council IMF International Monetary Fund IMPA Independent Medical Practitioners Association IMR Infant mortality rate IMS Indicator Measurement Systems IPS Institute of Policy Studies IPS PCS 2000 Institute of Policy Studies Private Clinic Study 2000 JCAHO Joint Commission on Accreditation of Health Care Organizations LHV Lady health volunteer MCGP Member of the College of General Practitioners of Sri Lanka MCH Medical college hospital MHSS Morbidity and health status survey MIS Management information systems xxiv · Health Policy Research in South Asia MOH Ministry of Health MOHFW Ministry of Health and Family Welfare MOS Medical Outcomes Study NACER National Council for Applied Economics Research NADHI North Arcot District Health Information NAMCS National Ambulatory Medical Care Survey NCQA National Committee for Quality Assurance NFHS National Family Health Survey NGO Nongovernmental organization NHA National health account NHE National health expenditure NIPORT National Institute of Population Research and Training NSS National Sample Survey NSSO National Sample Survey Organization OECD Organisation for Economic Co-operation and Development OPV Oral poliovirus vaccine ORS Oral rehydration solution ORT Oral rehydration therapy PAHO Pan American Health Organization PER Public expenditure review PHC Primary health center PIP Project Implementation Plan PRA Participatory rural appraisal PRSP Poverty Reduction Strategy Paper PSQ Patient satisfaction questionnaire PSU Primary sampling unit PWD Public Works Department QOST Quality of Service in Texas REACH Resource Group for Education and Advocacy for Community Health Acronyms and Abbreviations · xxv Rs. Rupees RTI Reproductive tract infection SEAR Southeast Asia region SEARO South-East Asia Regional Office SHA System of Health Accounts SLI Standard of living index SLNHA Sri Lanka national health account SLPFS98 Sri Lanka Public Facility Study 1998 SMR Standardized mortality rate STD Sexually transmitted disease STI Sexually transmitted infection TEH Total expenditures on health THC Thana health complex Tk. Takas TRC Tuberculosis Research Centre UHC Upazila health complex UHFWC Union health and family welfare center UNDP United Nations Development Programme UP Uttar Pradesh (India) USAID United States Agency for International Development WHO World Health Organization WONCA World Organization of National Colleges, Academies, and Academic Association of General Practitioners/Family Physicians SECTION I Introduction CHAPTER 1 Overview Abdo S. Yazbeck, World Bank David H. Peters, Johns Hopkins University Section I: Introduction South Asia is a region of contrasts. It covers eight sovereign nations ranging in population from 250,000 in the Maldives to 1 billion in India. The region can boast of impressive technological achieve- ments, yet it contains more than 40 percent of the world's poor, who live on less than a dollar a day. In the health sector, the con- trasts and needs are equally impressive. At one end of the spectrum, there are amazing achievements in health outcomes in Sri Lanka and in the southern Indian state of Kerala, where infant mortality rates (IMRs) are 15 and 16 deaths per 1,000 live births, respec- tively. At the other end of the spectrum, IMRs are 163 deaths per 1,000 live births in Afghanistan and 85 deaths per 1,000 live births in the northern Indian state of Uttar Pradesh. Similarly pro- nounced differences exist in other health indicators, including maternal mortality, fertility, and--to a lesser extent--nutrition (see table 1.1). Such large variations in health, nutrition, and fertility outcomes are no doubt the result of many different factors, some within and others outside what is traditionally considered the health sector. The factors outside the health sector that most likely affect these outcomes include determinants related to the performance of the economy, such as income and food security, as well as other factors, such as education, water and sanitation conditions, roads, and 3 4 · Health Policy Research in South Asia Table 1.1 Selected Health Indicators in South Asian Countries GNP INFANT TOTAL CHILDHOOD POPULATION PER CAPITA MORTALITY FERTILITY MALNUTRITION COUNTRY (2001) (2001) RATE (2000) RATE (2000) (1997­2000)a Afghanistan 27,247,944 -- 163 7 49 Bangladesh 133,405,392 370 60 3 48 Bhutan 828,044 640 58 5 19 India 1,033,389,824 460 69 3 47 Nepal 23,584,706 250 74 4 47 Pakistan 141,450,160 420 83 5 -- Sri Lanka 19,649,486 830 15 2 -- -- = Not available. Note: GNP = gross national product. a. Most recent data from 1997­2000. Source: World Bank, World Development Indicators 2002 database. social and cultural factors related to gender relations and tradi- tional beliefs. Health sector policymakers have little direct impact on such factors, but they do have considerable influence over the performance of the health sector and its effectiveness in addressing the needs of the population. When policymakers fully understand the strengths and weaknesses of the health sector, they can make adjustments and reforms that will make the sector more effective. To understand the sector fully, policymakers need access to up-to- date, accurate health policy research. It is this need for health pol- icy research that is the main focus of this book. The volume highlights the excellent research that has been conducted by South Asian institutions and researchers to support health policymaking in the region. It is clearly not feasible to cover all aspects of health policy research in one volume, so we decided to organize the book around four research areas that capture some of the best recent research efforts in South Asia: (a) inequality analysis, (b) expenditure analy- sis, (c) private sector analysis, and (d) consumer and provider per- spectives. In each research area, we selected a number of papers that covered different dimensions of the topic. For example, in the area of expenditure analysis, we have included a chapter that covers traditional public expenditure reviews, a chapter that covers the Overview · 5 sophisticated application of the national health accounts approach, and a chapter that focuses on facility-based expenditure and cost analysis. Most of the research involves survey techniques but does not rely on clinical epidemiology, demography, or anthropological methods. The inequality, expenditures, and private markets analy- ses and, to a lesser extent, the consumer and provider perspectives rely heavily on health economics and other forms of applied micro- economics. This volume does not cover basic science research and studies oriented toward specific health conditions or exposures to health risks. The driving force behind the increase in locally produced health policy research in South Asia has been a shift in how development agencies such as the World Bank, the World Health Organization (WHO), and bilateral donors finance research in the region. Increasingly, local researchers and institutions in South Asia have been contracted to conduct research meant to feed into regional needs and reform efforts. In fact, 7 of the last 12 chapters in this volume summarize studies supported by the World Bank; the United Kingdom Department for International Development (DfID), the WHO, and others supported the studies described in the remaining chapters. In choosing papers for this volume, we used the following selection criteria. First, we wanted research con- ducted by South Asian institutions or researchers. Second, we wanted innovative papers that used state-of-the-art methodologies for collecting and analyzing data. Third, we looked for papers that used research findings as a basis for policy recommendations. In other words, we chose applied practical research with a policy dimension over basic research and data collection. Finally, we wanted papers that used methods that could be tried out in other countries and thereby serve as a South-South exchange of ideas to promote further research. Not every paper in this volume meets all these criteria, but we tried to select papers that met most of them. Why focus on South Asian researchers and research institutions? This body of researchers is producing high-quality and innovative research. Included in this volume are pathbreaking pieces of work covering some of the most detailed subnational benefit incidence 6 · Health Policy Research in South Asia analysis (BIA) work to date (chapter 3), one of the best examples of applying a national health accounts approach to a low-income coun- try (chapter 7), innovative approaches in researching the ways the private sector can provide health services (chapters 10, 11, and 13), and new emphases on consumer redress (chapter 12) and enhancing the voice of citizens (chapter 14). In addition to the innovations, the book covers excellent examples of research that is directly linked to policy in the areas of equality (chapters 4 and 5), public expenditure analysis (chapter 6), cost-efficiency (chapter 8), and public-private collaboration in delivering health services (chapter 9). Local participation and ownership of research products are also important for other reasons. Locally managed research is likely to be more relevant to local policymakers and to be more closely and practically linked to politically feasible reforms and policies. In many cases, research by a local institution improves the quality of data collection and makes the findings more acceptable to policy- makers and civil society. As noted above, donors and multilateral agencies working in South Asia are increasingly contracting locally based researchers and research institutions (instead of international consultants) to do studies. By supporting local research institutions, donors are helping develop local capacity to make and to manage health policy. As health systems evolve, it is important to have the analytical capacity at the local level to monitor conditions and eval- uate the impact of reforms. International technical assistance may still be needed to share information and techniques gleaned from experience in other countries and regions of the world, but only in a supportive role. This volume is another way to share the findings of health policy research among the countries of South Asia and with countries in other parts of the world. In the remainder of this chapter, we briefly review the research presented in the four main parts of the book: inequality analysis, expenditure analysis, private sector analysis, and consumer and provider perspectives. We conclude the chapter by highlighting some of the methodological innovations represented in the volume and some of the policy-relevant outcomes that emerged from the research. Overview · 7 Section II: Inequality Analysis (Chapters 3­5) The global development community has increasingly focused on the plight of the poor and on the need not only to stress economic growth, but also to take into account the substantial inequalities in income and in socioeconomic measures such as health and educa- tion (IMF and World Bank 1999a, 1999b, 1999c, 2001; World Bank 2000). Similarly, the international health community has in recent years increasingly focused on the heath needs of the poor and on inequality both in health outcomes and in the use of health services (Claeson and others 2002; Peters and others 2002; WHO 1999, 2001; World Bank 1997). In 2000, the World Bank released the results of analytical work on health inequality (Gwatkin and others 2000) that had been done using demographic and health surveys (DHSs) financed by the U.S. Agency for International Development (USAID). Using DHS data on 44 low- and middle-income countries, Gwatkin and others (2000) took household responses to questions about assets and grouped them in wealth categories to show the differences in health, fertility, and nutritional outcomes, and service use between the rich and the poor. Four of the 44 countries analyzed were in South Asia; in all four of these countries, there were large gaps between the rich and poor in all outcomes (table 1.2). From the perspective of health sector policy, there is a qualitative difference between the inequality described in the first three rows of table 1.2 and that described in the last two rows. As discussed above, health sector outcomes, such as those in the first three rows, are determined by factors within as well as outside the health sector. While health outcomes are the ultimate objectives of a health sys- tem, policymakers have far more control over health sector outputs, such as those in the last two rows. The three chapters on health inequality research in this volume (chapters 3­5) focus on the deter- minants of inequality in health outcomes that fall within the health sector and, therefore, within the direct sphere of influence of health sector policymakers. The three chapters look at inequality in the health sector from three different perspectives. Chapter 3 is a 8 · Health Policy Research in South Asia Table 1.2 Inequality in Health Outcomes and Health Sector Outputs in South Asia (wealth quintiles) BANGLADESH INDIA NEPAL PAKISTAN 1996/97 1992/93 1996 1991 OUTCOMES AND OUTPUTS POOREST RICHEST POOREST RICHEST POOREST RICHEST POOREST RICHEST Infant mortality ratea 96 57 109 44 96 64 87 62 Child stuntingb 50 24 65 31 59 32 61 33 Fertility ratec 3.8 2.2 4.1 2.1 6.2 2.9 5.1 4 Full immunization 47 67 17 65 32 71 22 55 Attended delivery 2 30 12 79 3 34 4.6 55 aDeaths under age 12 months per 1,000 live births. bPercentage of children under age 5 years whose height for age is below ­2 standard deviation z-score. cBirths per woman ages 15­49 years. Source: Gwatkin and others 2000. detailed study of public sector subsidies for health in India as a whole and in its largest states and regions. Chapter 4 takes a national health accounts approach and compares inequality in outcomes, out- puts, and financing in three South Asian countries. Chapter 5 looks at geographic inequality in resource allocation in the health sector and suggests a different rationale for allocating resources. In chapter 3, Ajay Mahal focuses on equality in public subsidies to the health sector in India as a whole and in the 16 largest states and regions. The chapter summarizes what can be described as one of the most detailed and comprehensive benefit-incidence studies for the health sector in a low-income country (Mahal and others 2000). The work stands out in a number of ways. First, the study not only looked at the national subsidy but also disaggregated it by states and regions, thus making possible comparisons among them. Moreover, the analysis was disaggregated by gender, urban-rural residence, poverty status, and membership in a scheduled caste or tribe, recog- nized disadvantaged groups in India. Second, the study looked at the resource allocation decisions underlying the findings, such as alloca- Overview · 9 tions to the different levels of care, which have important policy implications for the future. Third, the limited availability of cost data in many Indian states led the researchers to develop innovative ways to assess the costs of services and to test two different method- ologies to ensure that the main findings were robust. Officials in the Indian government have already used the findings of the study to emphasize the need to make the health system more equitable. In chapter 4, researchers from the Institute of Policy Studies in Sri Lanka, Data International in Bangladesh, and Nepal Health Economics Association compared equality in the health sectors in three South Asian countries--Bangladesh, Nepal, and Sri Lanka-- using a national health accounts approach, among other methodolo- gies. This was the first such use of national health accounts in South Asia to compare equality among countries. Another important methodological element of the study was the use of state-of-the-art techniques such as concentration indexes for measuring inequality. The study covered a wider range of the different dimensions of equality than has been attempted in any other studies of equality in the health sector. The study looked first at inequality in health sec- tor outcomes (infant and child mortality and reporting of illnesses), then turned to inequality in heath sector outputs (use of health care services) and in benefits from government subsidies. It went on to look at the distribution of payments for health services and finally studied the redistributional nature of the different ways in which the governments of the three countries finance health services. The last chapter in section II--chapter 5--provides a natural transition to section III on expenditure analysis. Chapter 5, written by the team at the Health Economics Unit of the Ministry of Health and Family Welfare in Bangladesh, is an extension of a tradi- tional public expenditure review (PER) with a focus on equality in geographical resource allocation. Two features in chapter 5 provide valuable lessons for researchers and policymakers interested in equality in public expenditure. First, the chapter highlights the need to go beyond the basic and superficial description of budget and expenditure allocations to look carefully at the mechanisms under- lying historic allocation decisions. Second, the chapter introduces 10 · Health Policy Research in South Asia the concept of needs-based resource allocation to replace or correct the facility-based approach used in the health sector in Bangladesh and in most other countries in South Asia and elsewhere. Section III: Expenditure Analysis (Chapters 6­8) The simplest way to describe the essence of expenditure analysis in the health or any other sector is that it is an exercise that involves following the money. This seemingly simple exercise is especially important in the health sectors in South Asian countries. According to the global national health accounts maintained by the WHO, South Asian countries stand out in terms of having a weak public financial commitment to the health sector and a heavy reliance on private out-of-pocket spending. Five South Asian countries--India, Pakistan, Nepal, Afghanistan, and Sri Lanka--are ranked among the lowest in the world in terms of national public sector spending on health as a percentage of gross domestic product (GDP). In another measure of the lack of public commitment--the share of total health spending by private sources--India, Pakistan, and Nepal are in the top 20 countries (out of 190) in which private spending accounts for more than 75 percent of all health sector spending. The low levels of public spending on health in South Asia make it especially critical to determine the efficiency and equality of this expenditure. The three chapters in section II look at the equality of public allocations, while the three chapters in section III look at different dimensions of the efficiency of public and private spend- ing. Chapter 6 takes a traditional approach to PERs and links the objectives of health programs in Bangladesh with public expendi- tures on them. Chapter 7 uses a national health accounts method- ology to analyze private health expenditures in Sri Lanka. Chapter 8 focuses on the dominant element of public spending--facility- based expenditures--in Bangladesh and uses a facility survey to look at the cost-efficiency of the allocation decisions. These three chapters take different approaches, but they all highlight the Overview · 11 importance of analyzing expenditures in the health sector, and they all provide an empirical basis for policy options. Chapter 6, which was also written by the Health Economics Unit of the Ministry of Health and Family Welfare in Bangladesh, is an excellent example of how basic PERs in the health sector can support the implementation of programs and provide empirical inputs into the policy process. The chapter highlights the link between the PER exercise carried out by the Health Economics Unit and the main elements of the reform in the health sector in Bangladesh. The PER looks at the extent to which the govern- ment's policy of targeting the services most often used by the poor achieved its stated objective of reaching low-income groups. It also looks carefully at the government's expenditure allocations within the package of priority services. In addition to addressing country- specific policy issues, chapter 6 highlights the basic elements of a PER, including (a) looking at overall expenditures in the sector and the extent to which they are consistent with the original budget fig- ures; (b) reviewing the breakdown of expenditures by programs, levels of care, and inputs; (c) analyzing domestic sources of financ- ing as well as external support; (d) analyzing the overall resource envelope; (e) analyzing the equality of resource allocation; and (f) examining the budget processes that drive allocation decisions. In chapter 7, which was written by the Institute of Policy Stud- ies and the Ministry of Health in Sri Lanka, the authors used national health accounts techniques to collect and report on all health sector expenditures in Sri Lanka. While national health accounts are becoming standard exercises in member countries of the Organisation for Economic Co-operation and Development (OECD), very few developing countries have invested in collect- ing the data needed to conduct and maintain them. The Sri Lanka accounts summarized in chapter 7 are a way for low-income coun- tries to expand basic expenditure analysis in the health sector. Chapter 7 highlights the main distinguishing feature and strength of national health accounts, which is that they capture both private and public expenditures in the health sector. Without this feature, policymakers are working with an incomplete picture, typically 12 · Health Policy Research in South Asia consisting only of the government's expenditures. The authors summarize a number of important outcomes of the Sri Lanka national health accounts, including analyses of (a) trends in total expenditures, (b) sources of financing, (c) public and private expenditures disaggregated by function and level of care, (d) types of providers, (e) allocations of expenditures among geographic areas, and (f) international comparisons. Chapter 8 looks at efficiency in the health sector in Bangladesh using facility-based data. In this expenditure study, Ravi Rannan- Eliya and Aparnaa Somanathan challenge the long-standing assumption that lower-level facilities deliver health care at lower costs. They conducted a relatively low-cost data collection effort. The data from the facility survey allowed the researchers to go beyond the basic expenditure data reported in the government sys- tem. These data made it possible for them to directly observe inputs such as the availability and functioning of basic equipment in different departments in hospitals and clinics, the availability and functioning of basic utilities needed for equipment to function, staff availability and classification in facilities, time allocation pat- terns of providers by classification, use of services by type of ser- vice, and performance indicators such as bed occupancy rates. The survey and the budget data together allowed the researchers to cal- culate average-cost numbers for the different levels of care and to point out suboptimal investments to policymakers. Section IV: Private Sector Analysis (Chapters 9­11) Health policymakers and researchers have neglected the private sec- tor in South Asia for a long time. Nonetheless, the private sector is a large and growing part of South Asia's health systems, in terms of both financing and providing health services. Several chapters in this volume highlight the dependence of national health systems in South Asia on private financing for health services--49 percent of all health spending in Sri Lanka, 64 percent in Bangladesh, 77 per- cent in Nepal, and 82 percent in India (WHO 2001). Overview · 13 In addition to the large role that the private sector plays in financing health care, it provides a significant proportion of health services in low-income countries (Berman and Rose 1996). For example, in a comparison of DHS results in 38 countries, Gwatkin and others (2000) found that, on average, 77 percent of childhood diarrhea treatments were provided by the private sector, with even higher figures in Bangladesh (92 percent), Nepal (90 percent), and India (84 percent). The private sector is responsible for the largest segment of the health care market in South Asia, though there are important differences between countries and types of services. In chapter 9, V. R. Muraleedharan and Sunil Nandraj point out that the most numerous type of providers in India (about 1 million) are private, unqualified medical practitioners who largely cater to the rural poor. They also note that 70 percent of the country's quali- fied medical doctors work in the private sector, largely in urban areas. Quoting the National Sample Survey Organisation (NSSO 1998), the authors demonstrate that the private sector portion of both inpatient and outpatient services is growing, with more than 80 percent of outpatient visits and more than half of hospitaliza- tions now occurring in the private sector. At the other end of the spectrum, Rannan-Eliya and his colleagues (chapter 11) note that 45 percent of outpatient visits in Sri Lanka are to private general practitioners. Despite the growing role of the private sector in financing and providing health care in South Asia, many governments in the region view the private sector with suspicion and are not convinced of its appropriateness, quality, or cost-effectiveness. As a result, many governments have yet to decide how to approach the ques- tion of the private sector's role in health care. However, in order to make informed policy decisions and develop strategies to deal with the private sector, governments will need much more information about its nature, composition, activities, opportunities for collabo- ration, and areas of failure. At this stage, most of the analysis that has been done is descriptive in nature, meaning that it simply iden- tifies the size and location of different parts of the private sector and documents experiences. The research in this volume extends 14 · Health Policy Research in South Asia this body of knowledge a stage further by examining how the pri- vate sector is actually functioning in South Asia and analyzing what options governments have to deal with its various components. The authors of these chapters examine various dimensions of private sector health care in South Asia, which we define as all providers and suppliers of health inputs outside the government sector, including both for-profit organizations and nonprofit non- governmental organizations (NGOs). Private health providers themselves consist of a disparate group of traditional healers, drug vendors, shopkeepers, and unqualified practitioners, as well as pro- fessional pharmacists, laboratory technicians, nurses, and doctors. However, the distinction between public and private is not com- plete. The authors from India (chapters 9 and 10) and Sri Lanka (chapter 11) point out how public sector practitioners also practice privately, in some cases legally (in Sri Lanka) and in other cases illegally (in Uttar Pradesh, India). The chapters in this section shed light on how different parts of the private sector operate in India and Sri Lanka, using innovative analytical methodologies that yield new insights on the subject. In chapter 9, Muraleedharan and Nandraj synthesize the literature on private health care provision in India as a basis for a detailed expo- sition of government options for dealing with private providers. Recognizing the limited capacity of governments to monitor their contracts and regulate the private sector, they point to six specific ways in which governments can collaborate with the private sector to reach a set of policy objectives, including increased access, safety, and accountability of all health services. In his study of private health provision in India's largest state, Uttar Pradesh, (chapter 10), S. Chakraborty has conducted an innovative survey to assess how different parts of the health- provider market function--qualified and unqualified medical prac- titioners, small and large hospitals, and diagnostic centers. In addition to reporting hard-to-find empirical data on the motiva- tions, quality of care, and pricing practices of these private markets, he has identified the constraints that prevent different types of private providers from providing quality services and from Overview · 15 collaborating with the government. Chapter 11 reports on another empirical study of private practice behavior, this time a study of general practitioners in Sri Lanka. For the first time, Rannan- Eliya and his colleagues have been able to adapt survey methods used for private practitioners in Australia and the United States to assess the quality of care in a developing country using a nationally representative sample of private general practitioners. They pro- vide new quantitative information about the different levels of services provided by private practitioners, how much they charge, and how they operate. The authors conclude that private practi- tioners in Sri Lanka provide a substantial volume of good-quality clinical care, thus reducing the burden on government services. The practitioners themselves are concerned about the lack of vocational training available to them, weak referral mechanisms between the public and private sectors, and inadequate financing methods to ensure that future doctors will have incentives to go into private practice. Section V: Consumer and Provider Perspectives (Chapters 12­14) The views of the users of health services and providers have not been well researched in South Asia. Each of the three chapters in this section uses a different approach and brings a unique perspec- tive to this subject. In chapter 12, Bejon Misra reports on a study that looked specifically at consumer redress in public and private hospitals in India, a subject that is rarely examined. He conducted surveys both at health facilities and at the Consumer Courts, which are special forums that were established under the Consumer Pro- tection Act to deal with consumer complaints efficiently. The study found that hospitals themselves are not well equipped to deal with consumer complaints, most of which tend to be concerned with nonclinical issues such as sanitation, availability of water and power, and billing procedures. Unfortunately, the study found that the Consumer Courts are very inefficient and not very accessible, which leaves both consumers and defendants dissatisfied. As a 16 · Health Policy Research in South Asia result of these investigations, Misra is able to propose a wide vari- ety of measures that would make health services more accountable and provide genuine means of redress to consumers. In chapter 13, Prasanta Mahapatra takes another approach to examining the quality of health services in Andhra Pradesh, India. Defining "quality of care" in terms of health attainment (technical quality) and responsiveness (interpersonal quality), Mahapatra's group adapted research instruments from other countries to docu- ment the physical structures and clinical practices of public and private health providers and patient satisfaction with them both. In assessing the evidence from his study, Mahapatra concludes that technical quality may be slightly better in the public sector facilities that he studied than in the private sector facilities, but the private sector appears to provide better interpersonal quality of care. More important, he highlights the facts that quality-related practices are poorly developed in both the public and private sectors and that there is insufficient orientation toward clients throughout the entire health sector in Andhra Pradesh. Chapter 14 raises the issue of how the voices of consumers and other stakeholders can be heard in the process of health sector reform in Bangladesh. Nilufar Ahmad details an extensive consul- tation process among 34 stakeholder groups in the preparation of Bangladesh's Health and Population Sector Program that began in 1998. The process showed how consultation could be used to reach a consensus about which programs should be prioritized and which vulnerable groups, notably women and children, should be tar- geted. Through the participatory process, it became clear that a majority of Bangladeshi citizens are not aware of their right to receive health care and, therefore, are not taking advantage of the high-quality services that already exist. Partly influenced by the surveys that were conducted as part of the Health and Population Sector Program, the government has adopted a Patients' Rights Charter and established a National Steering Committee for stake- holder participation within the Ministry of Health and Family Welfare, consisting of government officials, women's organizations, and NGOs. Overview · 17 Methodologies and Implications As mentioned above, a major purpose of this volume is to share innovative research methodologies being used in South Asia to provide the empirical basis for health sector reform and policy development. In this section we highlight two elements of the methodologies used in chapters 3­14 (data collection and analysis techniques) and some of the emerging themes. Data Collection Efforts Given the nature of this volume (collection of research articles), it is not surprising to find a wide range of methodologies and associ- ated data collection efforts used. Two areas stand out in terms of innovation and effectiveness: flexible facility-based surveys and qualitative participatory techniques. A number of chapters used facility-based surveys to both measure performance and understand the constraints faced by the health sector. Chapter 8 used a facility survey to measure unit cost of service delivery at four different lev- els of care in Bangladesh. The survey allowed the researchers to pinpoint inefficiencies and the policies that caused them. Chapters 10 and 13 used facility surveys to look at multiple dimensions of the private health sector in two different states in India. These sur- veys helped, for the first time in India, to better define the size and makeup of the private sector (including the informal sector), mea- sure the quality of care of privately and publicly provided services, capture the perspective of providers in the private sector, and explore potential forms of public-private collaboration. The facility survey in chapter 11 successfully applied techniques from industrial countries to Sri Lanka. Since the survey was nationally representative, it allowed for estimation of the real con- tribution of the private sector to health care delivery in terms of volume of service, type of services, and cost of services. Moreover, the survey looked at the profile of providers and explored policy issues. And finally, the facility survey in chapter 12 captured critical dimensions of consumer redress in the health sector in three cities in India. 18 · Health Policy Research in South Asia While quantitative surveys for households and facilities can cap- ture what is happening in the health sector, they are limited in terms of reaching a deep understanding of why services are poor. Chapter 14 highlights the different ways qualitative techniques can be used. Participatory rural appraisal techniques are used for clients (poor rural women), community leaders, providers, and profes- sional associations to diagnose problems in the health sector in Bangladesh and to identify solutions and support reforms. A differ- ent set of qualitative tools were used in chapter 12 to supplement a facilities survey. Legal case reviews and key informant interviews were used to assess the consumer redress system. Crosscutting Implications and Emerging Themes The richness of the research papers in this volume makes it diffi- cult to quickly capture the main themes and implications of their research. But three repeated themes can be highlighted: equality of public spending, the role of the private sector, and the role of con- sumers. On the theme of equality in public expenditures, research in Bangladesh, India, Nepal, and Sri Lanka shows that in some parts of South Asia--such as south India and Sri Lanka--govern- ments do a much better job of distributing subsidies in the health sector than other regions. The different chapters that tackled this issue (chapters 3­7, 12, and 14) not only measure the inequalities, but also suggest different policy tools that include establishing geo- graphic allocations formulas, reallocating resources between differ- ent levels of care, educating vulnerable groups about their rights, introducing community-based monitoring, and strengthening redress mechanisms. Another repeated theme relates to the role of the private sector in health services delivery and finance in South Asia (chapters 4, 7, 9­11, and 13). The research overwhelmingly documents the domi- nance of the private sector in Bangladesh and India and finds a very strong private sector in Sri Lanka. The research also highlights dif- ferent policy instruments available to the government for working with the private sector to achieve health sector outcomes. Overview · 19 A third general theme is the role of consumers and the mecha- nisms available to them to influence health services delivery. The authors in this volume have supported the belief that individuals and households can make a difference in how health services are delivered. By documenting the limited ability of patients and households to influence health services, they propose a number of policy recommendations, which include improving consumer edu- cation, developing and disseminating a patient bill of rights, build- ing community watch committees to oversee service delivery, and improving consumer redress mechanisms. Underlying these policy recommendations is the need to strengthen the links between poli- cymakers and potential health care consumers, as well as to change the nature of the relationship between consumers and providers in the private and public sectors. While the three themes summarized above cut across several of the chapters in this volume, a more basic theme underlies all the chapters and is the main motivation for conducting health policy research. That theme is that empirical research can and should challenge basic assumptions about the health sector and will pro- vide policymakers some of the tools needed to improve and moni- tor the performance of the sector. This volume includes a number of clear applications of this theme, but two examples can make the point. In chapter 3, Mahal takes on the assumption that health spending in India is pro-poor and shows that the overwhelming majority of states in India have a strong pro-rich orientation. In chapter 8, Rannan-Eliya and Somanathan challenge the assump- tion that some of the lower-level health care facilities in Bangladesh are efficient by measuring the unit cost of delivery and linking the results to the level of utilization and staffing patterns of these facilities. Such findings about the unintended results in the performance of their health systems should motivate policymakers and planners to change course. By conducting and disseminating the results of their work, the South Asian institutions and researchers represented in this volume have the opportunity to provide valuable support for informed policymaking in the health sector. 20 · Health Policy Research in South Asia References Berman, P., and L. Rose. 1996. "The Role of Private Providers in Maternal and Child Health and Family Planning Services in Developing Countries." Health Policy and Planning 11: 142­55. Claeson, M., C. G. Griffin, T. A. Johnston, M. McLachlan, A. Soucat, A. Wagstaff, and A. S. Yazbeck. 2002. "Health, Nutrition, and Population." In A Sourcebook for Poverty Reduction Strategies, Vol. 2. Washington, D.C.: World Bank. Gwatkin, Davidson, Shea Rutstein, Kiersten Johnson, Rohini P. Pande, and Adam Wagstaff. 2000. Socioeconomic Differences in Health, Nutrition, and Population in India. HNP Publication Series, HNP/Poverty Thematic Group. Washington, D.C.: World Bank. IMF (International Monetary Fund) and World Bank. 1999a. "Poverty Reduction Strategy Papers--Operational Issues." Draft for discussion (poverty. worldbank.org/files/operational_issues.pdf). Washington, D.C. ------. 1999b. "Building Poverty Reduction Strategies in Developing Countries" (poverty.worldbank.org/files/building_english.pdf). Washington, D.C. ------. 1999c. "Heavily Indebted Poor Countries (HIPC) Initiative--Strengthen- ing the Link between Debt Relief and Poverty Reduction" (poverty. worldbank.org/files/strengthening_link_english.pdf). Washington, D.C. ------. 2001. "Poverty Reduction Strategy Papers--Progress in Implementation 2001" (poverty.worldbank.org/files/prspprogess.pdf). Washington, D.C. Mahal, Ajay, Janmejaya Singh, Farzana Afridi, Vikram Lamba, Anil Gumber, and V. Selvaraju. 2000. Who Benefits from Public Health Spending in India? New Delhi: National Council of Applied Economic Research. NSSO (National Sample Survey Organisation). 1998. Morbidity and Treatment of Ailments. NSS 52nd round, July 1995­June 1996. Peters, D. H., A. S. Yazbeck, R. R. Sharma, G. N. V. Ramana, L. H. Pritchett, and A. Wagstaff. 2002. Better Health Systems for India's Poor: Findings, Analysis, and Options. Washington, D.C.: World Bank. WHO (World Health Organization). 1999. World Health Report 1999: Making a Difference. Geneva. ------. 2001. World Health Report 2001: Mental Health: New Understandings, New Hope. Geneva. Overview · 21 World Bank. 1997. Health, Nutrition, and Population Sector Strategy. Washington, D.C. ------. 2000. "Poverty Reduction Strategy Papers--Internal Guidance Note 2000" (www.worldbank.org/poverty/strategies/intguid.pdf). ------. 2001. "Poverty Reduction Strategy Papers Sourcebook." Preface. Draft for Comments (www.worldbank.org/poverty/strategies/chapters/preface/pref0618. pdf). CHAPTER 2 A Framework for Health Policy Research in South Asia David H. Peters, Johns Hopkins University Abdo S. Yazbeck, World Bank The research presented in this volume offers a highly selective view of health research that should be conducted in South Asia. Each of the chapters deals with policy issues that have tended to be ignored in the past, and each represents a study that was conducted by South Asian institutions (some of the authors have since moved). The chapters are oriented toward providing the information needed to develop policies and strategies to improve the perform- ance of health systems in South Asian countries. To put this research in context, it should be recognized that health systems research is only one of several priority areas of health research in South Asia. The Ad Hoc Committee on Health Research Relating to Future Intervention Options (1996) high- lighted how knowledge obtained through research is responsible for much of the health gains made during the 20th century. Yet com- mitment to health research has been declining and has concentrated on health concerns of the industrial world. Arguing that collective action on health research is a public good and is needed to address global health and development challenges, the committee proposed a broader framework of health research priorities, organized around 23 24 · Health Policy Research in South Asia four main challenges, including the type of health systems research presented here: 1. The unfinished agenda for control of childhood illness, undernu- trition, and excess fertility 2. The challenge of changing and emerging major microbial threats 3. The emergence of neglected epidemics of noncommunicable dis- ease and injury 4. The inefficiency and inequality in health systems The Global Forum for Health Research (2000) further high- lighted the need to build research capacity in developing countries, to correct what it called "the 10/90 gap": only 10 percent of the health research funding is allocated to 90 percent of the disease burden facing the world. In developing systems to track resource flows into health research, the forum refined its taxonomy of health research into five categories (Global Forum for Health Research 2001): 1. Nonoriented, fundamental research 2. Health conditions (classified by disease) 3. Exposures and risk factors that affect health (determinants) 4. Health systems research 5. Research capacity building The studies in this volume would be classified as health systems research (category 4), particularly because the research deals with health policy and planning and health services delivery. But such classifications involve some overlap and ambiguity, as the research in this volume is also related to other categories. For example, many of the factors studied in this volume, such as the allocation of health resources or the quality of care, can also be considered as risk factors that affect health (category 3). Although not examined directly in this volume, health systems research is also concerned A Framework for Health Policy Research in South Asia · 25 with health conditions, as these are the eventual outcomes of health systems. The intention of the research in this volume was also to develop research and policy capacity in local institutions through involve- ment in the selection, design, conduct, and application of health systems research (category 5). Interestingly, none of the researchers was solicited through a competitive process. This was partly because only a limited number of local agencies are capable of doing such work, and partly because of the time constraints imposed on the research project by both policymakers and funding agencies. However, in the long term, as more institutions become capable of undertaking research and as the type of research meth- ods change to more experimental and longitudinal approaches, there will be greater tension between the desirability of competi- tive means for soliciting research and the need to encourage greater collaboration among researchers. Given that this research is intended to examine health systems, it is useful to have a common understanding of what is meant by a health system. Numerous definitions and frameworks have recently been offered as descriptions of health systems (Berman 1995; Cas- sels 1995; Frenk 1994; OECD 1992; Roemer 1991; WHO 2000). They all involve a purpose for the health sector (for example, improving people's health status), a set of actors on the demand side (for example, individuals, households) and the supply side (service providers, financiers), intermediating actors (regulators, institu- tional purchasers), and functions of a health system (service delivery, financing, human resources, and other inputs). Since the research in this volume is particularly interested in the application of research to policy, it is useful to examine what Roberts and others (2003) have described as five sometimes overlapping control knobs that policymakers use to improve the performance of a health system. 1. Financing. Financing interventions include all the mechanisms used to raise money for health sector activities, including taxes, insur- ance premiums, and direct out-of-pocket payments by patients. Each of the six chapters in this book dealing with equity and expenditure 26 · Health Policy Research in South Asia analysis (chapters 3­8) addresses financing concerns. Their common findings are that public financing levels are low and inequitable and that there is a high level of dependence on direct private payments from individuals, which is greatest in India and lowest in Sri Lanka. 2. Payment. The payment control knob involves the mechanisms that are used to pay health care providers, including doctors, para- medical workers, and hospitals and other health institutions. The main ways to pay individual providers are through fee-for-service, salary, and capitation payments, while institutions are generally paid through budgets, grants, or service payments. Providers respond to the different incentive systems these mechanisms cre- ate. Studies in this volume provide evidence of low efficiency of public facilities in Bangladesh based on their fixed budgets and rigid staffing norms (see chapter 8); yet fee-for-service payment mechanisms result in overproduction of services in the private sec- tor (chapters 9­11). 3. Organization. This control knob involves determining which organizations will be able to perform which functions (for example, deliver certain health services, enforce regulations, or provide insurance), as well as how they are able to perform these tasks (for example, the type of staff involved) and how they are rewarded. Studies in this volume demonstrate how public and private providers share responsibilities (chapters 9­11) and how equitable and efficient they are (chapters 3­5 and 8). 4. Regulation. These are the instruments by which the state places requirements on enterprises, citizens, and government itself. One aspect involves writing laws, orders, and other rules; another involves implementation through enforcement. Regulation covers all the state's efforts to influence the behavior of those in the health system, including providers, insurance companies, and patients. 5. Persuasion. This policy control knob applies to efforts to change the behavior of the public and providers. The research in A Framework for Health Policy Research in South Asia · 27 this volume does not analyze the main channels of communication used to persuade people, such as the mass media or encourage- ment of peers, but it does look at how involvement of various stakeholders can influence policy (chapter 14) and how parts of the legal system can be used (chapter 12). In chapters 13 and 14, the researchers also point out the role of consumer groups in pro- viding information on health practices as a means of improving quality. The research in this volume is intended to provide infor- mation that will persuade policymakers to improve how health systems perform. Although health systems research is intended to influence how these policy control knobs are used, in practice such research has often fallen short in delivering changes (Getting Research into Pol- icy and Practice 2003). Public policy is frequently rooted in politi- cal values and is often based upon expert opinion and the collective judgment of lawmakers and officials. Research may be discarded because it is seen as irrelevant, too complex, untimely, or anathema to the ideologies of policymakers or to the policy processes. We believe policy should be informed as much as possible by evidence, such as that provided by the researchers in this volume. Following the findings of Crewe and Young (2002), we argue that quality research, local involvement of researchers and policymakers, and effective use of communication channels are important determi- nants of whether research will influence policymaking. Many of the research studies in this volume were generated out of locally defined priorities. Public discussions with key stakeholders identified each topic as a priority. Many of the research projects involved collaboration with national policymakers and interested local and international agencies during the design and dissemination phases; researchers were often brought together for consultations during the analysis, design, and dissemination phases. According to Landry, Amara, and Lamari (1999), the biggest hurdle in utilizing research knowledge is the failure of researchers to transmit their findings. Local institutions can improve policy decisions by provid- ing timely and relevant evidence from researchers who are on the spot and, thus, can communicate directly with local policymakers 28 · Health Policy Research in South Asia and other stakeholders. We hope that the local nature of this research makes it relevant not only to local stakeholders, but also to international development agencies. International agencies can play an important role in supporting such research by encouraging more local engagement in conducting research and transmitting findings. International agencies can also bring together people, ideas, and experience from a broad range of countries in a collaborative approach that supports the legitimacy of research and policy net- works of developing countries, rather than simply pursuing exter- nally driven research agendas. The research presented in this book will contribute to health sector development by providing more opportunity to transmit the findings of locally relevant quality research to policymakers in South Asia. Having pointed out the selective nature of the health systems research of this volume and the intention to build research capacity to strengthen evidence-based policymaking, it is also important to recognize the areas that require further study in South Asia. Fur- ther research in health financing is sorely needed. With high dependence on out-of-pocket payments in South Asia, developing mechanisms to pool risks is clearly a priority area for future research. The studies in this volume address only a small part of the issues related to persuasion and regulation. Much more work is needed on the use of these mechanisms to influence the behavior of providers and consumers of health services. Further study is also needed in the area of testing different mechanisms for paying providers. Indeed, understanding and manipulating incentives for health service providers is a largely neglected area. Given that health services are an intensely human resource­based activity, much further work in human resources is merited. Organizational issues that deserve additional attention include a better understanding of the private health sector and how to use the tools to influence private providers' behavior (Mills and others 2002; Waters, Hatt, and Peters 2003). This is particularly challeng- ing in a South Asian context, where many of the private providers are not formally recognized and operate in a gray economy. At the A Framework for Health Policy Research in South Asia · 29 same time, it is also important to build on the studies that examine how to improve the efficiency and accountability of the public sec- tor and to look for new ways to reduce the rigid and bureaucratic approaches to delivering health services in the public sector. It is also imperative to further develop the research methods used in health systems research in South Asia. The studies in this volume are descriptive in nature, with analysis based on careful observation of current and past situations. We hope that future research will test experimental or quasi-experimental methodologies based on insights gained from the types of research found in this volume. For example, studies that examine strategies to influence the private sec- tor through randomized community trials would provide much- needed evidence about what can work to reform a sector that is critical to health care in South Asia (Waters, Hatt, and Peters 2003). Such research will not only provide more robust evidence about the effectiveness of health strategies, but also will build a broader base of experience with the various actors and interventions. By encour- aging local researchers and networks with policymakers, by pursu- ing the neglected issues that are important to the development of health systems in South Asia, and by building on the methods of research represented in this volume to more experimental approaches, it is our intention to improve the quality of health poli- cies and decisions so that health systems will be better equipped to enhance the health and welfare of the citizens of South Asia. References Ad Hoc Committee on Health Research Relating to Future Intervention Options. 1996. Investing in Health Research and Development. TDR/GEN/96.1. Geneva: World Health Organization. Berman, P. 1995. "Health Sector Reform: Making Health Development Sustain- able." In P. Berman, ed., Health Sector Reform in Developing Countries. Boston: Harvard School of Public Health. Cassels, A. 1995. "Health Sector Reform: Key Issues in Less Developed Coun- tries." Journal of International Health Development 7: 329­49. 30 · Health Policy Research in South Asia Crewe, E., and J. Young. 2002. Bridging Research and Policy: Context, Evidence, and Links. Working Paper No. 173. London: Overseas Development Institute. Frenk, J. 1994. "Dimensions of Health Sector Reform." Health Policy 27: 19­34. Getting Research into Practice and Policy. 2003. (http://www.grip-resources.org, accessed March 16, 2003). Global Forum for Health Research. 2000. The 10/90 Report on Health Research 2000. Geneva. ------. 2001. Monitoring Financial Flows for Health Research. Geneva. Landry, R., N. Amara, and M. Lamari. 1999. "Climbing the Ladder of Research Utilization: Evidence from Social Science Research." Paper presented at the Society for Social Studies of Science, San Diego. Mills, Anne, Ruairi Brugha, Kara Hanson, and Barbara McPake. 2002. "What Can Be Done about the Private Health Sector in Low-Income Countries?" Bulletin of the World Health Organization 80: 325­30. OECD (Organisation for Economic Co-operation and Development). 1992. The Reform of Health Care: A Comparative Analysis of Seven OECD Countries. Paris. Roberts, M. J., W. Hsiao, P. Berman, and M. R. Reich. 2003. Getting Health Reform Right: Improving Performance and Equity. New York: Oxford University Press. Roemer, M. I. 1991. National Health Systems of the World. 2 Vols. New York: Oxford University Press. Waters, H., L. Hatt, and D. Peters. 2003. "Working with the Private Sector for Child Health." Health Policy and Planning 18(2): 127­37. WHO (World Health Organization). 2000. World Health Report 2000--Health Sys- tems: Improving Performance. Geneva. SECTION II Analysis of Inequality CHAPTER 3 The Distribution of Public Health Subsidies in India Ajay Mahal, Harvard University Abstract The chapter assesses the distribution in India of public sub- sidies for health care across different groups, classified by per capita expenditure, using the benefit incidence analysis method. First, the analysis finds that health subsidies are not particularly well targeted to the poor in India, espe- cially among those living in rural areas and in the poorer states. Second, the distribution of health subsidies is primarily driven by the magnitude of subsidies and utiliza- tion patterns related to hospital-based care. The distribu- tion of subsidies for primary care and for several services associated with maternal and child health (pre- and postna- tal care, immunizations, and the like) is targeted more effectively than curative care. Third, the unequal distribu- tion of subsidies for inpatient stays may be consistent with the public sector performing a key role of insuring poorer patients against expensive illness episodes. This is sug- gested by the observation that poorer patients and poorer states use relatively more of publicly provided hospital services than private hospital services, compared with their 33 34 · Health Policy Research in South Asia richer counterparts. However, meeting the insurance objectives appears to involve a tradeoff: although there is greater insurance coverage for the poor, large amounts of public subsidies go to the rich, especially those living in rural areas. The findings of this chapter lend further empir- ical weight to the association of the distribution of public subsidies for health care with the problem of physical access to care. Distance seems to disproportionately affect the utilization behavior of the poor in India, as suggested by the distribution of inpatient days in public hospitals among residents of rural areas and the much more equi- table allocation of subsidies among urban respondents. Availability of private care also appears to influence the dis- tribution of subsidies. Introduction In India, government spending on health is somewhat lower than the world average but is still quite significant, accounting for 0.9 percent of gross domestic product (GDP) in 1999­2000. In 1995­96, the last year in which a detailed survey of household health care spending was undertaken, government spending accounted for less than 20 percent of all health spending, both pub- lic and private. Nearly 90 percent of public spending is routed through the state governments, because the Indian Constitution specifies that a large number of health-related activities should be run by the states (India 1996; Reddy and Selvaraju 1994). The cen- tral government spends most of the rest while a small portion is spent by local bodies (World Bank 1995a). Various key policy documents over the years have expressed con- cern about the equity of health care provision in India. For instance, a Government of India committee report of 1946 recom- mended that "no individual should fail to secure adequate medical care because of the inability to pay for it" (Lok Sabha 1985, p. 3). The Distribution of Public Health Subsidies in India · 35 India is also a signatory to the Alma Ata Declaration of 1978, which aspired to the goal of "Health for All" (our italics) by the year 2000 (Lok Sabha 1985, pp. 1­2). The government's 1983 National Health Policy document for India supported "planned efforts . . . to provide adequate care and treatment to those entitled to free care . . . to remove existing regional imbalances and to pro- vide services within the reach of all, whether residing in the rural or urban areas" (Lok Sabha 1985, p. 41). Finally, the Indian Con- stitution, both in its Preamble and in its enunciation of Fundamen- tal Rights and Directive Principles, repeatedly emphasizes the need to create a more equal society (India 1996). Given the obvious importance of the equitable allocation of health facilities in government policy in India, it is necessary to evaluate the extent to which government programs are achieving this goal. There are two additional reasons for carrying out such an assessment. First, evidence from other countries that have reached a level of economic development similar to India's suggests that government participation in the financing and provision of health care does not always promote the objectives of equity (Castro-Leal and others 1999; Meesook 1984; van de Walle 1992). Second, such an analysis might yield some insights into what policy changes might be needed. For instance, if the analysis were to show that spending on preventive care is more equally distributed than spending on curative care, it might be appropriate to reassign some resources from the latter to the former, especially if preventive expenditures are cost-effective. Furthermore, if the analysis were to show that the poor rely more on subsidized, public curative care than on privately provided care but that the nonpoor benefit dis- proportionately from public subsidies, this would indicate the need for better targeting of publicly funded care. In line with the rationale above, this chapter analyzes the distri- bution of public sector health spending in India, net of user charge revenues, using an approach referred to as benefit incidence analy- sis (BIA) (Selden and Wasylenko 1992). Because we focus on public expenditures, we will not examine the equity implications of how 36 · Health Policy Research in South Asia these expenditures are financed, because that would require an entire separate study. Only about 20 percent of tax revenues in India are raised through direct taxes, with the balance coming from indirect taxes such as sales taxes and excise and customs duties, which are more likely to be paid by less well-off sections of society (India 1999). Methodology and Data This section covers four topics: benefit incidence analysis, the household survey data used, the unit cost calculation, and the con- struction of inequality indexes. Benefit Incidence Analysis BIA involves four steps. The first is to rank all individuals accord- ing to an appropriate measure of their socioeconomic status, such as current income per capita or consumption expenditure per capita. The particular measure that analysts use for the ranking depends on which measure policymakers regard as most impor- tant. Planners and policymakers in India have put a high priority on raising incomes and reducing poverty, so these were natural candidates for our ranking devices (India 1993, 1999). The second step is to match each individual with the amount of public health services that individual uses. The third step is to estimate the net per-unit cost of service provision to the government and to multi- ply it by the number of units of publicly provided care each indi- vidual uses. The standard approach taken in the literature is to use the average cost per unit of providing the service minus any fees paid to the government for that service (Castro-Leal and others 1999). Using the average cost per unit of service also enables us to carry out a full analysis of how all the health subsidies provided by the government are allocated to people across the socioeconomic spectrum. The final step is to analyze the distribution of net gov- ernment health spending by the socioeconomic categories of the beneficiaries. The Distribution of Public Health Subsidies in India · 37 Organizing a BIA involves a number of conceptual issues. They include choosing between individuals and households as the unit of analysis, choosing between income per capita and expenditure per capita as the indicator of socioeconomic status, and interpreting the welfare implications of the results (Selden and Wasylenko 1992). In this chapter, we use per capita expenditures as the ranking device and the individual as the unit of analysis. Household Health Survey Data BIA relies on two major sources of data--utilization of health ser- vices and out-of-pocket spending on health care. Both kinds of data are usually available from household surveys in developing coun- tries. A major source of our data was the 52nd round of the National Sample Survey (NSS) in India that was fielded in 1995­96, the most recent year for which health-related data are available for a sufficiently large sample of Indian households. The survey sampled nearly 121,000 households--71,300 in rural areas and 49,700 in urban areas. The sample was selected from 7,663 vil- lages and 4,991 urban blocks that together represented all of the states and union territories of India. We used the survey data to obtain information on a number of key variables at the individual level. These included illness pat- terns, consumption expenditures, out-of-pocket spending on health, hospitalizations in the year prior to the survey, outpatient treatments over a reference period of two weeks, immunizations received by children, care received at the time of childbirth, and pre- and postnatal visits to clinics by women who had given birth in the previous year. The sampling process by which data were collected is described in detail in the National Sample Survey Organisation's Survey on Health Care, July 1995­June 1996 (NSSO 1998). The survey had a two-stage stratification procedure that assigned weights to each household, with the weight being the inverse of the probability of that specific household being selected as part of the sample. We used the inverse probability weights to derive the parameter esti- mates reported in this chapter. 38 · Health Policy Research in South Asia Unit Costs of Service Provision Perhaps the most satisfactory way to allocate public spending on health is to divide it up by type of service provided and then allo- cate the (gross) cost per unit of the service used among individuals. If certain expenditures cannot be allocated, then one can either omit them altogether or use some rule of thumb to distribute them across various groups. For the purposes of this study, we estimated the (gross) cost per unit of care from two sources--from available facility cost studies in India and from data on government expendi- tures on health during 1995­96. Three methods are commonly used to obtain the cost per unit of a health service. First, one can estimate the average costs per unit of a specific service based on some allocation rule, along the lines of Drummond and others (1997). There are few reliable facility- cost studies in India that directly estimate the cost per unit of ser- vice in this manner. The only work with which we are familiar that provides such estimates is a detailed time-and-motion study for dif- ferent types of hospitals and for a limited set of primary-care facili- ties in the state of Andhra Pradesh (World Bank 1997b). Data on total expenditures by facility--hospital or primary health care cen- ter--were more readily available. In terms of hospital costs, a key source of information for us was a survey by Sanyal and Tulasidhar (1995) of health care use and spending patterns in about 80 second- ary and tertiary public hospitals. The survey covered hospitals in 14 states and 3 union territories. The published version of the study presented data aggregated at the level of the state, and we were unable to obtain information at the hospital level. In addition, we obtained information on the total expenditures of two hospitals run by the Bombay Municipal Corporation, two maternity hospitals in Hyderabad, and a nonprofit hospital in Chennai (Mahal and others 2000). This information was supplemented by information from several different studies on the total expenditures of a few primary health care centers (IHSD 1996; Muraleedharan and others 1998; World Bank 1995b). Because of a lack of data on specific services, we used a second method that relied on assumptions about case The Distribution of Public Health Subsidies in India · 39 equivalence together with information on inpatient days, outpa- tient visits, and total facility expenditures to obtain the unit costs of a day of inpatient care and of a single outpatient visit. While useful, these two methods of estimating unit costs are essentially ad hoc ways of getting around a conceptual problem that typically arises whenever goods and services (in our case, health services) are jointly produced. It is difficult to talk meaning- fully of average costs. The more appropriate notion is marginal costs or an average marginal cost for a service. Estimating marginal costs requires data on health facility cost functions, which were not readily available to us. However, we were able to estimate a rudi- mentary hospital services cost function using expenditure data from the Sanyal and Tulasidhar (1995) study of public hospitals, even though it was aggregated at the state level. We have brought together all information on the cost of an outpatient visit and of an inpatient day of care at a public hospital from each of the methods in table 3.1. Table 3.1 Unit Cost Estimates from Facility-Level Data in India (1995/96 rupees) COST PER COST PER STATES SAMPLE STUDY INPATIENT DAY OUTPATIENT VISIT COVERED SIZE Primary care Muraleedharan and others 1998 310­2,147 13­17 Tamil Nadu 2 World Bank 1997a 73 8­10 Andhra Pradesh n.a. Public hospitals World Bank 1997a 134­187 34­46 Andhra Pradesh 2 >117­121 28­83 Andhra Pradesh 8 Based on data in Duggal 366­414 70­91 Maharashtra 1 and Nandraj 1994 Sanyal and Tulasidhar 1995 168­205 34­42 Maharashtra 1 Gupta and others 1992 245­290 43­54 All India State-level 150 25 Tamil Nadu 1 Immunization programs, pre- and postnatal care World Bank 1995a n.a. >11.75 n.a. n.a. IHSD 1996 n.a. 20­25 Orissa n.a. n.a. = Not available/not applicable. Source: Mahal and others 2000. 40 · Health Policy Research in South Asia Table 3.1 also presents costs per unit of service data for primary health centers using both the estimates from the time-and-motion study and from a case-equivalence approach. As table 3.1 shows, facility-level data for costing primary-level services (curative and preventive) were even less available than data for hospitals. In view of the limited quantity of facility-level costing data, we supplemented our findings with a BIA that used the average cost of providing health care services from a fourth source--government expenditures on health services. Our primary objective was to iden- tify elements of government health spending that could easily be allocated to those curative and preventive services for which we had utilization data from the household survey. We then divided the total government expenditures on each of these services by the number of units of services that were used under the relevant cate- gory (for example, "inpatient in a hospital" or "outpatient in a hos- pital"). The results that we obtained under this method were essentially similar to those that we obtained from direct facility costing studies; for that reason, we omitted them from the discus- sion in this chapter. (Interested readers can refer to Mahal and oth- ers 2000 for additional details.) Constructing Inequality Indexes As discussed previously, one key indicator of economic status that we used in this study was annual per capita consumption expenditure. We obtained these data for each individual by dividing the relevant household consumption expenditure per month by household size, then multiplying the result by 12. We then assigned each individual their household weights--first by rural or urban areas in each state and then rural-urban combined--and divided the sample into quin- tile groups. We constructed national-level quintile groups by pool- ing the various individuals across states in terms of their annual per capita consumption expenditures, accounting for their household weights by rural or urban areas and then rural-urban combined.1 To capture inequality in the use of public services or in the allo- cation of public subsidies, we used the concentration index (C) measure (Kakwani, Wagstaff, and van Doorslaer 1997; van The Distribution of Public Health Subsidies in India · 41 Doorslaer and others 2000) constructed from grouped socioeco- nomic status data. This index indicates the degree to which the poor and the nonpoor use services or benefit disproportionately from subsidies. A negative value for the use of outpatient services, for example, indicates that the poor use these services more than the nonpoor, while a positive value indicates the opposite. The more unequal the distribution across income groups, the larger (in absolute size) the concentration of service use. Higher use of these services by the poor may be considered equitable if the poor have a greater need for them than the nonpoor. Ignoring the element of need in inequality calculations assumes that need is invariant with socioeconomic status, which is typical in benefit incidence analyses. We follow this approach with one notable exception--in the way we treat the use of public services linked to reproductive and child health. Here we take the number of children in the appropriate socioeconomic group as an indicator of need. We measure inequity by the difference between inequality in utilization and inequality in need; if inequality in utilization is less than inequality in need, then our measure of inequity is positive, indicating that the inequity favors the nonpoor (van Doorslaer and others 2000). Main Findings Utilization of Public Health Services Inpatient Days. The three inpatient care columns in table 3.2 indi- cate the distribution of hospitalizations or, more precisely, inpa- tient days in public facilities, by patients' socioeconomic status--in our case indicated by per capita expenditure. The table distin- guishes between rural and urban areas and covers the estimated 88.5 million inpatient days spent in public facilities across India in the mid-1990s. Our summary measure is the concentration index of inpatient use of public facilities. Three key findings emerge from an examination of the inpatient columns of table 3.2. First, even if we ignore rural-urban differen- tials, there are considerable interstate differences in the utilization ALL and ­0.046 0.440 0.055 0.059 0.122 0.105 ­0.096 0.230 0.072 0.103 0.338 ­0.003 0.149 0.106 0.180 0.137 0.153 (ANMs) The CARE Mahal ALT in URBAN ­0.137 0.018 ­0.030 0.095 ­0.003 ­0.035 ­0.127 0.025 ­0.011 0.041 ­0.002 0.005 0.190 ­0.010 0.105 ­0.017 0.024 midwives ted PRENA considered. Status, nursey repor as also RURAL ­0.031 0.391 0.031 0.075 0.119 0.134 ­0.075 0.145 ­0.007 0.091 0.348 ­0.013 0.044 0.131 0.084 0.142 0.135 2000). auxiliar vice was ser by others YS) ALL DA ­0.071 0.408 0.146 0.445 0.613 0.176 ­0.083 0.307 0.157 0.392 0.412 0.150 0.349 0.052 0.634 0.226 0.287 "need" Socioeconomic ided health of andl TIENTA tor and prov (INP ublicp (Maha URBAN TH ­0.050 0.136 ­0.044 0.228 0.232 0.029 ­0.132 0.016 ­0.018 0.244 0.150 0.208 0.200 ­0.062 0.218 0.008 0.058 care indica relevant months CHILDBIR RURAL 0.041 0.286 0.154 0.484 0.649 0.297 ­0.069 0.353 0.047 0.266 0.339 0.098 0.210 0.071 0.653 0.195 0.265 postnatal the 24 additional State/Region, of and an and th vice, ALL bir ­0.038 0.061 ­0.056 0.011 0.104 0.006 ­0.128 0.051 ­0.078 0.096 0.056 ­0.016 0.056 ­0.073 0.063 0.025 0.005 Pre- care, Ser utilization of TIONS in prenatal between ypeT URBAN ­0.067 ­0.018 ­0.061 ­0.064 ­0.030 ­0.086 ­0.214 0.026 ­0.120 0.000 0.048 ­0.030 0.006 ­0.102 0.019 ­0.017 ­0.045 dispensaries. shares by IMMUNIZA and,y and children of RURAL ­0.002 0.051 ­0.038 0.029 0.102 0.053 ­0.099 0.048 ­0.010 0.091 0.038 0.011 0.040 ­0.037 0.038 0.053 0.039 quintile deliver Facilities centers, number . ALL ­0.04 0.319 Health ­0.143 0.137 0.212 0.087 ­0.081 0.183 ­0.047 0.010 0.103 0.157 0.281 0.062 0.156 0.048 0.104 healthy expenditure total vey CARE the sur immunizations, in the Public TIENTA primar capita URBAN for in ­0.159 0.126 ­0.341 0.036 0.163 ­0.056 ­0.046 0.055 ­0.115 0.041 0.205 0.080 0.274 0.044 0.161 ­0.053 ­0.004 per from shares OUTP vices from hospitals, indexes vations RURAL ­0.038 0.254 0.014 0.134 0.256 0.097 ­0.053 0.239 0.032 0.038 0.065 0.216 0.251 0.097 0.077 0.062 0.146 Ser quintile indexes) obser Care include ALL constructed capita 0.264 0.538 0.107 0.333 0.390 0.319 0.043 0.428 0.163 0.321 0.384 0.407 0.446 0.156 0.397 0.221 0.314 are concentration per outlying Health a a a CARE facilities of from of (concentration IENTTA URBAN ­0.019 0.331 0.113 0.358 0.279 0.071 ­0.058 0.169 ­0.059 0.280 0.178 0.247 0.451 0.058 0.403 ­0.039 0.092 health indexes removing INP construction Public constructed after Distribution 1995/96 RURAL 0.320 0.401 0.256 0.333 0.387 0.400 0.068 0.430 0.193 0.288 0.492 0.420 0.398 0.158 0.462 0.240 0.364 the Concentration For was 3.2 .P .P .P Pradesh. = need estimates Nadu .P Bengal P 2000. to of yana nataka theast India 42 ableT TE/A est ST REGION Andhra Bihar Gujarat Har Himachal Kar Kerala Madhya Maharashtra Nor Orissa Punjab Rajasthan amilT excluded. Uttar W All Note: is others index a Refers The Distribution of Public Health Subsidies in India · 43 of inpatient days in public facilities across per capita expenditure classes.2 The states in the south and west of India (Kerala, Tamil Nadu, Gujarat, and Maharashtra) have considerably lower values of the concentration index (which indicates a larger share of poorer socioeconomic groups) than the collection of poor states known as BIMARU (Bihar, Madhya Pradesh, Rajasthan, and Uttar Pradesh). Punjab and Haryana also have large concentration indexes, sug- gesting that the rich use public inpatient facilities more than the poor, relative to their shares of the population in these rich states of north India. Looking at patterns by rural-urban residence, the distribution of public sector inpatient days across socioeconomic groups is more biased against poorer people in rural areas than in urban areas. As before, there are considerable interstate variations in the distribu- tion of inpatient days across socioeconomic groups, even when we consider the data for rural and urban residents separately. The previous discussion might lead one to suspect that the pub- lic health system in India is unable to offer much financial security to the poorest in society. This would be the case if the fewer hospi- talizations reported by the poor either reflect discouraged dropouts from public facilities or substituted outpatient for inpatient care. However, if we suppose that needy cases are hospitalized, even among the poor, then information about which facility they use for that purpose is obviously of interest in terms of the financial pro- tection offered by the public system. The inpatient columns in table 3.3 provide information on the distribution of hospital inpatient days for the top and bottom expen- diture quintiles of the sample population, divided between public and all other sources of care (labeled as "private"), by state, by rural or urban residence, and by socioeconomic group. Overall, the poor tend to rely on the public sector for inpatient care while the rich rely mostly on private care, measured in terms of their respective propor- tions of inpatient days. Public facilities accounted for nearly 60 per- cent of all inpatient days utilized by the poorest 20 percent of the population, compared with 40 percent used by the wealthiest 20 per- cent. There is considerable variation across states, with the poorest 17.3 21.3 29.3 54.2 96.4 31.8 20.0 58.0 20.6 72.2 75.9 29.6 83.7 31.9 56.8 42.8 37.6 non- RICHEST CARE ALT clinics, PRENA tals, combined. 75.8 58.8 66.4 85.9 68.3 55.4 95.1 79.2 89.9 99.8 92.2 97.6 67.4 97.7 84.5 81.6 nts POOREST 100 hospi reside YS) DA owned 13.3 22.7 14.3 39.2 96.4 25.1 22.3 54.6 22.0 81.0 66.6 11.8 77.3 26.0 60.4 49.0 37.4 urban RICHEST State/Region, IENTTA and (INP privately Y rural vice, 57.7 30.1 52.2 53.4 50.4 64.1 94.4 66.5 79.2 48.8 95.1 72.5 73.9 98.5 69.9 100 100 for Ser DELIVER POOREST include of are ypeT 76.2 65.5 76.2 87.2 99.8 66.0 44.9 90.0 53.2 92.3 98.6 75.6 94.3 59.0 82.0 67.2 73.7 facilities shares by RICHEST TIONS Private share) quintile Facilities IMMUNIZA 95.7 86.5 98.9 93.4 97.8 96.4 80.6 97.9 98.6 96.6 95.6 94.1 98.9 88.5 90.6 98.2 93.9 capita POOREST (percent per Private dispensaries. The 7.9 7.8 7.9 7.7 and and 12.3 16.4 36.7 20.8 22.6 21.9 29.7 33.7 10.1 56.5 20.8 10.8 15.1 CARE RICHEST 1995/96 Public IENTTA centers, "public."y by OUTP 4.0 8.8 6.0 Quintile, 20.8 37.1 11.6 34.0 20.2 48.2 23.2 22.5 41.8 43.7 16.9 45.1 20.7 20.9 POOREST healthy categor vices the Ser primar of Expenditure 28.5 39.2 25.4 21.6 89.6 44.2 41.5 63.1 22.3 80.5 85.0 28.6 76.5 27.1 49.7 67.3 40.5 under Use CARE RICHEST of IENTTA hospitals, Capita subsumed INP Per 61.9 19.3 43.3 63.5 70.1 58.5 44.8 69.7 36.0 94.9 26.5 83.8 72.3 64.2 93.6 59.9 not POOREST 100 include Distribution and others facilities all 3.3 .P .P .P Nadu .P and Bengal Public ableT TE/A yana nataka theast est India ST REGION Andhra Bihar Gujarat Har Himachal Kar Kerala Madhya Maharashtra Nor Orissa Punjab Rajasthan amilT Uttar W All Note: profits, 44 The Distribution of Public Health Subsidies in India · 45 20 percent in less well-off states (such as Bihar, Madhya Pradesh, and the northeast region) relying less than the richest 20 percent on pub- lic care for inpatient stays, at least as measured by the public share in total inpatient stays. More generally, however, the pattern of the poor relying more on public facilities than the nonpoor persists even at the state level. This finding is sometimes diluted in the state-level data by the fact that, in rich states such as Punjab and Haryana, all groups tend to use private care, whereas in some extremely poor states such as Orissa and Rajasthan, there is much more use of public care across all socioeconomic groups. Outpatient Treatments. Outpatient treatments are distributed much more in favor of the poorer socioeconomic groups than inpatient days (table 3.2). This is certainly true at the national level, where the concentration index for treatments at public facilities is 0.104, much lower than the 0.314 for inpatient days. The pattern at the national level is also repeated at the state level. Presumably this reflects the fact that outpatient services in public facilities are more accessible to the poor. It may also reflect some of the unmet need among the poor for inpatient services, which translated into an increased use of outpatient treatment services. As with inpatient days, there is considerable interstate variation in the concentration index of outpatient treatments at public facili- ties. Much of this variation is similar to that for inpatient days, but with one key difference--overall, there is less interstate variation in the value of concentration indexes for outpatient treatments than in those for inpatient stays. We believe there are at least two rea- sons why this is the case. First, services provided by primary health centers and public dispensaries constitute about one-third of all outpatient treatments at public facilities (Mahal and others 2000). This contrasts with their small share in days of inpatient stays in public facilities--barely 5 percent at the national level. This fact matters in rural areas, because outpatient treatment is now accessi- ble even to poor people, as many primary health care centers have recently been opened in rural areas. Second, the poor may be sub- stituting outpatient treatment for inpatient days more frequently 46 · Health Policy Research in South Asia precisely in those states where it is relatively difficult to gain access to inpatient services at public facilities. Another contrast between the distribution of inpatient days and outpatient treatments by public and private care is apparent from the information on the share of outpatient treatments provided in public facilities (see table 3.3). Nearly 80 percent of all treatments received by the poorest 20 percent of the population were provided in the pri- vate sector, compared with 85 percent received by the top quintile. These are significantly greater shares than for inpatient days. This trend is also evident at the state level, but there are significant inter- state variations in the role of the private sector in outpatient care. Immunizations. We also had information on the immunization status of children from birth to four (full) years old. Information had been collected on the total number of diptheria-pertussis-tetanus (DPT) doses (including booster doses, if any), oral poliovirus vaccine (OPV) doses (including booster doses), and Bacillus Calmette Guerin (BCG) and measles vaccinations that they received. Except in the case of measles, respondents were also asked whether they had received the dose at a government or private facility. In our analysis, we exclude measles and look specifically at immunizations among children from birth to 24 months old at the time of the sur- vey. We did this for two reasons. First, some of the immunization doses, such as booster shots for OPV and DPT, are given when the child is more than one year old. Thus, focusing only on children under one year of age would have yielded a downward-biased esti- mate of total immunizations received by children in the previous year (NSSO 1995b).3 Second, focusing only on currently living children who were more than one year old but less than 24 months old would also have yielded downward-biased results, since it would have excluded children who might have been immunized during their first two years of life but who died in the previous year.4 Another potential source of bias was that the database from the 52nd round of the NSS has far fewer children ages 12 to 24 months than from birth to 12 months old. On the other hand, including both groups in our calculations would have biased the result for the The Distribution of Public Health Subsidies in India · 47 total number of immunizations received during the previous year in an upward direction, balanced by the fact that we do not include measles vaccinations. The numbers in table 3.2 are the differences between the con- centration index for publicly provided immunization doses and the concentration index for need, as measured by the proportion of children between birth and 24 months in each group. (There are quite significant differences in the latter across groups, with more of these children among the poor.) The data indicate that the immunization doses provided by the government are fairly equi- tably distributed, with concentration indexes hovering around zero. A few remarks are in order, however. Although immunization doses are not directly comparable with outpatient treatments and inpa- tient stays in view of the different concentration indexes used, the distribution of immunization doses appears to be fairer in terms of its allocation to poorer groups than outpatient treatments and inpatient stays. In fact, if we did not adjust for the greater propor- tion of children belonging to the poorer groups, the distribution of utilization would favor them even more. In most states, poorer groups receive a significant share of total government immunizations. However, in Gujarat, Kerala, Maha- rashtra, and Tamil Nadu, the poor receive much higher shares of public sector immunizations than in the other states, a pattern that was also true for outpatient treatments and inpatient stays. Table 3.3 also provides information on the proportion of total immunizations that was provided by the public sector to children under 24 months of age. More than 90 percent of all immunizations given to children in the poorest expenditure quintile came from the government. For the top quintile, the share of the public sector was somewhat smaller but still quite significant, at 74 percent. Overall, the government provided the bulk of the immunization services-- 89 percent of all doses of DPT, OPV, and BCG received by children for whom our table has information. Moreover, the public sector provided a higher proportion of all immunization doses in rural areas than in urban areas--93 percent and 78 percent, respectively (Mahal and others 2000). 48 · Health Policy Research in South Asia Childbirth. Indians also use public facilities for childbirth, relying on inpatient days and other forms of care. The 52nd round of the NSS specifically distinguished hospitalization for this purpose from inpatient days for other illnesses for households with a child between birth and one year of age at the time of the survey (NSSO 1995a, 1995b, 1998). According to estimates based on NSS data from the 52nd round, childbirth and associated medical factors resulted in nearly 15.95 million inpatient days in public facilities during 1995, the bulk of which were spent in hospitals. Nearly two-fifths of all inpatient days were accounted for by urban residents; within that group, 90 percent stayed in public hospitals (Mahal and others 2000). The upper expenditure quintiles accounted for a disproportion- ately large share of these inpatient days, especially in rural areas. This fact is reflected in the magnitudes of the concentration index for rural and urban areas (see table 3.2).5 Moreover, these rural- urban differences hold true for individual states as well. Another general pattern--which has already been noted in the case of other service utilization categories--is that, in the southern states (Andhra Pradesh, Karnataka, Kerala, and Tamil Nadu) and in Gujarat, Maharashtra, and West Bengal, inpatient days related to childbirth are more evenly distributed among different socioeconomic groups than in Bihar, Orissa, Rajasthan, Uttar Pradesh, and the northeast. Notice also that the concentration indexes for inpatient days related to childbirth are greater than those for outpatient treatments and immunizations, indicating yet again that inpatient days in public facilities are disproportionately used by the rich, irrespective of the cause. Table 3.3 shows that the share of private care increases from the bottom to the top of the socioeconomic ladder as measured by per capita expenditure, although considerable interstate differences persist. Some states, particularly Andhra Pradesh, Gujarat, Kar- nataka, Kerala, Maharashtra, and Punjab, used more private services than others. The poorer states, such as Madhya Pradesh, Orissa, Rajasthan, and West Bengal, tended to rely on public services. Although not reported in the tables, auxiliary nurse midwives (ANMs) attended an estimated 1.46 million births in the year pre- The Distribution of Public Health Subsidies in India · 49 ceding the survey; the overwhelming majority (90 percent) of such births were to rural residents. Women from the lower expen- diture quintiles typically accounted for a disproportionate share of the births attended by ANMs, even after accounting for the socioeconomic distribution of the total number of children born during the period in question. Thus, 62 percent of the births attended by ANMs in India were to women in the bottom two quintiles, although their share of the number of children born was 49 percent. This picture is broadly reflected in state-level data as well. Prenatal Visits. Of the estimated 52.5 million visits for prenatal care all over India in the year prior to the survey, about 60 percent (or 30.8 million) were handled by public facilities and the remainder by the private sector. Excluding care provided by ANMs, public hospitals accounted for nearly 56 percent of all prenatal visits in the public sector (Mahal and others 2000). Table 3.2 presents concentration indexes of the distribution of prenatal visits to all public facilities (including hospitals). As dis- cussed earlier, the socioeconomic distribution of such visits to pub- lic hospitals is more unequal than the distribution of visits to other kinds of public facilities, such as primary health centers (PHCs). There are essentially three main findings in this category. First, the distribution of pre- and postnatal visits in rural areas is fairer to the poor than to the nonpoor. Second, there are interstate differences along the lines discussed earlier. Third, the concentration indexes for pre- and postnatal visits are typically lower in magnitude than those for inpatient stays. Table 3.3 presents the proportion of prenatal visits to public facilities. The data indicate that the public sector was the major provider of such care in most states. The share of total visits to pri- vate facilities, however, increased with per capita consumption expenditure; at the national level, people in the top quintiles made more than three times the share of visits to private facilities than those in the lowest quintile. This pattern was also observed among both rural and urban residents when considered separately, as well 50 · Health Policy Research in South Asia as in individual states. Private facilities appeared to be particularly popular in the southern and western Indian states. In addition to the institutional sources discussed above, ANMs constituted an important provider of prenatal care. ANMs handled an estimated 19.4 million visits, with rural residents accounting for 96 percent of that total. Given that ANMs are likely to be cheaper than other sources, it is not surprising that 60 percent of the total number of visits to ANMs were by people in the bottom 40 percent of the population. The relative importance of this source of care varied substantially from state to state, ranging from a low of 3 per- cent of all visits (excluding ANM visits) to public facilities in Kerala to 100 percent in the case of Gujarat (Mahal and others 2000). Gross and Net Costs of Health Care Services This section provides an additional aspect to the methodology that we used to calculate government subsidies to a specific public health service. To estimate the subsidy, we needed first to estimate the gross per-unit cost to the government of providing a specific service--for example, the cost per inpatient day, per outpatient visit, per immunization received, or per prenatal visit. This infor- mation has been provided in table 3.1. We then subtracted any amounts that the government may have recovered through user charges from unit cost estimates to get an estimate of the subsidy received per unit of utilization of a specific service. Thus, this sec- tion focuses on user charges. Given that the objective of a BIA is to estimate the net subsidies that accrue to individuals in different socioeconomic groups, the amounts patients paid for the use of health facilities are a key element in these calculations. We had access to two sources of information about fees paid by users of public facilities. The first was the NSS itself. The database of the 52nd round contained information about charges paid for inpatient care in hospitals and other health facilities. Unfortu- nately, information about any fees that might have been paid for the use of outpatient facilities was not directly available, as they were subsumed under the general category of "medical expendi- tures." Nor was any information available about fees that might The Distribution of Public Health Subsidies in India · 51 have been paid for prenatal care, immunization, and inpatient days related to childbirth. However, the NSS data did include informa- tion on immunizations and inpatient stays associated with child- birth that were received free of charge at public facilities. Our second source of information was government budget doc- uments, which typically include data on actual revenues received in previous years (Selvaraju 2000). We were able to compare the information on revenues presented in those documents with the information reported by respondents in the 52nd round of the NSS. Although the estimates of user fee revenues from the two sources differed, this did not affect the results of our BIA in any significant way (Mahal and others 2000). In this chapter, we use the estimates of user charges derived from the survey data rather than those from the budget documents. Table 3.4 provides survey-based estimates of the share of user fees paid to public facilities by (a) the richest 40 percent of the pop- ulation, (b) rural and urban residents separately, and (c) state. Table 3.4 Share of Richest 40 Percent of the Population in Total Hospital Charges Paid to Public Health Facilities by State and Region, 1995/96 (percent) STATE/REGION RURAL URBAN COMBINED Andhra Pradesh 91.6 54.5 92.8 Bihar 92.5 83.5 98.0 Gujarat 60.4 78.3 71.6 Haryana 67.2 89.0 80.5 Himachal Pradesh 93.0 60.7 92.9 Karnataka 55.4 80.7 59.5 Kerala 50.6 65.1 63.5 Madhya Pradesh 58.5 64.9 72.1 Maharashtra 74.6 70.8 89.8 Northeast 84.6 80.1 90.2 Orissa 93.7 89.8 91.3 Punjab 67.9 92.6 89.6 Rajasthan 88.1 98.3 98.1 Tamil Nadu 68.4 88.1 91.8 Uttar Pradesh 93.6 84.0 92.1 West Bengal 92.5 80.6 90.0 All India 84.2 81.7 86.4 Source: Mahal and others 2000. 52 · Health Policy Research in South Asia These figures indicate that those in the top expenditure quintiles pay an overwhelming proportion of the user charges the govern- ment received. In India as a whole, people in the top two expendi- ture quintiles pay for more than 86 percent of these charges. Even within states, the share of user charges paid by those in the top quintiles is disproportionately high, although the precise share varies widely. For example, the share of hospital fee revenues paid by the top two expenditure quintiles ranged from a low of about 60 percent in Karnataka and Kerala to 98 percent in Rajasthan and Bihar. In most of the poorer states, the top quintiles paid very high proportions of total statewide fee revenue, and this share exceeded the share of inpatient days used by people in these quintiles in all of the states. Assuming that the quality of care received is similar across all socioeconomic groups, then it can be argued that the greater financial burden borne by richer groups makes up to some extent for the bias against the poor in the distribution of public subsidies for inpatient care. When we examined the data by rural and urban respondents separately, we observed a similar picture. We also found that the distribution of free immunization ser- vices closely paralleled the pattern of utilization. This is not sur- prising, as these services are provided mostly free by the public sector under various initiatives such as the universal immunization program. (Ninety-nine percent of immunization doses were pro- vided free.) We assumed that "free" immunization services were just that--completely subsidized--whereas fees were charged for "nonfree" services sufficient to cover the costs of providing such services. Moreover, each immunization dose received was consid- ered a separate visit for the purposes of our BIA. The survey also provided us with data on the distribution of inpa- tient days for childbirth spent in free wards in public facilities. These data could be disaggregated by rural-urban residence, by various indicators of socioeconomic status, and by individual states. A total of 13.3 million inpatient days of care (or roughly four-fifths of all child- birth-related inpatient care at public facilities) were spent in free wards. Our primary finding is that the pattern of how women from various expenditure quintiles gave birth in free wards closely follows the distribution of inpatient care, with a slight bias in favor of poorer The Distribution of Public Health Subsidies in India · 53 groups and against richer groups. This, generally speaking, is the pic- ture at both the national and the state levels and in both rural and urban areas. Some states--Andhra Pradesh, Haryana, Karnataka, Kerala, Punjab, and West Bengal--had a noticeably fairer distribu- tion of free ward days. Thus, if we assume that there are no hospital charges for inpatient days in free wards, the implication is that public subsidies to inpatient care in free wards is more equitably distributed than public subsidies to inpatient care (for childbirth) in general. We had to make a few other assumptions before proceeding with the BIA. We assumed that no user fees were charged for outpatient visits. We had two main reasons for making this assumption. First, no direct estimates of revenues from these charges were available because they were subsumed under the category of medical expendi- tures in the NSS data. Second, and more crucially, the estimates in the NSS data for hospital charge revenues alone exceed the total rev- enue estimates from user fees reported in state government budgets by a large margin, so we thought it would be prudent not to make any guesses about any additional amounts that might have been raised from outpatient visits. Given the available evidence that any amounts paid by users are likely to have been positively correlated with socio- economic status (in other words, to have been paid by those most able to afford them), excluding them as we did probably biased the esti- mated distribution of public health subsidies toward greater inequity. We treated all visits to ANMs as visits to public facilities, given the substantial numbers of health subcenters that operate at village lev- els. Moreover, we treated such visits and any assistance received from ANMs in childbirth as being either fully paid for or fully subsidized. Results of Benefit Incidence Analysis We used the following common set of unit cost estimates for calcu- lating subsidies based on table 3.1 and additional evidence: · One inpatient day of curative care or childbirth at a public hospi- tal (270 rupees) · Outpatient treatment or prenatal visit at a public hospital (50 rupees) 54 · Health Policy Research in South Asia · One inpatient day at a primary care or other health facility (150 rupees) · Outpatient treatment or pre- or postnatal visit at a PHC or other facility (25 rupees) · Immunization dose per visit or prenatal visit to an ANM (15 rupees) · Childbirth attended by an ANM (75 rupees) The first two estimates are the midpoint of the range of numbers reported in table 3.1 and are close to our estimates of average incre- mental costs for these services. The estimates for inpatient stay and outpatient visits to PHCs are a modified form of the data in Muraleedharan and others (1998), corrected for the upward bias that results from using low utilization of inpatient beds when calcu- lating unit costs.6 The cost for a dose of immunization or for a pre- natal visit was estimated to be 15 rupees, which is the midpoint of the cost per visit reported in World Bank (1997a) and the upper bound estimates reported in table 3.1. No estimates of the cost of a single childbirth attended by an ANM/lady health volunteer (LHV) were available, but we assumed it to be the cost of half a day of work, or about 75 rupees (see also Muraleedharan and others 1998). For our analysis, we first allocated user fee revenues to five dif- ferent socioeconomic groups (per capita expenditure quintiles) based on the distribution in the NSS data and on the assumptions discussed in the previous section. Table 3.5 presents our results on the distribution of public subsi- dies on health in India by state and by rural-urban classification using concentration indexes. Elsewhere, we report the results by level of care--public hospitals, primary health facilities and other lower-level care, and immunizations (Mahal and others 2000). These results fol- low closely the utilization patterns reported earlier in this section, so we have excluded them from our analysis for reasons of space. Our primary finding is that for the types of care considered here, public health subsidies are disproportionately distributed in favor of the richer groups: some 31 percent of total subsidies go to those in The Distribution of Public Health Subsidies in India · 55 Table 3.5 Distribution of Public Subsidies by State/Region and Socioeconomic Status, 1995/96 (concentration indexes) STATE/REGION RURAL URBAN COMBINED Andhra Pradesh 0.140 ­0.091 0.116 Bihar 0.296 0.234 0.419 Gujarat 0.142 ­0.082 0.000 Haryana 0.231 0.096 0.201 Himachal Pradesh 0.350 0.225 0.340 Karnataka 0.281 ­0.041 0.209 Kerala 0.014 ­0.113 ­0.041 Madhya Pradesh 0.308 0.076 0.292 Maharashtra 0.107 ­0.160 0.060 Northeast 0.178 0.196 0.219 Orissa 0.340 0.161 0.282 Punjab 0.245 0.211 0.223 Rajasthan 0.288 0.305 0.334 Tamil Nadu 0.090 ­0.050 0.060 Uttar Pradesh 0.348 0.232 0.304 West Bengal 0.164 ­0.075 0.157 All India 0.261 ­0.012 0.214 Note: The formula used to construct concentration indexes from quintile shares in total public health subsidies was reported in Mahal and others 2000. the highest quintile, while only 10 percent go to those in the lowest quintile. Table 3.5 demonstrates this through a concentration index of the distribution of public subsidies. For India as a whole (both rural and urban), the index is positive--0.214--which suggests a bias in the distribution of public subsidies toward the richer groups. Moreover, as our discussion about utilization patterns across five different types of services suggests, this result is driven primarily by the distribution of public subsidies on inpatient care. Nevertheless, there are considerable interstate differences. As expected, selected southern states (Andhra Pradesh, Kerala, and Tamil Nadu), Gujarat, Maharashtra, and West Bengal have a distribution of public subsi- dies that is fairer to poorer groups than do the other states. Health subsidies are also much more equally distributed among urban residents than among rural residents. Thus, the top quintile of rural residents accounted for about 39 percent of all subsidies in comparison with the bottom quintile's 10 percent, while for urban residents these proportions were 15.7 percent and 16.1 percent, 56 · Health Policy Research in South Asia respectively (Mahal and others 2000). This is primarily because subsidies for public hospitals are much more unequally distributed in favor of the upper quintiles in rural areas than in urban areas. Again, interstate differences show up here along the expected lines. Bihar, Madhya Pradesh, Orissa, Rajasthan, and Uttar Pradesh, which are among the poorest states in India, have the least egalitar- ian distribution of subsidies, although Himachal Pradesh, a state with a growing economy, also has a highly unequal distribution of subsidies to public health. The reasons clearly need to be analyzed further in the future. Policy Implications The calculations in the previous section suffer from a number of limitations, two of which are particularly significant. First, the fact that we were able to access only a limited number of facility costing studies for India implies that our approach to allocating subsidies failed to account for interstate and interquintile differences in the cost of providing services. Consequently, the distribution of subsi- dies in our results is biased toward states with poor-quality facili- ties. Moreover, if it can be assumed that improving the quality of care leads to increased use of public facilities, then the total amount of the subsidies is itself likely to be underestimated. Within any state, the lack of information about the quality of care available to different groups may cause a bias if richer individuals get better- quality care. The second limitation, obviously, is our use of a case-equivalent approach for estimating unit costs from government expenditure data. In this case, we were able to use state-level data to obtain unit cost information specific to each state, but we were forced to fall back on assumptions about relationships between these costs for different types of care because of the difficulty of obtaining govern- ment expenditure data in sufficient detail (Mahal and others 2000). Although the exact implications of these assumptions for the distri- bution of subsidies are not clear, our analysis points to the need for The Distribution of Public Health Subsidies in India · 57 a more careful study of government expenditures in individual states in the future. Even taking these limitations into account, several clear mes- sages emerge from our analysis. First, health subsidies are not par- ticularly well targeted to the poor in India, especially the poor who live in rural areas and in the poorer states. States in the south of India, such as Kerala and Tamil Nadu, do considerably better in this regard than their poorer counterparts in the north, such as Bihar and Uttar Pradesh. Second, the allocation of subsidies across the different quintiles is driven by the size and distribution of subsidies to hospital-based care. This is unfortunate because the distribution of subsidies for primary care and for inpatient stays related to childbirth are rea- sonably evenly distributed across different socioeconomic groups. Indeed, programs associated with maternal and child health (such as programs dealing with pre- and postnatal care and immuniza- tions) that are linked to schemes sponsored by the central govern- ment appear to be targeted much better, from a distribution perspective, than purely curative care. Third, the unequal distribution of subsidies for inpatient stays that we have found in our analyses of the NSS data does not neces- sarily imply that the public sector fails to insure poor patients against expensive episodes of illness. We observed that poorer patients and most of the poor states use more publicly provided hospital services than private hospital services, compared with their richer counterparts. The catch is that meeting the insurance objec- tives of the poorer sections of society appears to involve a trade- off--in providing insurance for the poor, the public sector is also handing over large amounts of public subsidies to the rich, espe- cially those who live in rural areas. So what do our findings imply for policy? Richer groups may simply have access to better-quality public health facilities because of their greater bargaining power, a fact that is obvious to anyone who has tried to be admitted to a public hospital in India. However, this lack of access to high-quality care in the public sector may be driven more by the fact that the rural poor tend to live a long 58 · Health Policy Research in South Asia distance from medical facilities and that they must take off of work and lose income if they decide to seek health care (Gertler and van der Gaag 1990). This is suggested by the fact that hospital subsidies are distributed much more evenly among the urban population than among the rural population. Richer individuals also consume more health care, a natural out- come if such care is considered a normal good. This is well known empirically and internationally, and recent work by Gumber (1997) and Duraisamy (1998), among others, has established that this is also the case in India. This fact alone can have a substantial impact on the distribution of public subsidies. In light of the above, it is obvious that increasing household income and developing infrastructure that increases access to hos- pital care would help to improve the allocation of subsidies. This assertion is supported by preliminary analyses that we undertook linking inequalities in the allocation of health subsidies to road densities across the 16 Indian states and regions that we examined in this study. For example, the correlation coefficient between the ratio of the share of the top and bottom expenditure quintiles in public sector immunizations and road density was ­0.78; in other words, higher road density (roads per square kilometer) was associ- ated with lower inequality in public health subsidies. Richer states, as measured by per capita expenditure, and states with lower levels of inequality in per capita expenditures also had lower levels of inequality in the allocation of public health subsidies. This obser- vation suggests that there need not be a tradeoff between growth and the allocation of health services to the poor. Further work is necessary to explore these issues fully and to test the robustness of these early findings. Apart from income growth and infrastructure development, there are other actions that the government could take. Our analysis sug- gests that one such action would be to improve the quality of care provided at primary care centers. If existing expenditure levels on primary care and the cost of services per unit are anything to go by, simply increasing the share of expenditures may not be the only rem- edy for a cash-constrained government. Rather, a more effective The Distribution of Public Health Subsidies in India · 59 approach might be to make care providers at the level of primary care more accountable. This could be accomplished, perhaps, by transferring some of the care provision responsibilities to private providers; by decentralizing control over revenues, expenditures, administration, or all three (see Mahal, Srivastava, and Sanan 2000); or by introducing other kinds of democratic participation in service provision. This approach is supported by the observation that some of the better-performing states (such as Gujarat, Maharashtra, and West Bengal among the poorer states) have also experienced a greater degree of administrative decentralization in the past 30 years. These suggestions, particularly the suggestion to improve the quality of public sector care, run into the problematic observation made by Besley and Coate (1991) that greater equity can be achieved and insurance for the poor improved if the quality of available public services is not "too high." They argue that the interests of the poor could be served by the public sector if the richer groups start using private care or unsubsidized public facili- ties such as paid inpatient wards. For this to happen, however, the nonpoor would have to perceive the quality of care in the private, unsubsidized facilities as being better than that of subsidized public services. If this line of reasoning is correct, then the much more even distribution of subsidies in some states in our data may reflect the fact that high-quality private sector options are available to the richer groups in those states. In other states and regions (such as Bihar, Madhya Pradesh, Rajasthan, Uttar Pradesh, and the north- east), such options may not be available either because of a general absence of private sector services or a lack of regulatory standards and enforcement. Indeed, these states and regions have close to the smallest bed-to-population ratios among all Indian states, suggest- ing a lack of options for users and a consequent bias in public health care in favor of the wealthier and more influential groups (Gumber 1997). Poorer states may also have poorer-quality regula- tory systems that, in turn, may promote a low quality of care. In these circumstances, there may be a case for developing additional government facilities in these states and regions and for improving the regulatory regime that oversees the provision of health care. 60 · Health Policy Research in South Asia Notes 1. Constructing national quintiles and joint rural-urban quintiles requires making the uncomfortable assumption that consumption expenditure levels are comparable across regions and rural-urban residences, despite the fact that the consumption baskets we used in constructing indexes for individual states differed. 2. We have not distinguished between public hospitals and facilities such as primary health care centers in our analysis of inpatient days. Although the distribution of inpatient days in primary health care centers is much more egalitarian than the distribution of public hos- pital inpatient days, it accounts for only a very small proportion of all inpatient days spent in public facilities (Mahal and others 2000). 3. This is also true for the measles vaccine. 4. This problem exists even if we focus on children between birth and 12 months of age, although obviously to a lesser degree. 5. The proportion of children from birth to 12 months old in each socioeconomic group is our indicator of need. 6. Specifically, we calculated unit costs on the assumption that inpatient beds were fully utilized, and thus, we scaled up the inpa- tient day numbers reported in Muraleedharan and others (1998) for this purpose. We used a case-equivalent methodology of six outpatient visits being equal to one inpatient visit in terms of cost to obtain our estimates. References Besley, Tim, and Steven Coate. 1991. "Public Provision of Private Goods and the Redistribution of Income." American Economic Review 81: 979­84. Castro-Leal, Florencia, Julia Dayton, Lionel Demery, and Kalpana Mehra. 1999. "Public Social Spending in Africa: Do the Poor Benefit?" The World Bank Research Observer 14(1): 49­72. Drummond, Michael, Bernie O'Brien, Greg Stoddart, and George Torrance. 1997. Methods for the Economic Evaluation of Health Care Programmes. 2nd ed. London: Oxford Medical Publications. The Distribution of Public Health Subsidies in India · 61 Duggal, Ravi, and Sunil Nandraj. 1994. "Health Finances of the Bombay Munici- pal Corporation." Background paper for the seminar on Improving Public Hospitals in Bombay City, Medico Friend Circle (Bombay group), June 26. Duraisamy, P. 1998. "Morbidity in Tamil Nadu: Levels, Differentials, and Deter- minants." Economic and Political Weekly XXXIII(17): 982­90. Gertler, Paul, and Jacques van der Gaag. 1990. The Willingness to Pay for Medical Care: Evidence from Two Developing Countries. Baltimore: The Johns Hopkins University Press. Gumber, Anil. 1997. "Burden of Disease and Cost of Ill-Health in India: Setting Priorities for Health Interventions during the Ninth Plan." Margin 29(2): 133­72. Gupta, J. P., B. Sharma, Y. Gupta, Reeta Dhingra, and K. Nair. 1992. Financing Health Care in Nonstate Sector in India. New Delhi: National Institute of Health and Family Welfare, Department of Planning and Evaluation. IHSD (Institute for Health Sector Development). 1996. Orissa Health and Family Welfare Project: Impact Assessment Study. London. India, Government of. 1993. "Report of the Expert Group on Estimation of Pro- portion and Number of Poor." New Delhi: Planning Commission, Prospec- tive Planning Division. ------. 1996. The Constitution of India. New Delhi: Law Ministry. ------. 1999. Economic Survey 1998­99. New Delhi: Ministry of Finance. Kakwani, Nanak, Adam Wagstaff, and Eddy van Doorslaer. 1997. "Socioeconomic Inequalities in Health: Measurement, Computation, and Statistical Infer- ence." Journal of Econometrics 77(1): 87­103. Lok Sabha. 1985. National Health Policy. New Delhi: Government of India Press. Mahal, Ajay, Janmejaya Singh, Farzana Afridi, Vikram Lamba, Anil Gumber, and V. Selvaraju. 2000. Who Benefits from Public Health Spending in India? New Delhi: National Council of Applied Economic Research. Mahal, Ajay, Vivek Srivastava, and Deepak Sanan. 2000. "Decentralisation and Public Service Delivery in Health and Education Services: Evidence from Rural India." In Jean Dethier, ed., Governance and Development in China, India, and Russia. Dordrecht, Netherlands: Kluwer Academic Publishers. 62 · Health Policy Research in South Asia Meesook, Oey. 1984. "Financing and Equity in the Social Sectors in Indonesia." World Bank Staff Working Papers #703. World Bank, Washington, D.C. Muraleedharan, V., and the members of the Health Sector Project Working Group. 1998. Cost Analysis of Two Types of PHCs. Chennai, Tamil Nadu: DANIDA. NSSO (National Sample Survey Organisation). 1995a. Schedule 25.0: Survey on Health Care, July 1995­June 1996. New Delhi: Government of India, National Sample Survey Organisation. ------. 1995b. Instructions to Field Staff: Design, Concepts, Definitions, and Proce- dures, Vol. I. New Delhi: Government of India, National Sample Survey Organisation. ------. 1998. Morbidity and Treatment of Ailments. Report #441. New Delhi: Gov- ernment of India, Department of Statistics. Reddy, K. N., and V. Selvaraju. 1994. Health Care Expenditure by Government of India: 1974­75 to 1990­91. New Delhi: Seven Hills Publication. Sanyal, S. K., and V. B. Tulasidhar. 1995. Utilisation Pattern and Financing of Public Hospitals: A Report. New Delhi: National Institute of Public Finance and Policy. Selden, Thomas, and Michael Wasylenko. 1992. "Benefit Incidence Analysis in Developing Countries." Working Paper Series #1015. World Bank, Country Economics Department, Washington, D.C. Selvaraju, V. 2000. Government Expenditures in the Health Sector, 1995­96. Draft report. New Delhi: National Institute of Public Finance and Policy. van de Walle, Dominique. 1992. "The Distribution of the Benefits from Social Ser- vices in Indonesia 1978­87." Policy Research Working Paper Series #871. World Bank, Public Economics, Country Economics Department, Washing- ton, D.C. van Doorslaer, Eddy, Adam Wagstaff, Hattem van der Burg, Terkel Christiansen, Diana De Graeve, Inge Duchesne, Ulf-G Gerdtham, Michael Gerfin, Jose Geurts, Lorna Gross, Unto Hakkinen, Jurgen John, Jan Klavus, Robert E. Leu, Brian Nolan, Owen O'Donnell, Carol Propper, Frank Puffer, Martin Schellhorn, Gun Sundberg, and Olaf Winkelhake. 2000. "Equity in the Deliv- ery of Health Care in Europe and the U.S." Journal of Health Economics 19: 553­83. World Bank. 1995a. India: Policy and Finance Strategies for Strengthening Primary Health Care Services. Report #13042-IN. Population and Human Resources Division, South Asia Country Department II, Washington, D.C. The Distribution of Public Health Subsidies in India · 63 ------. 1995b. India's Family Welfare Program: Toward a Reproductive and Child Health Approach. Report #14644-IN. Population and Human Resources Divi- sion, South Asia Country Department II, Washington, D.C. ------. 1997a. Health, Nutrition, and Population. Human Development Network, Washington, D.C. ------. 1997b. India: New Directions in Health Sector Development at the State Level: An Operational Perspective. Report #15753-IN. Population and Human Resources Division, South Asia Country Department II, Washington, D.C. CHAPTER 4 Equity in Financing and Delivery of Health Services in Bangladesh, Nepal, and Sri Lanka Institute of Policy Studies, Sri Lanka Abstract Using a national health accounts (NHA) approach, existing data sources were used to assess the equity in distribution of financing and health system resources in Bangladesh, Nepal, and Sri Lanka. Simple tables of incidence by decile, Lorenz curves, Kakwani indexes, Suits indexes, and concentration indexes were used for the analysis. NHAs already existed in official form for Bangladesh and Sri Lanka, but not for Nepal. All three countries have health systems in which the predominant sources of financing are taxes and direct out-of- pocket payments by households. Despite the overall similar- ity in financing and provision systems, significant differences are observed in actual performance between Bangladesh and Sri Lanka. In Sri Lanka, both tax and out-of-pocket pay- ments were progressive means of financing, and government health care expenditures were pro-poor in their distribution. In Bangladesh, the same financing mechanisms were mod- estly regressive, and the distribution of government health expenditures was not pro-poor. This difference in national 65 66 · Health Policy Research in South Asia performance appears to be due to (a) the fact that poor peo- ple in Bangladesh are less likely than the nonpoor to recog- nize and to respond to their own episodes of illness, and (b) a lack of any consumer-level differences in quality between the private and the public sectors, which encourages the rich to prefer to use private services instead of public ones. Sri Lanka's experience suggests that increasing equity in Bangladesh would require increasing health awareness among the Bangladeshi poor, substantially increasing their access to modern medical services, and improving the pro- gressiveness of taxation in general.1 Introduction Equity in the design and operation of health care systems is an important system goal for nations. The notion of fairness in financing as recently proposed by the World Health Organization (WHO 2000) is only one component of equity out of many that are of concern to governments. Equity in access to and use of health services is commonly an important goal for policymakers in countries in the Southeast Asia region (SEAR) and elsewhere. Alleviating poverty through the redistributive effect of public health spending can be another important welfare goal for devel- opment planners in many countries. However, realistic assess- ments of the extent to which these goals are achieved by the health system have been rare outside the industrial world until recently. Current WHO efforts to measure fairness in financing as defined by WHO's Evidence and Information for Policy may be a useful contribution to the understanding of health systems, but they do not address the equity concerns of policymakers. There is little dispute that, from the perspective of improving overall health status, equity in access to health care or in provision within a health system is an important intermediate objective, because inequalities in such access have an important impact on the overall health status of any national population. Nevertheless, Equity in Financing and Delivery of Health Services · 67 equity in access is an important objective in its own right for most countries (van Doorslaer, Wagstaff, and Rutten 1993). In the con- text of the poorer SEAR countries, equity in access to public health services is also an important objective for alleviating poverty. For example, it has been argued that provision of basic health services in Malaysia and Sri Lanka has been an important mechanism for mitigating the impact of poverty in rural areas (Alailima and Mohideen 1983; Meerman 1979). It is also sometimes argued that public financing and provision of health services is desirable as an indirect means of redistributing material resources within a society. So in addition to the objectives of equal health outcomes and equal access, policymakers can be concerned with the redistributive impact of government health expenditures. This connects to the related interest of policymakers in the equity of the financing mechanisms for health care. Health systems and their reforms are often judged not only in terms of their implications for the distri- bution of access and outcomes, but also in terms of the distributive burden of financing. The evaluation of equity in financing is not only an important objective in its own right, but is intimately con- nected with the evaluation of the provision and supply side of health care systems. Much of the pattern of equity found in a particular health system is a function of the existing patterns of social and economic equity outside the health system itself. Nevertheless, it is generally assumed that the structure and organization of the delivery and financing of health care can play a significant role in determining the equity in financing and access to and use of health services. However, with the exception of the one study in Organisation for Economic Co-operation and Development (OECD) countries, there has been little systematic analysis in developing countries of the equity characteristics of alternative national health care financing and delivery systems (van Doorslaer, Wagstaff, and Rutten 1993). Despite the importance that SEAR policymakers give to equity as a goal, there has been little empirical analysis of how equitably overall health financing and delivery strategies in each country per- form or of how individual financing mechanisms compare with one 68 · Health Policy Research in South Asia another within the same country. Although many examples of indi- vidual health programs or projects are being evaluated in terms of their impact on equity, these evaluations do not tell policymakers much about the performance of overall national health systems or strategies. Moreover, because changes in one part of a health sys- tem can often have secondary, and unintended, consequences for the operation of other parts of the health system, focusing on sin- gle components of the health system in isolation may not reveal the full equity implications of any program. For example, shifting a segment of the population to health care funded by social insurance might reduce critical political support for maintaining the level of general revenue funding for publicly funded health services. On the other hand, encouraging private provision of health care may contribute to equitable provision if it is successfully combined with targeting public expenditures to the poor. This type of linkage between different elements of the health care system implies that we must examine not only the health system as a whole but also the financing and provision sides in combination. In recent years, the WHO has been in the forefront of emphasiz- ing the importance of these aspects of equity as policy criteria when reforming health care systems, especially when attention has been focused too much on efficiency considerations. For example, publi- cations by the WHO examining the experience with user fees in the 1980s have reminded policymakers and donors of the importance of taking potential negative equity effects into account (Creese and Kutzin 1995). In Europe, the WHO made equity in health services a central goal in the attempt to achieve health for all in the year 2000 (Whitehead 1990). In the Americas, the Pan American Health Organization (PAHO) has stressed the importance of the concept of equity in both health status and the distribution of health services in its work in the region (Shaw 2002). The WHO's own framework for assessing health sector financing reform options has emphasized the importance of collecting adequate and systematic data on the equity of health care systems (Kutzin 1995). Research, such as the European Concerted Action Programme on Quality Assurance study that examined equity in the health care systems of several industrial countries, led to many debates about Equity in Financing and Delivery of Health Services · 69 national differences in the financing of health systems and their policy implications. At the Asia Pacific Health Economics Network (APHEN) conference in Bangkok in April 1997, the representa- tives of several countries expressed considerable interest in collabo- rating on similar work in the Asian region, partly to strengthen intraregional research collaboration and partly to provide a geo- graphic contrast to the studies being carried out in other regions. Countries expressing strong interest in collaborating on a regional Asian study included Bangladesh, Nepal, Sri Lanka, and Thailand. This study is the initial outcome of that collaborative interest. Given the interests of SEAR member states, the general impor- tance attached to the equity goal in health care by the WHO, and the opportunity for comparing this analysis with several other equity studies outside SEAR, this study was designed with the fol- lowing objectives: · To the extent possible using existing data, document empirically equity in delivery of health care services in three SEAR coun- tries (Bangladesh, Nepal, and Sri Lanka) using a national health accounts (NHA) framework--in other words, by examining both public and private delivery systems and, to the extent possible, by addressing vertical and horizontal equity issues raised by the analysis. · To the extent possible using existing data, evaluate equity in the financing of health care services in the three countries, including both government and private methods of payment. · Compare the performance of the existing health care systems and policies in achieving equity in the three countries and draw appropriate lessons about the major health financing mechanisms for other SEAR countries. · Determine the feasibility of this type of system assessment in the data-scarce environments of SEAR countries. Given data and funding constraints, this study focuses on Bangla- desh and Sri Lanka. Only limited resources were available for the Nepalese component of the study, and the lack of an existing NHA 70 · Health Policy Research in South Asia in Nepal prevented any significant new analysis. However, some important work was done to identify the future steps for establishing NHAs in Nepal and to analyze public sector expenditures on health. There is no agreed-upon definition of what constitutes equity with respect to health systems. This lack of a consistent definition stems from underlying differences in philosophical approaches and notions of justice. Distributional objectives in health care can stem from two sources: equity or social justice on the one hand and altruism or caring on the other. In general in health care equity studies, equity objectives have been analyzed independently of any distributional objectives that are motivated by altruism. Two theo- ries of social justice have tended to dominate discussions about equity. The libertarian view holds that analysts should focus on the extent to which people are free to purchase the health care that they want. State involvement should be minimal and limited to providing a minimum standard of care for the poor. However, in practice, most countries' policymakers are concerned with the alternative egalitarian approach to equity. Egalitarians focus on ensuring equality in access to health care and tend to favor a strong role for the state in the financing and provision of health care. In many countries, health care is financed and delivered by a mixture of systems, and there are traces of both ideologies in policymaking, with the emphasis often changing with changes of government (van Doorslaer and Wagstaff 1998). A variant on the egalitarian approach is one whose concern is not with the distribution of health services itself, but with the dis- tribution of income within society. In this variant, a reduction in the inequality of final incomes is regarded as the equity goal. Van Doorslaer, Wagstaff, and Rutten (1993) contended that in indus- trial countries, this type of equity goal is rarely used in arguments about health services. However, it may be more important in devel- oping countries, where public spending is more constrained and where alleviating poverty may be a higher priority than other social objectives (Rannan-Eliya and others 1999). This study empirically evaluates the distribution of health ser- vices, the distribution of the burden of financing of those services, Equity in Financing and Delivery of Health Services · 71 and their net redistributive impact. It thus relates to the concerns of the egalitarian approach for equality of access as well as to the concerns about the distribution of income within a society. In con- trast to other studies, it also places this analysis within the context of an overall analysis of the health financing system using an NHA approach. Country Situations Although all three countries in this study are in South Asia and are low-income economies, they are dissimilar in terms of both their health and socioeconomic indicators (table 4.1). Bangladesh and Nepal are among the poorest low-income countries, while Sri Lanka was on the verge of graduating to lower-middle-income sta- tus in 1997. Nepal is a largely mountainous, landlocked country; Bangladesh's territory consists for the most part of a large delta; and Sri Lanka is an island. A more important difference is that Sri Lanka has health indicators that are more akin to those of an upper-middle-income economy than to a developing country, while Bangladesh and Nepal continue to experience health condi- tions typical of countries at their income level. All three countries are densely populated. Health conditions are relatively poor in Bangladesh; life expectancy was 58 years at birth in 1997, and the infant mortality rate was 90 per 1,000 live births in 1997. Health conditions in Nepal are similar, with a life expectancy of 58 years in 1998 and an infant mortality rate of 93 per 1,000 live births in 1996. In both countries, major public health problems include disorders related to reproductive health and pregnancy, childhood diarrhea, malaria, tuberculosis, and acute respiratory infections. There are significant gender and regional differences in disease prevalence and inci- dence. By contrast, Sri Lanka has achieved both low mortality and low fertility rates. By 1997, despite an income level of less than US$800 per capita, Sri Lanka had reduced its infant mortality rate to 15, its child mortality rate to 18, and its total fertility rate to 72 · Health Policy Research in South Asia Table 4.1 General Indicators for Bangladesh, Nepal, and Sri Lanka, 1997 INDICATOR BANGLADESH NEPAL SRI LANKA Socioeconomic Population (millions) 124 23 18 Area (thousand km2) 130 143 66 GDP per capita (US$) 265 213 773 GNP per capita (US$) 270 210 800 GNP per capita (PPP$) 1,050 1,090 2,460 Gini index 33.6 36.7 30.1 Ratio of population living below poverty level (%) 43 42 25 Rural population (%) 81 88 78 Rate of illiteracy of women (%) 71 72.4 13 Health status Life expectancy at birth of males (years) 58 58 71 Life expectancy at birth of females (years) 59 57 75 IMRa 90 93 15 Maternal mortality rateb 35 54 8 Population Total fertility ratec 3.1 4.4 2.0 Crude birth rated 21 36.9 19 Crude death ratee 5.5 11.6 6 Notes: GDP = Gross domestic product. GNP = Gross national product. PPP$ = Purchasing power parity dollars. aDeaths under age 12 months per 1,000 live births. bDeaths per 10,000 live births. cBirths per woman age 15­49 years. dBirths per 1,000 people. eDeaths per 1,000 people. Sources: Central Bank of Sri Lanka, Annual Report 1998; World Bank, World Develop- ment Indicators 2000 database. below the replacement level of 2.0, and raised its life expectancy at birth to 75 years for women and 71 years for men. Variations in health status between different subgroups of the population are not great, with minimal differences between the urban and rural popu- lations. This performance, notably, has been achieved without expending a higher proportion of its national resources on its health care system than other countries in the region. The national health systems and national health strategies of the three countries are superficially quite similar. All three place government-financed care, with public provision of medical ser- vices, at the center of overall national policy. Government services Equity in Financing and Delivery of Health Services · 73 are generally tax financed, with no significant involvement of social insurance and very limited public sector user fees. At the same time, all three countries permit private sector financing and provision of medical services, and the general government attitude to the private sector is one of laissez-faire. In all three countries, the private sector accounts for a substantial share of overall ambulatory health care provision plus a smaller share of inpatient provision. In general, Sri Lanka is characterized by a greater degree of state involvement in both financing and provision than Bangladesh and Nepal. Table 4.2 provides a brief overview of each country's health system. Methods The tri-country study's primary goal was to examine the overall dis- tribution of financing burden and health care expenditures in each country, paying specific attention to each financing and provision mechanism, while using NHAs as the estimating and synthesizing basis. NHAs were constraining mainly in the analysis of the role of Table 4.2 Background Information on National Health Systems of Bangladesh, Nepal, and Sri Lanka BANGLADESH NEPAL SRI LANKA HEALTH SYSTEM INFORMATION (1997) (1995) (1997) Health spending Total health expenditure (US$ millions, 1997) 313.18 228.92 481.25 Total health expenditure (% of GDP) 3.9 5.45 3.19 Public health expenditure (% of GDP) 1.3 1.58 Total health expenditure per capita (US$, 1997) 10.6 11.5 26 Composition of total health expenditure Public (%) 34 23 50 Private (%) 64 77 50 Health sector facility provision and service utilization Public sector beds per 1,000 capita 0.24 0.15 3.08 Private sector beds per 1,000 capita 0.06 0.07 0.13 Admissions per 100 capita in public sector <1 <1 18 Physician contacts per capita per year 2.0 <2.0 4.5 Sources: Bangladesh Bureau of Statistics 1998; Central Bank of Sri Lanka, Annual Report 1998; Government of Sri Lanka and the Institute of Policy Studies (IPS) 2001; Hotchkiss and others 1998. 74 · Health Policy Research in South Asia household spending, while providing additional detail necessary for the systematic examination of government spending. In effect, the purpose of the study was to determine the distribution of the expen- diture aggregates already identified in the NHAs of each country and thus to extend the use of each country's NHAs to analyze equity characteristics. As explained below, Bangladesh and Sri Lanka had most of the data that we needed, but only Sri Lanka had all of the necessary data. We modified our framework for Nepal to accommodate data inadequacies. However, not all objectives of the study were achiev- able for Nepal, also because, in part, our resources did not enable us to do a more detailed analysis of the available data. Our analysis explored equity issues in the financing and distribution of health spending across the dimensions of income, gender, urban and rural location, and geographic regions. Data: Variables and Assumptions National Health Accounts. NHA estimates for 1997 are the main source of data used in this study for Bangladesh and Sri Lanka. The year 1997 is the most recent year for which reasonably reliable NHA estimates are available for both countries. In each case, NHA estimates had been compiled from several data sources, ranging from audited expenditure records of the government to survey data from private health care providers and pharmaceutical companies. Other reports gave detailed information on the compilation and structure of NHA in the two countries (Data International 1998; IPS 2001). In this study, we attributed to specific beneficiaries all expendi- tures for services provided by each category of provider groups. In addition, the NHA estimates provided expenditures by geographic regions (provinces in Sri Lanka and divisions in Bangladesh). The distribution of government subsidies was thus calculated at a regional level and then aggregated using population weights to obtain the national distribution. Since a complete set of NHAs was not available for Nepal, govern- ment expenditure data were obtained directly from the audited Equity in Financing and Delivery of Health Services · 75 accounts of the Ministry of Health (MOH). Although the Ministries of Defense, Home, Education, and Finance contribute roughly 10­12 percent of total health spending, no disaggregated data were available on their expenditures. Survey data on expenditures by private providers either did not exist or could not be made available for analysis in this study. Given these data restrictions, the Nepal compo- nent of this study examines only the distribution of MOH resources. Facility Cost Studies. It was necessary to carry out facility cost studies to disaggregate budgetary expenditures at the facility level in the public sector between inpatient and outpatient care. In Sri Lanka, this information was provided by the IPS Public Facility Survey 1997, which collected data on a national sample of more than 210 government medical facilities (Somanathan and others 2000). In Bangladesh, the Health Economics Unit/Data International (HEU/DI) Public Facility Survey 1997 was a very similar survey based on the Sri Lankan design, which collected almost identical information from a national sample of 121 government medical facilities (Rannan-Eliya and Somanathan 1999). Household Surveys. Two types of survey were used. General-purpose consumption surveys were used to estimate tax incidence and, in the case of Nepal and Sri Lanka, the distribution of health expendi- tures, utilization rates, and health status. Special household health surveys were also used in Bangladesh to obtain data on the utiliza- tion of health care facilities and household health expenditures. A description of the household surveys used in each country fol- lows in table 4.3. In each case, the following types of information were extracted for analysis in the study: · Total incomes and expenditures of households by income level · Total and disaggregated expenditures by households on medical care · Perceived illnesses or chronic conditions · Type of care sought, type of provider, and frequency of visits 76 · Health Policy Research in South Asia Table 4.3 Overview of Survey Data Used in the Study SAMPLE RECALL INFORMATION SURVEY YEAR (HOUSEHOLDS) PERIOD TYPE OBTAINED Bangladesh MHSS 1997 11,126 14 days and Single-topic Sickness and 90 days for health survey health care sickness and utilization and expenditure expenditure DHS 1996/97 8,682 14 days for DHS IMR by income sickness level HES 1995/96 7,420 1 month for National Household nonfood consumption expenditures expenditure survey Nepal Nepal Living 1995/96 3,338 1 month for National Sickness, Standards sickness and consumption health care Survey expenditure survey utilization, and household expenditures DHS 1996 8,082 14 days for DHS IMR by income sickness level Sri Lanka CSD HIES 1995/96 21,220 Various National Used to (1­4 weeks) consumption estimate tax survey incidence CB CFS 1996/97 8,880 14 days in National Sickness and health module consumption health care survey with utilization health module DHS 1993 9,230 14 days for DHS Data on IMR sickness inequalities Note: CB CFS = Central Bank of Sri Lanka Consumer Finance Survey. CSD = Census and Statistics Department. DHS = Demographic and health surveys. HES = Household expendi- ture survey. HIES = Household income and expenditure survey. IMR = infant mortality rate. MHSS = Morbidity and health status survey. Method of Analysis Estimating the Distribution of Need. The only reliable mortality data by socioeconomic status that we have is for child and infant mortal- ity rates as measured by DHSs in the three countries. Unfortu- Equity in Financing and Delivery of Health Services · 77 nately, we do not have estimates of mortality differences by income level from the Sri Lankan DHS data. Consequently, the recent analyses of DHS data by Gwatkin and others (2000) at the World Bank to estimate inequalities in health status by socioeconomic sta- tus are available only for Bangladesh and Nepal. Estimating the Distribution of Health Care Expenditures. We exam- ined inequalities in the utilization of services simply by looking at the distribution of use of different types of services across the beneficiary groups. We assessed the distribution of health care resources, in particular that of public subsidies, by distributing national health expenditures as estimated in the NHAs across the beneficiary groups. Our analysis estimates the distribution of public subsidies taking regional differences into account. Sri Lanka's NHA system disag- gregates all expenditures by province. The distribution of expendi- tures by provincial councils is straightforward. For support services provided by the central ministry, we imputed expenditures using secondary data. Bangladesh's NHA as originally developed does not provide regional breakdowns of spending. However, we used internal administrative data to allocate all program expenditures to specific divisions for this study. Each program expenditure total was allocated to the relevant type of facility. Data from the facility costing studies were then used to allocate these facility-level expenditures into inpatient and outpa- tient services. We then used survey data to distribute expenditures of individual public programs or activities across population sub- groups, making use of information on relative levels of utilization. We used utilization data from the household surveys to allocate total amounts for each program at the regional level among indi- viduals in the household data set according to their reported levels of utilization of the specific services. This assumes that each unit of service utilized within a region involved the same resource inten- sity. The units of service used for the analysis were outpatient visits and inpatient admissions. 78 · Health Policy Research in South Asia No regional-level analysis was done for distributing expendi- tures across demographic and rural-urban sector groups. Public expenditures on each program were distributed across each benefi- ciary group using the utilization levels of each of the groups at a national level as reported in the relevant household survey. The totals for private expenditures were derived from the NHAs and distributed according to information available from the house- hold surveys. Results Distribution of Health Status and Need As discussed above, we were forced to use data on inequalities in mortality risk as proxied by child mortality rates to infer underly- ing differences in health status and in the need for health services. Tables 4.4 and 4.5 show the extent of inequalities by socioeco- Table 4.4 Infant and Under-Five Mortality Rates by Income Quintile of Equivalent Consumption for Bangladesh, Nepal, and Sri Lanka INCOME QUINTILES BANGLADESH (1997) NEPAL (1996) SRI LANKA (1993) Infant mortality rate 1 96 96 2 99 107 3 97 104 4 89 85 5 57 64 Overall 90 93 25 Under-five mortality rate 1 141 156 2 147 164 3 135 155 4 122 118 5 76 83 Overall 128 139 7 Note: Estimates by income quintile are not available for Sri Lanka. Sources: Gwatkin and others 2000. Sri Lanka DHS 1993. Equity in Financing and Delivery of Health Services · 79 Table 4.5 Infant and Under-Five Mortality Rates by Education of Mother and Place of Residence for Bangladesh, Nepal, and Sri Lanka INDICATOR BANGLADESH NEPAL SRI LANKA Infant mortality rate Education of mother None 98 97 45 Primary 82 80 33 Secondary or higher 65 53 22 Urban-rural residence Urban 73 61 21 Rural 91 95 24 Estate (61) Under-five mortality rate Education of mother None 145 149 62 Primary 112 99 42 Secondary or higher 78 61 27 Urban-rural residence Urban 96 82 26 Rural 131 143 30 Estate (84) Note: Estate mortality rates are estimated from very small samples and thus are not reliable. Sources: Bangladesh DHS 1997; Nepal DHS 1996; Sri Lanka DHS 1993. nomic status in infant and under-five mortality rates in the three countries. As the tables show, overall health status is worse in Bangladesh and Nepal than in Sri Lanka. Within each country, there are clear socioeconomic differences in mortality. In Bangladesh and Nepal, infant and child mortality rates are generally 40 percent less in the richest quintile than in the poorest quintile. We do not have the equivalent data for Sri Lanka, but the general picture is that mor- tality rates become lower with increasing socioeconomic status, both in terms of income and educational level. Children whose mothers have secondary education or higher in all three countries experience only half the mortality risks of children whose mothers have no education. However, note that children in Sri Lanka whose mothers have no education still experience lower mortality risks 80 · Health Policy Research in South Asia than mothers with secondary education or higher in Bangladesh and Nepal, and than mothers in the richest quintile in both of the other countries. Because the incomes of households in the highest quintiles in both Bangladesh and Nepal are probably higher than the incomes of Sri Lanka households whose mothers have no edu- cation, it is likely that children of poor women in Sri Lanka have other environmental advantages. In general, these data confirm that health status is worst in the lower socioeconomic groups and among the most disadvantaged, and that, correspondingly, the need for health services is greatest among these groups. As discussed above, we decided not to use reported sickness as a direct measure of actual need for health care as we expected self- report health status to be a poor proxy for actual health status. This assumption was confirmed in the household data for our three countries. In all three countries, contrary to the evidence that mortality rates are worse for those at lower socioeconomic levels, reported sickness rates are no higher in the poorer income deciles than in the richest deciles. In Bangladesh, reported sickness rates are rela- tively equal across the income range, while they increase modestly with increasing income in both Nepal and Sri Lanka. To varying extents, higher socioeconomic status in all three countries is associ- ated with a higher propensity to report sickness and presumably to be aware of having an actual illness. This runs completely counter to the objective evidence of worse health status at lower socioeco- nomic levels. The second notable feature of these data is that reported sick- ness across the three countries varies in the opposite direction to mortality rates. Sri Lankans with mortality rates close to those of European nations are twice as likely to report being sick as Bangladeshis at all income levels. Taking into account the fact that the Nepal survey used a 1-month recall period in contrast to the 14-day recall periods used in the Bangladesh and Sri Lanka sur- veys, the likelihood of a respondent reporting an episode of illness in any given time period in Nepal is probably lower than in Bangladesh. Equity in Financing and Delivery of Health Services · 81 The differences in reported sickness by age are not the same in Bangladesh and Sri Lanka. While the reported sickness rates in Sri Lanka are lowest among older children and young adults, as one would expect, reported morbidity rates are generally the same across all adult age groups in Bangladesh. There may be a common link between these patterns. The health transition is generally associated with increasing awareness of and sensitivity to illness. Sri Lanka, with its low mortality rates, is the most advanced of the three countries in its health transition and thus has the most illness-sensitive population. Sri Lankans report more sickness than Bangladeshis or Nepalese. At the same time, the health transition is an ongoing phenomenon. Younger Bangladeshis may simply be more sickness-sensitive than older Bangladeshis because they have benefited more from recent cul- tural shifts favoring increased consciousness of illness, which may be countering the likely higher underlying morbidity in the older adult age groups. Similarly, in Sri Lanka, the existence of higher sickness rates at higher income levels is an indicator that significant socioeconomic differences remain in the sensitivity to illness. Distribution of Health Services and Resources Inequalities in Utilization of Services. In Bangladesh and Sri Lanka, there are no substantial differences between the poor and the rich in terms of their likelihood of seeking outside treatment if they have reported being sick. In Sri Lanka, however, because richer individuals are more likely to report being sick, the overall utiliza- tion of outside treatment increases with income. There are significant differences in the pattern of use when treatment outside the home is sought. In Sri Lanka, 87­92 percent of all care is sought from modern qualified providers, with no dif- ference between income levels, compared with only 15­42 percent in Bangladesh. In Bangladesh, most people from all income deciles seek care from nonmodern or nonqualified providers, and there is a significant socioeconomic difference, with higher-income individu- als being more likely to use modern qualified providers and much 82 · Health Policy Research in South Asia more likely to use inpatient services. Although making direct com- parisons of the levels of utilization is hazardous given different lev- els of recall bias in each survey, the data also show the much lower rate of utilization of inpatient services by Bangladeshis at all income levels. This is consistent with administrative data showing that admission rates per capita in Sri Lanka are approximately 10 times higher than in Bangladesh. Figures 4.1 and 4.2 illustrate the patterns in use of public sector providers by sick people who resort to use of modern qualified providers. Sri Lankans in all income deciles are more likely to use public providers than are those in the corresponding deciles in Bangladesh, whether for outpatient or inpatient services. In Sri Lanka, poor respondents are more likely than better-off people to Figure 4.1 Use of Public Outpatient Providers in Bangladesh and Sri Lanka Percentage of visits 100 90 80 70 60 50 40 30 20 10 0 1 2 3 4 5 6 7 8 9 10 Income Decile Bangladesh Sri Lanka Sources: Bangladesh Bureau of Statistics MHSS 1997; CB CFS 1996/97. Equity in Financing and Delivery of Health Services · 83 Figure 4.2 Use of Public Inpatient Providers in Bangladesh and Sri Lanka Percentage of visits 100 90 80 70 60 50 40 30 20 10 0 1 2 3 4 5 6 7 8 9 10 Income Decile Bangladesh Sri Lanka Sources: Bangladesh Bureau of Statistics MHSS 1997; CB CFS 1996/97. rely on public provision. This distinction is not evident in Bangladesh. In both countries, public providers are more likely to be used for inpatient care than for outpatient care, while in Sri Lanka, people in the poorest deciles rely almost wholly on public providers for inpatient services. Inequalities in the Distribution of Benefits. Table 4.6 shows the esti- mated incidence of government subsidies by decile. We estimated them by using the distributional information in the household sur- veys to allocate the distribution of subsidies. In distributing subsi- dies for outpatient services, we assumed that the average subsidy per visit was identical for all public sector visits in a given province or division. Similarly, we distributed subsidies for inpatient services 84 · Health Policy Research in South Asia Table 4.6 Distribution of the Benefits of Government Expenditures on Personal Medical Services in Bangladesh and Sri Lanka (percent) OUTPATIENT SERVICES INPATIENT SERVICES INCOME DECILES BANGLADESH SRI LANKA BANGLADESH SRI LANKA 1 9.1 10.1 5.1 9.2 2 9.9 9.5 6.4 10.4 3 9.8 10.4 9.6 11.6 4 10.7 10.9 8.6 6.9 5 10.3 11.7 7.7 9.1 6 10.3 9.7 11.2 10.4 7 10.2 10.7 10.2 11.4 8 10.3 8.9 9.3 12.0 9 9.9 9.3 15.7 9.9 10 9.5 8.9 16.3 9.0 Overall amount Tk. 5.2 billion Rs. 2.7 billion Tk. 8.4 billion Rs. 5.9 billion Concentration index ­0.00022 ­0.0732 0.1633 ­0.0506 Note: Tk. = Takas. Rs. = Rupees. Sources: Bangladesh Bureau of Statistics MHSS 1997, CB CFS 1996/97. by decile on the assumption that the average subsidy per admission was identical within a given province or division. As discussed above, Sri Lankans are more likely than Bangla- deshis to report sickness, to use modern providers, and, when doing so, to use public providers. In addition, there is a clear socioeconomic distinction in terms of who chooses to use public providers, with the poor making more use of them than the rich do, especially in the case of outpatient services. This trend in using public health care services runs opposite to the trend in reported sickness in Sri Lanka. Consequently, the distribution of the bene- fits of government spending on health is relatively equal across all income deciles in Sri Lanka for both inpatient and outpatient ser- vices. However, the incidence in the case of outpatient subsidies is slightly more pro-poor than for inpatient services, and is reflected in a slightly more negative concentration index for outpatient ser- vices than for inpatient services.2 This probably reflects the fact that middle-income and rich people in Sri Lanka have fewer options to switch to private provision in the case of inpatient ser- vices than in the case of outpatient services. Equity in Financing and Delivery of Health Services · 85 In Bangladesh, the lack of any such socioeconomic distinctions in the choice between public and private providers means that peo- ple in all income deciles benefit equally from public outpatient service subsidies, but that the rich benefit much more than those in lower income deciles in the case of inpatient services. The latter pattern can be related to the extensive use of all modern providers and of all inpatient services by the rich. In contrast to Sri Lanka, the rich in Bangladesh do not tend to use private inpatient services, so their high overall utilization rate means that the rich benefit more from government spending than people in lower income deciles, which results in positive values for both of the concentra- tion indexes. Table 4.7 gives the overall incidence of all government subsidies in both countries and the related concentration indexes. The con- centration index for all services is 0.0544 in Bangladesh, which indicates that the overall distribution of benefits favors the rich. Table 4.7 Distribution of the Benefits of Government Expenditures on Collective Services and on All Services in Bangladesh and Sri Lanka (percent) COLLECTIVE SERVICES ALL SERVICES INCOME DECILES BANGLADESH SRI LANKA BANGLADESH SRI LANKA 1 10.0 10.4 7.4 9.8 2 10.0 10.4 8.3 10.2 3 10.0 11.4 9.7 11.3 4 10.0 9.8 9.6 8.7 5 10.0 10.2 9.0 10.0 6 10.0 9.7 10.6 10.0 7 10.0 10.4 10.2 10.9 8 10.0 9.7 9.7 10.6 9 10.0 9.1 12.6 9.5 10 10.0 8.8 12.8 8.9 Overall amount Tk. 4.0 billion Rs. 2.5 billion Tk. 17.6 billion Rs. 11.1 billion Concentration index 0.0000 ­0.0771 0.0544 ­0.0670 Note: Tk. = takas. Rs. = rupees. The distribution of collective services is estimated at the national level for Bangladesh and at the provincial level for Sri Lanka. Sources: Bangladesh Bureau of Statistics MHSS 1997, CB CFS 1996/97. 86 · Health Policy Research in South Asia For Sri Lanka, the index is ­0.0670, indicating a pro-poor distribu- tion of all patient subsidies. The available data did not enable us to analyze the distribution of subsidies by income level in Nepal. However, it was possible to estimate the relative incidence of government health spending in all three countries by region. These results are given in table 4.8. In all three countries, government health expenditures are highest in the region containing the capital city. However, a notable feature is the extent of variation between regions. In Bangladesh and Sri Lanka, funding varies from ­20 percent to +40 percent of the national average of all regions. However, in Nepal the funding lev- els in the peripheral regions are only one-tenth the level of funding in the capital region. The cause of this great disparity appears to be the concentration of national and teaching facilities in the capital city of Nepal, which is not balanced by any other investment in other areas. Overall, the distribution of health subsidies in Sri Lanka is pro- poor, whereas in Bangladesh and Nepal it is pro-rich. Figure 4.3 contrasts the estimated concentration indexes for Bangladesh and Sri Lanka with those of some other countries and a selection of states within India. The majority of developing countries for which estimates are available are characterized by pro-rich distributions of government health subsidies. Bangladesh is typical in this respect, and in fact its pro-rich bias is less marked than in many other developing economies, in India as a whole, and in its neigh- Table 4.8 Variations in Government Expenditures on Health by Region NEPAL DEVELOPMENT BANGLADESH DIVISION REGION SRI LANKA PROVINCE (TK. PER CAPITA, 1996/97) (RS. PER CAPITA, 1998/99) (RS. PER CAPITA, 1997) Khulna 113 Western 7 Northwestern 432 Sylhet 117 Midwestern 10 Uva 447 Rajshahi 117 Far Western 10 Sabaragamuwa 474 Chittagong 120 Eastern 57 North-Central 547 Barisal 126 Central 198 Southern 547 Dhaka 196 Northern-Eastern 550 Central 553 Western 804 Equity in Financing and Delivery of Health Services · 87 Figure 4.3 Benefit Incidence of Government Health Expenditures-- International Comparisons Concentration index 0.65 0.55 India India 0.45 Guinea Bihar, India India 0.35 Rajasthan, India Bengal, Pradesh, 0.25 India Ghana Indonesia West Vietnam Nadu, 0.15 Peru Bulgaria Andhra China Kenya Africa India Rica Tamil Bangladesh Lanka 0.05 South Kerala, Honduras Sri Chile Costa Malaysia Uruguay ­0.05 ­0.15 ­0.25 ­0.35 Source: Based on chart provided by Anil Gumber using data compiled by David Peters of the World Bank. The Bangladesh and Sri Lanka data are from this study. boring Indian state of West Bengal. Sri Lanka, on the other hand, appears to be one of a small group of countries, such as Costa Rica, Honduras, Malaysia, and Chile, and the state of Kerala in India, with a pro-poor incidence of government health subsidies. Most of the latter group of countries are usually identified as superior health performers, with health indexes that are better than expected given their national income levels. This is almost certainly not coincidental. All the latter group of countries in the chart have mixed public-private provision and financing of health care. With the exception of Chile, none explicitly targets its government health expenditures to the poor, emphasizing instead universal access to government-funded health services. Targeting in Malaysia, Costa Rica, and Kerala appears to similar to that in Sri Lanka--the rich are encouraged to use private services, thus reserving public services 88 · Health Policy Research in South Asia for poorer individuals. The resulting distribution of government subsidies remains pro-poor because the poor are as likely to use modern health services as the rich, as in Sri Lanka. This is not the case in Bangladesh, where the poor are less likely than the rich to think of themselves as sick and, when seeking medical care, are less likely to seek out modern qualified medical providers. Distribution of Payments for the Financing of Health Services Figure 4.4 shows the sources of funding used in each health system. All three countries rely primarily on government financing and out-of-pocket spending. There is no significant social insurance financing in any of the three countries, and private insurance is all but negligible except in Sri Lanka. Even in Sri Lanka, private insurance accounts for less than 2 percent of total health care Figure 4.4 Sources of Funding in the National Health Systems of Bangladesh, Nepal, and Sri Lanka Bangladesh (1997) Nepal (1995) Sri Lanka (1997) 0 10 20 30 40 50 60 70 80 90 100 Percent Taxes/donors Firms Commercial insurance Nonprofits Out-of-pocket Sources: Data International 1998; Hotchkiss and others 1998; IPS 2001. Equity in Financing and Delivery of Health Services · 89 financing. Sri Lanka relies mostly on government financing, with household spending accounting for less than half of total health care financing. In Bangladesh and Nepal, household spending finances most health care expenditures. In both Bangladesh and Sri Lanka, private insurance typically reimburses patients' out-of-pocket expenditures, so the distribu- tion of household payments in the survey data includes such pay- ments. Consequently, to examine the overall distribution of payments for health care, it is sufficient to analyze the distribution of payments by households in the form of tax payments and out-of- pocket expenditures for health. The distribution of payments in the form of households' out-of- pocket spending on health care was relatively easy to derive from the available household survey data. The incidence of tax payments across households was more difficult to estimate, but we attempted it par- tially for both Bangladesh and Sri Lanka. The difficulty arose because of the large share of overall taxes in each country that are not direct taxes such as income or retail consumption taxes. The Sri Lankan analysis is based on an analysis of data from the Census and Statistics Department Household Income and Expenditure Survey (CSD HIES) 1995/96, while the Bangladeshi estimates are derived from results published in Ensor, Hossain, and Miller (2001). Their results are based on analysis of household survey data for 1984/85, but we assumed that there has been little change in the indexes since then, because the tax structure has remained substantially unchanged. Distribution of Payments in Sri Lanka. Table 4.9 gives the estimated distribution of household payments for health care, in the form of both out-of-pocket payments and taxes, for Bangladesh and Sri Lanka. The distributions of tax payments and out-of-pocket spend- ing are relatively similar in Sri Lanka. Overall, both mechanisms are, in fact, progressive in Sri Lanka, with richer households con- tributing more than their share of total household consumption. The top quintile makes more than 50 percent of overall payments to the health care system, while the bottom quintile makes less than 5 percent of overall payments. 90 · Health Policy Research in South Asia Table 4.9 Distribution of Payments for Health Care for Each Source in Bangladesh and Sri Lanka by Income Decile (percent) BANGLADESH SRI LANKA INCOME DECILES TAXES OUT-OF-POCKET TAXES OUT-OF-POCKET 1 4 5 2 2 2 6 6 3 2 3 6 6 4 4 4 7 7 5 5 5 8 8 6 6 6 9 9 7 7 7 10 10 9 9 8 12 11 11 11 9 14 15 15 17 10 25 23 36 38 Overall amount Tk. 18.7 billion Tk. 34.4 billion Rs. 14.05 billion Rs. 13.9 billion Share of national funding (%) 34 64 50 48 Note: Tk. = takas. Rs. = rupees. Taxes as a share of national funding refers to both taxes and donor contributions. Four percent of national health expenditures are accounted for by nonprofits and firms in Sri Lanka that are not included above. Sources: Data International 1998; IPS 2001. Table 4.10 gives the estimated Kakwani and Suits indexes for each mechanism. A positive Kakwani index indicates a progressive payment with redistribution, while a negative index indicates the opposite. Only tobacco taxes are estimated to have negative Kak- wani and Suits indexes (Kakwani 1977). This finding points to a potential conflict between the goal of progressive financing of health care and the goal of improving health status through dis- couraging tobacco use. Increasing tobacco taxes is desirable from the perspective of reducing tobacco consumption, but these taxes are regressive in practice. This fact suggests that a policy of increasing taxes on tobacco should be accompanied by more con- certed efforts to change smoking behavior in the poorest house- holds if it is not to have regressive tax implications. An important feature of the two main financing mechanisms in Sri Lanka is that out-of-pocket payments are more progressive Equity in Financing and Delivery of Health Services · 91 Table 4.10 Kakwani and Suits Indexes for Different Payment Mechanisms for Health Care in Sri Lanka PAYMENT MECHANISM KAKWANI INDEX SUITS INDEX Income tax 0.53 0.77 Capital taxes 0.46 0.67 Sales taxes 0.05 0.07 Petrol tax 0.50 0.74 Liquor tax 0.15 0.21 Tobacco tax ­0.03 ­0.05 All taxes 0.091 0.134 Out-of-pocket spending 0.548 0.803 All payments 0.099 0.145 Note: The Kakwani index for overall payments was estimated by assuming that the inci- dence of remaining taxes was similar to that of those for which estimates had already been made. Source: Sri Lanka HIES 1995/96. than tax payments. However, the data on utilization patterns sug- gest that this is not due to the inherent progressivity of out-of- pocket financing. Rather, it is the outcome of a general reliance by the poor on subsidized, mostly free health care provision by the government and the switch to unsubsidized private provision by people in the higher-income deciles. Public and private provision combine in Sri Lanka to produce a system in which voluntary pri- vate payments for health care are largely progressive. The overall Kakwani index for the health care system is 0.099, which indicates progressivity in overall payments for health care. Distribution of Payments in Bangladesh. As shown in table 4.9, in Bangladesh the general distribution of tax payments is quite similar to that of out-of-pocket payments for health care. This finding reflects the general fiscal reliance on indirect and consumption taxes, as in Sri Lanka. However, the distribution is not as skewed toward payments by the rich as in Sri Lanka. For example, those in the top quintile make less than 39 percent of both tax payments and out-of-pocket payments for health care, in contrast to the more than 51 percent paid by those in the top quintile in Sri Lanka. 92 · Health Policy Research in South Asia Unlike in Sri Lanka, out-of-pocket spending is more regressive than tax funding. This reflects the fact that the rich do not neces- sarily use private providers, as is the case in Sri Lanka. Poor people in Bangladesh are just as likely to use private providers as rich peo- ple and, as a consequence, pay a higher proportion of overall pri- vate financing than in Sri Lanka. The overall Kakwani index for Bangladesh is ­0.072, indicating that the health care financing sys- tem as a whole is regressive. Redistributive Impact of Government Financing of Health Services Tax financing is progressive in Sri Lanka but seems to be slightly regressive in Bangladesh. The same can be said about government provision: government subsidies in Sri Lanka benefit the poor as much as the rich, whereas in Bangladesh they benefit the rich slightly more than the poor. To the extent that health status is worse among the poor in both countries, health spending is clearly not equitable in relation to need in Bangladesh and almost cer- tainly not so in Sri Lanka (see tables 4.11 and 4.12). Table 4.11 Redistributive Impact of Government Taxation and Financing of Health Care Services in Sri Lanka, 1995/96 (by income decile) SHARE OF TAXES AND BENEFITS (%) NET BENEFITS AS PERCENTAGE INCOME OF HOUSEHOLD DECILES TAXES BENEFITS RS. MILLIONS EXPENDITURE 1 2.5 9.8 1,026 4.88 2 3.5 10 909 3.07 3 4.5 10.9 908 2.68 4 5.2 8.8 512 1.32 5 6.1 9.9 546 1.24 6 7.3 9.7 329 0.64 7 8.7 10.6 271 0.44 8 11.2 10.7 ­77 ­0.10 9 15.0 10.1 ­699 ­0.73 10 36.0 9.5 ­3,725 ­1.96 Overall 100 100 0.0 0.0 Sources: Estimated from CSD HIES 1995/96 (tax payments); CB CFS 1996/97 (subsidies). Equity in Financing and Delivery of Health Services · 93 Table 4.12 Redistributive Impact of Government Taxation and Financing of Health Care Services in Bangladesh, 1996/97 (by income decile) SHARE OF TAXES AND BENEFITS (%) NET BENEFITS AS PERCENTAGE INCOME OF HOUSEHOLD DECILES TAXES BENEFITS TK. MILLIONS EXPENDITURE 1 4.5 7.4 515 1.25 2 5.6 8.3 471 0.77 3 6.2 9.7 626 0.84 4 6.7 9.6 502 0.57 5 7.7 9.0 220 0.21 6 8.5 10.6 373 0.31 7 10.0 10.2 34 0.03 8 12.1 9.7 ­424 ­0.26 9 13.9 12.6 ­227 ­0.11 10 24.7 12.8 ­2,090 ­0.49 Overall 100 100 0.0 0.0 Source: Estimated from CSD HIES 1995/96 (tax payments); CB CFS 1996/97 (subsidies). Nevertheless, government financing and provision of health care in Sri Lanka has a net redistributive impact. Those in the poorest decile pay less than 3 percent of total taxes but capture almost 10 percent of total government spending on health care services. In contrast, those in the richest decile pay 36 percent of total taxes but receive less than 10 percent of total government health subsidies. The net gain for the poorest households from government financ- ing and provision is significant, equivalent to 5 percent of their gross household consumption before receipt of subsidies. Conclusions This study has attempted to assess equity in the distribution of financing, health care expenditures, and health status in three countries within an NHA framework. Because of data limitations, we were able to carry out the full analysis for only two: Bangladesh and Sri Lanka. Although a lack of resources prevented us from car- rying out a full analysis for Nepal, we did determine that sufficient data exist in that country both to construct NHAs in the future and 94 · Health Policy Research in South Asia to conduct the empirical analyses that we carried out for the other two countries. We used the existing NHAs for Bangladesh and Sri Lanka to estimate the overall distribution of health care expenditures. Meth- ods developed elsewhere for analyzing equity in the financing and delivery of health care services can be easily adapted to an NHA framework. NHAs can, in general and with few new data require- ments, be extended to analyze the distributional aspects of health system financing. As this can provide a more rigorous comparison of private and public expenditures than other methods, it also allows for a more reliable assessment of the overall equity of the health system and more consistent comparisons of different financ- ing and provision mechanisms. The study confirmed our original expectation that household survey data on self-reported health status do not provide a reliable measure of underlying need. In all three countries, the poor suffer from higher mortality rates yet are less likely than the nonpoor to report suffering from any sickness in surveys. Controlling for self- reported ill health is thus not, in itself, sufficient to determine whether the distribution of health expenditures is meeting the underlying need for services. Therefore, we did not attempt to standardize health care expenditures directly according to any measure of underlying health status. Nevertheless, the weak but reasonable assumption that the poor are in worse health than the nonpoor is sufficient to support the conclusion that overall health care resources are not equitably distributed in relation to need in any of the countries we studied. Socioeconomic differences in perceived sickness appear to be a major contributor to differences in the amount of health care households use. The poor in all three countries appear to be less likely than the nonpoor to report episodes of illness, although less so in Sri Lanka. Not acknowledging sickness means that the poor use less medical care than they may actually need. In Sri Lanka, there is no difference between the poor and the nonpoor in their readiness to seek care from qualified modern medical providers when reporting themselves sick, but such a difference does exist in Equity in Financing and Delivery of Health Services · 95 Bangladesh. The poor in Bangladesh are less likely than the rich to use a modern medical provider when they report being sick, and, when they do so, they are no more likely to choose public providers than the rich. In Sri Lanka, although rates of utilization of all modern providers (both public and private) are more or less equal among people of all income levels, the poor are more likely than the nonpoor to choose public providers. Consequently, gov- ernment subsidies favor the poor in Sri Lanka but not in Bangladesh. Using concentration indexes as a numerical measure to compare the pro-poor incidence of government health spending across countries reveals that that the pro-rich incidence of subsidies in Bangladesh is not unusual; in fact, it is less pro-rich than in many Indian states. However, some countries, including Sri Lanka, and some Indian states, such as Kerala, do manage to achieve a pro- poor distribution of government subsidies, indicating that there is room for improvement in Bangladesh. Examining the differences in government expenditures by region revealed some inequalities. In all three countries, subsidy expendi- tures are highest in the region that contains the capital city; this is largely due to a higher concentration of teaching and national spe- cialist hospitals in the capital cities. However, it is notable that the disparity in government expenditures between the capital region and other regions is much greater in Nepal than the other two countries. In Bangladesh and Sri Lanka, the peripheral regions all receive at least 80 percent of the national average, but in Nepal, some districts receive only one-tenth of the per capita subsidies received by other districts. The large geographical disparities in subsidy levels in Nepal are probably associated with a much greater pro-rich bias in overall government spending than in Bangladesh. In the three countries, the health care systems are all essentially financed through just two mechanisms: (a) general revenue taxa- tion, which is used to fund mostly free, government health care services, and (b) out-of-pocket payments by households, which are mostly used to purchase medical care from the private sector. The extent of state financing and provision varies among the countries, and is highest in Sri Lanka, lowest in Nepal. 96 · Health Policy Research in South Asia In Bangladesh and Sri Lanka, the rich make greater payments through both mechanisms than the poor. In Sri Lanka, payments by the rich are greater for both mechanisms than their share of household consumption, making both mechanisms progressive in their impact. Overall, out-of-pocket payments are somewhat more progressive in Sri Lanka than taxation. However, this is a conse- quence of the availability of free government services funded through taxation and used predominantly by poorer groups, with richer individuals voluntarily choosing to use private providers. Most taxes in Sri Lanka appear to be progressive, but tobacco taxes are regressive. The poor pay more in tobacco taxes than their share of overall household consumption. This points to potential con- flicts between different health policy goals. In Bangladesh, the distribution of payments through both taxa- tion and household out-of-pocket spending is similar to the distri- bution in Sri Lanka, but the rich make smaller payments in relation to their ability to pay or their overall consumption than the poor. Consequently, both tax payments and household out-of-pocket financing are regressive financing mechanisms in Bangladesh. In contrast to Sri Lanka, out-of-pocket payments are more regressive than tax payments. This reflects the lack of any sorting between public and private provision by income level. Although taxation is generally regressive in Bangladesh, this does not imply that the overall fiscal impact of government health spending on the poor is negative. Since government health subsi- dies are less unequally distributed than tax payments, the net impact of tax financing of health care is to redistribute income in favor of the poor. In Bangladesh, those in the poorest quintile receive a net fiscal gain equivalent to 1 percent of their presubsidy household consumption through the combined impact of taxes and government health subsidies. In Sri Lanka, where the overall distribution of government health subsidies is actually pro-poor and taxation is progressive, the net fiscal gain for those in the poorest quintiles is much greater. The poorest quintile receives a gain of 45 percent of gross household consumption. At least in the Sri Lanka context, government financing of health care does con- Equity in Financing and Delivery of Health Services · 97 tribute significantly to overall poverty reduction and income redistribution. Notes 1. This study was funded by a grant from the South-East Asia Regional Office (SEARO) of the World Health Organization, for which we express our gratitude to Dr. Than Sein and Dr. Shambhu Acharya. The study itself was conducted by a team of researchers from Data International in Dhaka, Institute of Policy Studies in Sri Lanka, and the Nepal Health Economics Association, led by Ravi Rannan-Eliya and James Killingsworth. This final report was pre- pared by Tahmina Begum, Tamara Dorabawila, Badri Raj Pande, Ravi Rannan-Eliya, and Aparnaa Somanathan. We express our thanks to our many colleagues in each of the three countries from whose work and input we have benefited in conducting this study. 2. A value of zero for the concentration index denotes an equal distribution of benefits across all people. A negative index denotes a distribution that favors the poor, and a positive index denotes a distribution that favors the rich. References Alailima, Patricia J., and Faiz Mohideen. 1983. "Health Sector Commodity Requirements and Expenditure Flows." Unpublished report. National Plan- ning Department, Colombo, Sri Lanka. Bangladesh Bureau of Statistics. 1998. "Report on Survey of Private Health Service Establishment 1997/98." Ministry of Planning, Government of Bangladesh, Dhaka. Central Bank of Sri Lanka. Various years. Annual Report. ------. 1999. Report on Consumer Finances and Socioeconomic Survey, Sri Lanka 1996/97. Colombo. Central Bureau of Statistics, National Planning Commission. 1996. Nepal Living Standard Survey Report, Kathmandu. 98 · Health Policy Research in South Asia Center for Education and Technical Studies. 1993. Private Sector Health Expendi- tures in Nepal, Kathmandu. Creese, Andrew, and Joseph Kutzin. 1995. Lessons from Cost-Recovery in Health. Geneva: World Health Organization. Data International. 1998. Bangladesh National Health Accounts, 1996/97. Health Economic Unit, Ministry of Health and Family Welfare, Dhaka. Department of Census and Statistics. 1995. Sri Lanka Demographic and Health Sur- vey 1993. Colombo. ------. 2000. Sri Lanka Household Income and Expenditure Survey 1995/96. Colombo. Department of Health Services, Ministry of Finance, His Majesty's Government. Various years. Annual Report. Colombo, Sri Lanka. Ensor, Tim, Atia Hossain, and Nigel Miller. 2001. Funding Health Care in Bangladesh: Assessing the Impact of New and Existing Financing. HEU Research Paper No. 24. Health Economics Unit, Dhaka. Government of Sri Lanka and the Institute of Policy Studies of Sri Lanka. 2001. Sri Lanka National Health Accounts. Ministry of Health, Colombo, Sri Lanka. Gwatkin, Davidson, Shea Rutstein, Kiersten Johnson, Rohini P. Pande, and Adam Wagstaff. 2000. Socioeconomic Difference in Health, Nutrition, and Population in Bangladesh. Washington, D.C.: World Bank. Hotchkiss D. R., J. J. Rous, K. Karmacharya, and P. Sangraula. 1998. "Household Health Expenditure in Nepal: Implications to Health Care Financing Reform." Health Policy and Planning 13(4): 371­83. IPS (Institute of Policy Studies). 2001. Sri Lanka National Health Accounts: Sri Lanka National Health Expenditures 1990­99. Colombo. Kakwani, N. C. 1977. "Measurement of Tax Progressivity: An International Com- parison." Economic Journal 87(343): 71­80. Kutzin, Joseph. 1995. Experience with Organizational and Financing Reform of the Health Sector. Geneva: World Health Organization. Meerman, Jacob. 1979. Public Expenditure in Malaysia: Who Benefits and Why. New York: Oxford University Press. Equity in Financing and Delivery of Health Services · 99 Ministry of Finance, His Majesty's Government. 1990­1999. Government Budget (Red Book). Various issues. Kathmandu. ------. 1990­1999. "Budget Speech by the Hon'ble Minister for Finance." Vari- ous years. Kathmandu. ------. 1999/2000. Economic Survey, Fiscal Year 1999/2000. Kathmandu. Mitra, S. N., Ahmed Al Sabir, Anne R. Cross, and Kanta Jamil. 1997. Bangladesh Demographic and Health Survey 1996/97. National Institute of Population Research and Training (NIPORT), Bangladesh: Mitra and Associates, Bangladesh; Macro International, USA. National Planning Commission. 1997. The Ninth 5-Year National Plan (1997­2002). Nepal. National Planning Commission/UNICEF/UNDP. 1985. Restructuring Budget and Aid in Nepal, 1986­97. Nepal. Rannan-Eliya, P. Ravi, and Aparnaa Somanathan. 1999. Bangladesh Health Facility Efficiency Study Report. IPS HPP Occasional Paper No. 12. Colombo: Institute of Policy Studies. Rannan-Eliya, P. Ravi, D. Senagama, D. Weerakoon, and H. Aturupane. 1999. Monitoring the 20/20 Compact on Budget and Aid Restructuring in Sri Lanka. Health Policy Series Paper No. 1. Colombo: Institute of Policy Studies. Shaw, R. Paul. 2002. "World Health Report 2000 Financial Fairness Indicator: Useful Compass or Crystal Ball." International Journal of Health Services 32(1): 192­203. Somanathan, Aparnaa, Kara Hanson, Tamara Dorabawila, and Bilesha Perera. 2000. Operating Efficiency in Public Sector Health Facilities in Sri Lanka: Measure- ment and Institutional Determinants of Performance. Small Applied Research Paper No. 12. Bethesda, Md.: Partnerships for Health Reform Project, Abt Associates, Inc. van Doorslaer, Eddy, and Adam Wagstaff. 1998. Measuring and Testing for Inequity in the Delivery of Health Care. ECuity Working Paper No. 12. Rotterdam: Erasmus University. van Doorslaer, Eddy, Adam Wagstaff, and Frans Rutten, eds. 1993. Equity in the Financing and Delivery of Health Care: An International Perspective. Oxford: Oxford University Press. 100 · Health Policy Research in South Asia Whitehead, Margaret. 1990. The Concepts and Principles of Equity and Health. WHO Discussion Paper, EUR/ICP/RDP414. Copenhagen: WHO Regional Office for Europe. WHO (World Health Organization). 2000. The World Health Report 2000--Health Systems: Improving Performance. Geneva. World Bank. 1999. World Development Report 1998/99: Knowledge for Development. Washington, D.C. ------. 2001. World Development Report 2000/01: Attacking Poverty. Washington, D.C. CHAPTER 5 Geographic Resource Allocation in Bangladesh Tim Ensor, University of York Atia Hossain, Priti Dave Sen, Liaquat Ali, Shamin Ara Begum, and Hamid Moral, Ministry of Health and Family Welfare, Bangladesh Abstract One of the core objectives of the Health and Population Sector Program of Bangladesh is to improve the health status of the most vulnerable groups--particularly the poor, women, and children. The current system of geo- graphic resource allocation may be an impediment to attaining this goal. Resources are currently allocated to localities largely according to the size of inpatient facilities and number of staff in those areas. This practice leads to wide differences in per capita allocations across districts for both the revenue and development budgets. Analysis suggests that the allocations have no significant positive relationship to health needs and may even be inversely related to the general deprivation of an area. Low-, middle-, and high-income countries around the world with centrally managed public health care systems attempt to adjust simple per capita allocations by the age-sex struc- ture of the local population and aggregate measures of health, such as infant and standardized mortality rates. 101 102 · Health Policy Research in South Asia Three basic principles emerge from the literature on these international efforts: · Measures of need are of necessity selective and approxi- mate. · The allocation method and variables representing need should be transparent. · Allocation targets may be defined by an objective princi- ple, but the speed at which the targets are approached is a political choice. Existing statistics could be used in introducing a system of needs-based resource allocation in Bangladesh. Simple preliminary simulations suggest that resources would be moved to areas where mortality is highest. These simula- tions are mainly illustrative; a final formula could be devel- oped only through a consultative process that determines which statistical proxies are most suitable for measuring the health needs of the vulnerable groups for which the Health and Population Sector Program was ultimately designed. Introduction The Government of Bangladesh, through its Health and Popula- tion Sector Program (HPSP), is aiming to improve the health sta- tus and health care access of the most vulnerable groups in society, particularly women, children, and the poor (MOHFW 1998b). To achieve its goal, the government has created an Essential Services Package consisting of services known to be most needed by the vul- nerable groups, and it has been increasing the resources devoted to these services and the facilities through which they are delivered. At the end of the first two years of its operation, the HPSP can claim some success, having increased the proportion of public spending going into essential services to 60­70 percent in fiscal 2000 (MOHFW 2000). Yet surveys continue to show that the poor Geographic Resource Allocation in Bangladesh · 103 still find it difficult to obtain high-quality essential services (CIET- Canada 2000). Many factors affect the extent to which large national allocations for essential health services reach vulnerable groups. One factor, the subject of this paper, is how well the allocations are targeted to those most in need.1 If the spending is not well targeted--that is, if allocations to different areas of the country are not in relation to the need and the size of vulnerable groups--then the failure of such spending to improve the condition of vulnerable groups will not be surprising. Indeed, government health spending in Bangladesh is allocated on the basis of the existing capacity for service delivery and norms, rather than on measures of health status and vulnerability. The allocation system is thus not well suited to efforts to improve the status of particular groups. This paper begins by describing the current system of resource allocation and its effect on spending at district and upazila (subdistrict) levels. The second half of the paper gives examples of resource allocation methods that are more closely related to health need. Allocation of Government Health Spending in Bangladesh Budgets for most categories of regular government expenditure, including health, are determined centrally. The system can best be characterized as one of line budgets based on capacity and histori- cally determined norms, that is, on physical capacity--as measured by the number of facilities, staff, or beds (see table 5.1)--and on funding provided to the same facilities in previous years. Historic patient flows are taken into account in setting the budgets for food. The use of line budgets means that resources are allocated for specific line items (such as food, medicines, and medical supplies), and a transfer between lines, whether by those at the top or the bottom of the administrative structure, is not possible, unless it is within the same broad economic code (for example, within the staff pay and allowance category or within the supplies category). local local use when fund; direc- the and FINANCIAL vises vises vises general OF from wise, MANAGEMENT general, AUTHORITY surgeon surgeon surgeon budget surgeon, super tender super tender super of met contingency other tor CMMU, PWD AND SOURCE Civil Civil Civil Director Civil staff General. UNION 75,000 other n.a. Tk. n.a. 4 n.a. gas. Director = and DG fund staff 2,000) linens, tment. TION Tk. AZILA Depar UP 30 15,000 25,000 nurses, other than doctors, reagents, DESTINA Tk. Tk. Tk. 9 10 23 Less contingency (about orks W bandages, Public fund = staff 2,000) Tk. PWD than DISTRICT 30 22,000 40,000 doctors, nurses, other instruments, Unit. Tk. Tk. Tk. 11 27 30 Less contingency (about surgical Facilities n facility tance Maintenance BASIS Public decision spending patter includes and to TION capacity impor bed/per DG ALLOCA bed-day Per op-downT from Historical ehicleV balance Utilization Political facility the Funding Per · · · · · · Per Submissions Management of drugs; percent Allocation 1,000 Construction costing Tk. renovation) 70 = applicable. 5.1 other than (equipment, than Not CMMU a = fuel, less ableT More ITEM Operations Food MSR Maintenance, Staff Capital construction, n.a. a. Note: 104 Geographic Resource Allocation in Bangladesh · 105 Current Methods of Allocation The budget-setting process begins in August. Civil surgeons sub- mit a list of requirements to the director general's office. (For the basis of the civil surgeons' calculations, see table 5.1.) Budgets are consolidated by the director general's office and sent to the Min- istry of Health and Family Welfare (MOHFW) and the Ministry of Finance for approval. Ultimate ratification is by Parliament in the first week in June, when the consolidated budget for the entire government is presented. Thus, districts have little flexibility over the use of funds. The constraints extend to a requirement that 70 percent of the listed drugs be procured from the Essential Drugs Company, a govern- ment-owned institution. The only real flexibility is in the use of the maintenance budget, from which the civil surgeon may spend up to Tk. 5,000 for vehicles and up to Tk. 1,500 for building repairs using local contractors. Maintenance expenditures exceeding these amounts are managed by other administrative units. Total Budget Allocations for Districts and Upazilas Facilities are distributed nearly equally across administrative areas according to a norm of 31 beds for upazila health complexes and 50, 100, or 200 beds for district hospitals. Given the unevenness of population across administrative areas, the near-equal distribution of facilities means that the distribution of beds per capita is highly unequal. Moreover, although district and upazila facilities provide both inpatient and outpatient care, the allocation is determined solely by reference to the numbers of beds, bed-days, and staff and takes no account of the outpatient load. The distribution of beds and staff essentially determines the level of revenue funding for each area (table 5.1). In fiscal 2000, a 50-bed district hospital might have received an allocation of around Tk. 7 million and an upazila facility Tk. 3.5 million to 5 million (table 5.2). More than two-thirds of this allocation is for staffing. On average, the government funds one clinic (a union health and family welfare center [UHFWC]) for every three 106 · Health Policy Research in South Asia Table 5.2 Typical Budgets for a District Hospital and an Upazila Health Complex UNITS TOTAL COST PERCENT OF ITEM UNIT COST (TAKA) (TAKA) FACILITY BUDGET District hospital (50 beds) Staff (annual salary) Doctors 104,000 11 1,144,000 16.4 Nurses 74,000 27 1,998,000 28.6 Other 42,000 30 1,260,000 18.0 Staff total n.a. n.a. 4,402,000 63.0 Supplies Food (per bed-day) 30 18,250 547,500 7.8 MSR (per bed) 22,000 50 1,100,000 15.7 Maintenance (per year) 40,000 1 40,000 0.6 Other (per year) 900,000 1 900,000 12.9 Supplies total n.a. n.a. 2,587,500 37.0 Total n.a. n.a. 6,989,500 100 Upazila health complex (31 beds) Staff (annual salary) Doctors 104,000 9 936,000 25.9 Nurses 73,000 10 730,000 20.2 Other 40,000 23 920,000 25.4 Staff total n.a. n.a. 2,586,000 71.5 Supplies Food (per bed-day) 30 11,315 339,450 9.4 MSR (per bed) 15,000 31 465,000 12.9 Maintenance (per year) 25,000 1 25,000 0.7 Other (per year) 200,000 1 200,000 5.5 Supplies total n.a. n.a. 1,029,450 28.5 Total n.a. 3,615,450 100 n.a. = Not applicable. Note: Hospitals assumed to be operating at full capacity with all posts filled. Source: For salaries, Quayyum and Howlader 2000. unions, at around Tk. 240,000 per clinic, which is mostly for the salaries of the support staff and four paramedics. In addition, the government funds another six or seven field staff health workers in each union.2 Finally, each district's civil surgeon's office, which provides administrative support for the entire district and also delivers a public health function, costs an average of approximately Tk. 7.5 million. For the average district, with 6 upazilas and 24 unions, these allocations total roughly Tk. 52 million (table 5.3). Geographic Resource Allocation in Bangladesh · 107 Table 5.3 Distribution of Spending on Public Health in an Average District by Type of Facility PERCENT OF DISTRICT TYPE OF FACILITY UNIT COST (TAKA) UNITS COST (TAKA) SPENDING District hospital 6,989,500 1 6,989,500 13.4 Civil surgeon's officea 7,500,000 1 7,500,000 14.4 Upazila health complex 3,615,450 6 21,692,700 41.7 UHFWCb 88,200 18 1,587,600 3.1 Field staff 39,600 360 14,256,000 27.4 Total 52,025,800 100 Note: Average district contains 6 upazilas and 24 unions. UHC is upazila health complex; UHFWC is union health and family welfare center. a. Average budget allocation for district with five to seven upazilas (Bangladesh Health Bulletin 1997). b. MSR allocation of Tk. 75,000 plus four staff. The result of the current allocation process is that, in principle, budgets and expenditures bear little relation to either the size of the population or the number of patients treated (as measured by the number of admissions and outpatient consultations). If, for example, the occupancy levels were to fall to only 50 percent, the budgetary allocations would, again in principle, fall by only around 3 percent, and most of that decline would reflect lower food requirements. In practice, the director general does have some flex- ibility through the selective filling of vacancies, and vacancies of doctor positions of 35 percent or more exist in many facilities (Quayyum and Howlader 2000). The vacancy rate may exaggerate the degree of flexibility available, however, because many areas, particularly upazilas, are unattractive to doctors and nurses, who generally prefer to work in urban facilities. District Allocations. Just as the number of beds per capita varies enormously across administrative areas, so does the spending: apart from Dhaka, it ranged in fiscal 2000 from Tk. 54 in Gazipur to Tk. 256 in Bandarban. At more than Tk. 426, spending per capita in Dhaka is the highest, but Dhaka provides much of the specialist care for the whole country. 108 · Health Policy Research in South Asia Variations in spending are apparent both within and between divisions (table 5.4). Within divisions, spending tends to be higher in districts with medical college hospitals. The higher spending reflects the training function of these institutions and the fact that they serve patients from surrounding districts. Yet even outside such districts, substantive variations persist; the maximum alloca- tion to a district exceeds the minimum by about 50­70 percent in most divisions and by more than 300 percent in Chittagong.3 Assessing the fairness of health spending relative to the many dimensions of need is a complex task. One way to analyze the ques- tion is to compare current spending patterns with summary measures of health outcomes. A simple correlation between a district's infant mortality rate (IMR) and its expenditures yielded no significant posi- Table 5.4 Range of Public Health Spending per Capita across Districts by Division and Presence of a Medical College Hospital (MCH), Fiscal 2000 (Taka) DIVISION AND DISTRICT GROUP AVERAGE MINIMUM MAXIMUM Rajshahi All 89.34 66.52 176.12 Without an MCH 78.79 66.52 113.99 Khulna All 85.76 67.03 138.70 Without an MCH 79.87 67.03 97.74 Barisal All 90.73 61.25 153.18 Without an MCH 78.24 61.25 87.63 Dhaka (excluding Dhaka district) All 83.14 54.50 170.36 Without an MCH 75.05 54.50 92.95 Chittagong All 121.69 59.17 255.73 Without an MCH 122.21 59.17 255.73 Sylhet All 86.67 62.12 147.24 Without an MCH 66.48 62.12 71.10 Total 90.13 54.50 255.73 Geographic Resource Allocation in Bangladesh · 109 tive association (r2 = 0.03). A similar correlation with the age-gender standardized mortality rate also was not significant (r2 < 0.01). The human development index generated by the United Nations Development Programme (UNDP) offers a more sophis- ticated measure of general development by combining per capita measures of infant mortality, literacy, and gross domestic product (GDP) in a single index (in which 1 represents the highest level of development, 0 the lowest). A computation of the index for each district of the country (UNDP 1996) suggests that districts at a lower level of development receive lower funding (figure 5.1). The picture is slightly more complex, however. Those districts with the highest rating (above 0.43) receive the highest allocation per capita (an average of Tk. 118). Districts with the lowest rating (less than 0.3) receive around Tk. 93, and the middle districts receive Tk. 83. Figure 5.1 The Relationship between the U.N. Human Development Index and Public Spending per Capita in Bangladesh, 1996 Public expenditure per capita 300 250 200 150 100 50 0 0.2 0.3 0.3 0.4 0.4 0.5 0.5 0.6 0.6 Human Development Index Source: UNDP 1996. 110 · Health Policy Research in South Asia The overall conclusion is that, when measured against broad indi- cators of health outcome and general human development, patterns of health spending at the district level do not appear to reflect need. Allocations to Upazilas. At the upazila level, resources are allocated to inpatient facilities, outpatient care, and community services. Among the 369 upazilas outside large urban areas, we find that total allocations in fiscal 2000 averaged Tk. 13.7 million per upazila (figure 5.2). The size of the allocation an upazila receives depends primarily on the number of staff, because that is the explicit basis for the allo- cation formula. But much of the variation is also attributable to the size of the upazila's population, probably because allocations, par- ticularly under the development budget, reflect a calculation of Figure 5.2 Distribution of Total Public Spending on Health by Upazila, 1999/2000 Levels of spending 50 Std. Dev. = 0.51 40 Mean = 1.37 N = 369.00 30 20 10 0 0.00 0.25 0.50 0.75 1.00 1.25 1.50 1.75 2.00 2.25 2.50 2.75 3.00 Distribution Geographic Resource Allocation in Bangladesh · 111 local health need, which in turn is related to population. The posi- tive relationship of allocation to population, at least at the subdis- trict level (per capita allocations vary more widely at the district level) is an encouraging finding; it suggests that any move toward a formula based in part on population, as we propose below, would not be too out of line with current practice.4 A Needs-Based Allocation Formula Here we present a formula for allocating resources according to measures of need. The scenarios are illustrative; in practice, any formula would require substantial consultation and development. Moreover, at least two issues would require careful consideration before the adoption of any particular formula--maintaining fiscal discipline and treating the distinction between the revenue budget and the development budget appropriately. The issue of fiscal discipline arises because a needs-based alloca- tion system would not use norms based on line items, a move that would seem to threaten financial control and accountability. To pre- serve sound budgetary planning and financial reporting under a needs-based system, each district should still submit its plans for spending by line item at the budget-planning stage. Once the budget has been adopted and the fiscal year has begun, large deviations from a line item might be allowed, but only if adequately explained. The preparation of district budgets in this way is closely related to the development of local-level planning now being piloted in a number of districts. Skills in local planning can be considered a prerequisite to the successful development of a needs-based alloca- tion system. A second issue is the current division in allocations between the revenue budget and the development budget. Logical reasons exist for treating recurrent and capital spending differently when under- taking resource allocation, but maintaining the distinction in both the revenue budget and the development budget complicates the creation of needs-based formulas. One option would be to develop 112 · Health Policy Research in South Asia three budgets: one for revenue, another for the recurrent portion of the development budget, and a third for capital. Background to the Use of Allocation Formulas Allocating resources for public health care according to historic convention and capacity-based norms is common to many centrally managed public health services. A small but growing international literature has focused on ways of reorienting resource allocation so that systems become more responsive to local health needs. Coun- tries as diverse as Canada, South Africa, Zambia, the United King- dom, and Australia now allocate resources on the basis of need as defined by the size and characteristics of the population (Birch and Chambers 1993; Bourne and others 1990; Gilbert, Gibbeerd, and Stewart 1992; Lake 2000; Mays 1995). Although the details of each formula are different, a number of common guiding principles emerge from the literature. Measures of Need Are Necessarily Selective and Approximate. A formula can only partially capture the complex pattern of local circum- stances that determine health status and health service needs in a particular area. Unlike simple measures of capacity, needs formulas are indirect measures. Any method for allocating resources is inevitably an estimate, but a formula more obviously attempts to measure the whole by way of its parts. Therefore, a formula is more open to question and debate than, say, a simple measure of capacity. Need should be measured using accepted general criteria based on statistics that are regularly collected for each district, and the allocation formula should produce results that are consistent with the need measures. In this paper we offer some examples of how resources might be allocated in Bangladesh using common meth- ods of measuring health need. Ultimately, however, the variables chosen should be subject to open debate on their suitability in rep- resenting need before they are formally adopted. The Allocation Method and Variables Representing Need Should Be Transparent. Because a new method of calculation is certain to be Geographic Resource Allocation in Bangladesh · 113 disputed, as it will lead to some significant changes in resource allocation patterns, it must be as simple and transparent as possible. A strategy that several countries, including Zambia and the United Kingdom, have adopted is to introduce a relatively simple formula and later increase its sophistication as data and experience develop. More sophistication can help account for complex socio- economic and health-related determinants of demand for services. But care must be taken because sophistication is also likely to mean that fewer will understand and critique the methodology. A more subtle problem arises when apparent sophistication cov- ers up less objective criteria for choices. In the United Kingdom, an initially clear and transparent formula introduced during the 1970s was complicated during the 1980s when an apparently sophisticated statistical method was introduced to adjust for rela- tive differences in morbidity between areas (Carr-Hill 1990). The new formula had the effect of shifting resources to those areas most supportive of the government and led to charges that the changes were politically inspired. The methods used were later shown to be scientifically flawed. A simple but transparent formula may have much greater chance of winning general acceptance and reduces the chance that flawed methods will be obscured by complexity. Allocation Targets May Be Defined by an Objective Principle, but the Speed at Which the Targets Are Approached Is a Political Choice. The method of final allocation can be determined by a formula that is as objective as possible; in contrast, the introduction of such a for- mula and the speed of its implementation is mostly a political, not a scientific, choice. The transition can be made quickly (within a few years) or extended over a longer period. An immediate transition is probably not advisable as local services will take time to adapt to smaller or larger allocations. Toward a Needs-Based Formula for the Geographic Allocation of Resources The factors affecting the need for health services in a particular area can be divided into two components: (a) those explained by 114 · Health Policy Research in South Asia the size and structure of the population and (b) those explained by other health and socioeconomic factors.5 Size and Structure of the Population. The most important factor deter- mining the need for health service in an area probably is the size of the population. The age and gender structures of the population are, together, perhaps the second most important factor. In most coun- tries, including Bangladesh, the need for health services tends to be highest for the very young (less than one year old), women of child- bearing age, and the elderly. Public spending by age group partly illus- trates these tendencies (figure 5.3), but utilization may reflect need more accurately. These demographic factors vary little across districts. Other Health and Socioeconomic Factors. Differential health needs also exist within individual age-gender groups. Some groups are more Figure 5.3 Public Spending on Health by Age Group in Bangladesh, 1998 Spending per capita (taka) 700 600 500 400 300 200 100 0 0­4 5­9 10­14 15­25 25­34 35­44 45­59 60­69 70+ Age group Source: MOHFW 1998a. Geographic Resource Allocation in Bangladesh · 115 vulnerable to disease. A strong link is frequently found between poverty and inequality on the one hand and mortality and morbid- ity on the other. Locational factors, such as proximity to rivers, also may also predispose certain groups to higher levels of diseases such as malaria. Identifying a suitable indicator of disease that can be used to weight resource allocations is not an easy task. One way would be to compare differential patterns of morbidity within given age- gender groups. The problem with this method is that statistics can record only those episodes of illness that are reported to the health system. Most reporting systems are biased in this respect. The use of hospital statistics, for example, biases the results toward those with access to hospitals. Household surveys can yield information on actual disease prevalence, but the sample is often not suffi- ciently large or wide to provide a guide to differences between dis- tricts and subdistricts. A good proxy for morbidity that is less likely to be biased, used in both the United Kingdom and South Africa, is the standardized mortality rate (SMR).6 The recording of mortality should be rela- tively free of bias in that a statistical system is required to record all deaths. For populations of more than 100,000, the SMR is often a good predictor of actual morbidity (Carr-Hill 1990). One study has questioned the reliability of Bangladesh's mortal- ity statistics and has suggested that deaths in the country are under- reported (Streatfield 2000). For the purposes of an allocation formula, the consistency of reporting across areas is more impor- tant than whether such reporting is accurate in an absolute sense. Nonetheless, for Bangladesh, the infant mortality rate (IMR) may be a more reliable mortality indicator than the SMR; hence, calcu- lations based on the IMR are also provided here. Although the IMR emphasizes infant deaths rather than the general burden of disease, much of the general mortality rate in fact consists of infant and maternal deaths. Given that current health policy concentrates on improving infant and maternal health, the bias of the IMR may be acceptable. 116 · Health Policy Research in South Asia The Costs of Care. The cost of providing services to a defined popu- lation may vary from area to area. The factors determining these costs are complex and include regional differences in staffing costs, differential transport costs, and variations in prices of basic foods.7 Ideally, all these factors should be included in a medical price index that can be used to adjust allocations. Bangladesh does not have such an index (nor do many other countries). One factor affecting cost is location. Per capita costs of care tend to rise in sparsely populated areas because of increased transport costs and treatment times. Part of this cost may be borne by patients through longer journeys to health facilities and part by the health system through the provision of smaller facilities scattered across the area. To represent the location effect, and in the absence of other local measures of cost, population density has been used to construct an index for cost. Although density captures part of the cost differen- tial, it fails to account for other cost drivers, such as variations between areas in salary levels or costs of supplies. One justification for the use of population density is that national pay scales and cen- tral procurement of many supplies mean that costs of inputs are broadly similar across the country. Some further work is required to indicate whether these costs are indeed similar. Resource Allocations Allocations for each area are obtained by using the mortality rate (indicator of need) and population density (indicator of location- related costs) to adjust the population of each district up or down. The adjusted population is then used as a basis for proportional allocations. A higher mortality rate and lower density than average each will cause the population to be adjusted upward and will result in larger proportional allocations. The converse occurs for districts with lower-than-average needs and costs. For any district i, the total target allocation is given as the product of need and cost fac- tors, as follows: Allocationi = PerCap × POPi × (1 + ai) × (1 + ni) × (1 + c) Geographic Resource Allocation in Bangladesh · 117 where PerCap is the national budget (excluding Dhaka City) divided by the total national population, POPi is the population of district i, a is the district age-gender adjustment, n is the needs adjustment, and c is the cost adjustment.8 For the country as a whole, each of these fac- tors is zero by definition. The sign on each indicator denotes whether its value is above (+) or below (­) the national average. Adjusted populations and corresponding allocations were calcu- lated using the above formula and available data. For the district of Madaripur, for example, the adjusted allocation becomes: AllocationMadaripur = PerCap × POPMadaripur × 1.09 × 0.69 × 0.96 These values indicate that the district has an above-average pro- portion of the very young and very old, but a below-average SMR and index of cost (higher population density). For the purposes of the simulations below, Dhaka district is excluded. The per capita allocation to Dhaka is several times larger than that to the next largest district. In addition, many travel to Dhaka from other districts to receive treatment for diseases requir- ing specialized inpatient care (see section on cross-boundary flows below). Using both the SMR and IMR as indicators of need, the simula- tions lead to significant increases in allocations to Chittagong (10 percent increase) and Sylhet (47 percent increase) (figure 5.4). In contrast, the target allocation for Rajshahi declines about 18 per- cent. No significant change is recorded for Dhaka (not including Dhaka district) and Khulna. All divisions, however, contain some significant gainers and losers. Cross-Boundary Use of Services An objection to the needs-based formula is that it does not take account of "cross-boundary flow"--that is, care provided in some districts, especially those with medical college hospitals or tertiary facilities, to residents of other districts. To investigate the magnitude of cross-boundary flows, we sur- veyed approximately 1,100 inpatients and outpatients in five medical 118 · Health Policy Research in South Asia Figure 5.4 Cumulative Effect of Weighting on Fiscal 2000 Division Allocations for Health in Bangladesh by Weighting Factor Crore taka 300 250 200 150 100 50 0 Current Capitation Age/sex SMR SMR + cost IMR + cost Rajshahi Div. Dhaka Div. Khulna Div. Chittagong Div. Barisal Div. Sylhet Div. Note: SMR = Standardized mortality rate. IMR = Infant mortality rate. 1 crore = 10 million. facilities--three district hospitals and two medical college hospitals. The patients were receiving care across a range of specialties. The results suggest that, with the exception of the Dhaka medical college hospital, these facilities are predominantly used by those residing in the hospital's district (table 5.5). More than 90 percent of both inpatients and outpatients came from the same district for the three hospitals surveyed. Indeed, between 60 percent and 80 per- cent of outpatients and between 20 percent and 50 percent of inpa- tients came from the upazila in which the hospital is located. The use of the Mymensingh medical college hospital is similar to that of the district hospitals, with more than 75 percent of inpa- tients and almost 100 percent of outpatients coming from the Mymensingh district. More than 50 percent of inpatients at the hospital came from the Sadar upazila. Geographic Resource Allocation in Bangladesh · 119 Table 5.5 Inpatient and Outpatient Use of District Hospital (DH) and Medical College Hospital (MCH) Facilities, by Selected District and Proximity of Patient Residence to Facility (percent) PATIENT STATUS AND RESIDENCE RELATIVE CHITTAGONG DHAKA KISHORJANJ MANIKGANJ MYMENSINGH TO FACILITY DH MCH DH DH MCH TOTAL Inpatients Sadar (city) 20.0 15.3 38.5 52.0 50.0 32.7 In district 94.4 41.8 96.2 98.0 75.0 72.6 In adjacent districts 3.3 22.0 3.8 0.0 22.0 14.5 Other 2.2 36.2 0.0 2.0 3.0 12.9 Number of 90 177 78 50 164 559 inpatients Outpatients Sadar (city) 59.0 41.9 77.1 68.3 79.3 62.1 In district 95.1 81.9 99.0 98.4 98.9 92.6 In adjacent districts 0.0 9.7 1.0 1.6 1.1 3.9 Other 4.9 8.4 0.0 0.0 0.0 3.5 Number of 61 155 96 63 87 462 outpatients The Dhaka medical college hospital is used relatively widely by patients from outside Dhaka district, especially inpatients. Only about 40 percent of inpatients and about 80 percent of outpatients come from the Dhaka district. More than half the inpatients com- ing from outside Dhaka district (36 percent of all outpatients) come from districts beyond those adjacent to Dhaka, particularly for urology services. The survey is indicative rather than representative, and further work would be required to investigate whether these patterns are repeated elsewhere in the country. If they are representative, then they suggest that little need exists for significant financial compen- sation for treating cross-border patients except for the Dhaka facil- ity and perhaps other medical college hospitals, the Mymensingh results notwithstanding. (Perhaps that facility was surveyed at an unusual time.) A prerequisite to adjusting for cross-boundary flows, then, is to adequately measure them in an ongoing way by, for example, adding district of residence to the data collected on patients in the 120 · Health Policy Research in South Asia hospital's management information system. With information on the size of cross-boundary flows, the issue could be tackled in sev- eral ways. One way would be to make a greater national allocation to districts offering such services. Another way, given that facili- ties tend to be used by patients from the same division, is to allo- cate resources to each division on the basis of weighted capitation and then permit divisions themselves to make an adjustment for cross-boundary flows by reallocating some of the resources from each district to the division-level facilities (mostly medical college hospitals). Another important aspect of cross-boundary flow is the influ- ence of major roads on the case mix and volume of activity in cer- tain hospitals. Casual observation indicates that some district and upazila facilities situated close to main roads are frequently crowded, predominantly with trauma cases resulting from traffic accidents. Many of these accident victims are from outside the area. Some allowance for this factor will be required when refining the formula, and doing so would again be aided by adding relevant information to hospital admissions data, especially the particular cause of admission. Formula versus Special-Needs Allocations One potential disadvantage of a formulaic approach is that it does not take into account needs in certain areas that are not reflected in the regular statistical reporting--for example, the effects of peri- odic excessive flooding or of epidemic or endemic diseases. Although special allocations for particular vertical programs are probably not to be encouraged, especially in the context of a sec- tor-wide approach, an argument could be made for reserving part of the annual allocation for disbursements apart from those calcu- lated by the main formula. Taken further, a separate formula could be devised for allocating specific elements of the budget. This is indeed the approach in the United Kingdom, where separate for- mulas are used to allocate funding for acute care, care for the men- tally ill, and community health services (Peacock and Smith 1995). Geographic Resource Allocation in Bangladesh · 121 In summary, to attain the objective of transparency, it is sug- gested that one formula be used for the majority of the allocation. If special allocations are desired, then a small part of the budget should be retained for allocation outside the formula. Introducing Needs-Based Allocations A weighted needs-based allocation formula along the lines of the one described above could not completely replace current methods overnight. In the United Kingdom, for example, moving all alloca- tions to within a few percent of the weighted targets took about 15 years. Services need a transition period to adapt to the prospect of increased or reduced allocations. The actual length of the transi- tion period is a political decision. One method of phasing in a resource allocation formula is through the allocation of real (that is, inflation-adjusted) budget growth. In this method, the needs-based allocation to any division or district would be based on the historic allocation, with addi- tional funding from real budgetary growth for areas that lag behind their needs-based per capita targets. The speed of the change would depend on the rate of economic growth. If, for example, the health budget grows at a real rate of 3 percent per year, all divisions could be within 5 percent of the target weighted allocations within five years (figure 5.5). Within divisions, adjustment would take longer unless real reductions were made in some district allocations to permit intradivision reallocations of funding. This method shifts that share of the total budget. The gradual implementation and the reallocation of real as opposed to nominal growth in the budget permit allocations to continue to rise in all divisions, but with vary- ing rates of increase. Practical implementation of a needs-based budget can take a number of forms, depending on the level of local decentralization. Both options described below require reallocations to be planned in advance, preferably over a three- to five-year period. The fore- casting could be modeled within the framework of a medium-term financial strategy for the health sector. 122 · Health Policy Research in South Asia Figure 5.5 Potential Transition of Actual Spending to Needs-Based Health Spending Targets in Bangladesh by Division Percentage from target 30 20 10 0 ­10 ­20 ­30 ­40 ­50 ­60 1 2 3 4 5 6 Year Rajshahi Div. Dhaka Div. Khulna Div. Chittagong Div. Barisal Div. Sylhet Div. Option 1. Normative Budgeting within Needs-Based Allocations-- District and Upazila Levels. Under this approach, the MOHFW, in collaboration with the Ministry of Finance, allocates funds to dis- tricts on the basis of a needs analysis similar to the one proposed above. District and upazila budgets would then be composed using normative methods, but within the overall limits of the needs- based allocation.9 The normative methods could be dropped sooner in divisions with a longer transition to needs-based district allocations. Limitations on transfers between line items could con- tinue as a way of ensuring administrative control. This option may be the most feasible for the short or medium term given the lack of delegated financial powers or decentralized management decisionmaking. Geographic Resource Allocation in Bangladesh · 123 Option 2. Local-Level Planned Budgets. A more sophisticated option is for the district to be notified of its needs-based budget in advance, but for district and upazila managers to decide on what activities and services should be financed. The planning process would not be strictly constrained by norms, but instead would fol- low a justification based on the needs of the community. Financial control during the year would still be strictly imposed, and local- level planners would have to spend within the budgeted line items. The system could be integrated with the local-level planning process already started in a number of districts. In fact, both approaches--local budgeting and normative budget- ing within needs-based allocations--could be developed simultane- ously. Some districts could plan their own budget structure, while spending in others could be determined at the national level. But spending under each approach would be constrained by overall district-based resources. Allocation of Capital An important objection to a standard formula based on current population health care needs is that it does not adequately account for past excesses or deficiencies of investment in the health sector in a particular area. Allocations to one area that are greater than the needs-based allocation might be justified now on the grounds that they would make up for deficient levels of resources allocated historically. An answer to this objection is that the resource alloca- tion formula described above is aimed at the allocation of the recurrent elements of the budget, for both revenue and develop- ment. The allocation of funding for capital development is a more complicated issue precisely because explicit account must be taken of past levels of investment. Doing so would require the assessment and valuation of facilities, and future allocations would take into account the capital stock and investments made in priority areas. Capital allocation is a complex task, and further work would be needed should the current system be regarded as unsatisfactory. In the meantime, however, a system based on need is less likely, rather than more likely, to perpetuate inequalities of historic 124 · Health Policy Research in South Asia investments. Historic norms tend to base recurrent allocations on the current size of facilities. This automatic link to the past is bro- ken in moving toward a needs-based system. Summary The level of public finance for health care provided to each geo- graphic area in Bangladesh is largely determined by the size of facil- ities and employment of staff, rather than by health status or socioeconomic need of the population. Such a situation is common to many centrally managed and publicly funded health care systems. This historical approach to budgeting has two fundamental weaknesses. First, it tends to thwart the objective of allocating resources to the most vulnerable population groups and may pre- vent national equity objectives from being carried down to the local level. Second, because allocations are not related to the level of activity in the facilities but rather to their capacity, it does noth- ing to reward a greater level of efficiency. The system can, there- fore, be both inequitable and inefficient. The advent of the sector-wide approach to budgeting provides an ideal opportunity to begin changing the system of allocation. Under the previous project-oriented approach, changing the allo- cation system would have meant changing dozens of separate budget allocations with specific purposes. The gradual unification of allocations under the revenue and development budgets allows the allocation of resources through a small number of channels in a more equitable way. The local-level planning initiative provides a further opportu- nity to develop needs-based allocations. The initiative requires dis- tricts and, later, upazilas to decide priorities according to local needs. The process is not likely to increase the need to decentralize budget decisionmaking because it will increase the effectiveness of the local-level planning and prioritization process. The process is also likely to draw attention to apparent inequities of the current funding system. Geographic Resource Allocation in Bangladesh · 125 Reforming the system of geographic resource allocation could profoundly improve the effectiveness of resource use by channeling more resources to needy areas and reducing the incentive to pre- serve capacity at the expense of addressing local needs. Notes 1. Another factor is the efficiency with which services are deliv- ered at the local level, an important subject for future investigation. 2. Most of the staff under the Directorate General of Family Planning are excluded because they were funded under the devel- opment budget during 1999/2000. Beginning with 2000/01, these workers fell under the revenue budget. 3. In most districts, the government's development allocations closely match the allocations made under the revenue budget (cor- relation coefficient of +0.89, p < 0.01). The development budget includes donor funding through the government but not direct program aid. Information on the latter is not available by district. 4. Our analysis also suggests that, holding population and other factors constant, upazilas in Syhlet division generally receive lower allocations than upazilas in other divisions. 5. The data used in this section were those readily available from sample surveys, the 1991 census, and regular management infor- mation system reports from hospitals. 6. Mortality rates can be standardized by comparing the observed rate in an area with the rate that would be expected given the local age structure and national age-gender specific mortality rates. Use of unadjusted mortality rates is inaccurate because a dis- trict may record a higher mortality rate as a result, for example, of a larger elderly population, a condition which itself indicates a rela- tively successful health care system rather than relative deprivation. 7. In Bangladesh, as in other centrally planned systems, it could be argued that, because items such as staff wages are nationally determined, local variation does not occur. This is true in terms of the impact on expenditures. However, a hidden cost appears in 126 · Health Policy Research in South Asia remoter areas, where staff are reluctant to work and positions are often not filled. In the end the cost is borne by patients, in terms of lower quality care, rather than the health system. 8. This formula is replicated from Peacock and Smith (1995). 9. Some norms may need to be relaxed to stay within a needs- based allocation. For example, a district facing a reduced allocation may not have the resources to fill the entire complement of posts required by the facility normative. References Birch, S., and S. Chambers. 1993. "To Each According to Need: A Community- Based Approach to Allocating Health-Care Resources." Canadian Medical Asso- ciation Journal 149(5): 607­12. Bourne, D., W. Pick, S. Taylor, D. McIntyre, and J. Klopper. 1990. "A Methodol- ogy for Resource Allocation in Health Care for South Africa: Part 3: A South African Health Resource Allocation Formula." South African Medical Journal 77: 456­59. Carr-Hill, R. 1990. "RAWP Is Dead: Long Live RAWP." In A. J. Culyer, A. K. Maynard, and K. W. Posnett, eds. Competition in Health Care: Reforming the NHS, London: Macmillan. CIET-Canada. 2000. Service Delivery Survey: Second Cycle, 2000 Preliminary Find- ings. Report for the Health and Population Sector Program. Dhaka. Gilbert, R., B. Gibbeerd, and J. Stewart. 1992. "The New South Wales Resource Allocation Formula: A Method for Equitable Health Funding." Australian Health Review 15 (June 21). Lake, S. 2000. "Modeling `Need' in Health Sector Resource Allocation Formulae: Application in a Low-income Country." Health Systems Financing in Low- Income African and Asian Countries. CERDI, Clermont Ferrand. Mays, N. 1995. "Geographical Resource Allocation in the English National Health Service, 1971­1994: The Tension between Normative and Empirical Approaches." International Journal of Epidemiology 24(S1): S96­S102. MOHFW (Ministry of Health and Family Welfare). 1998a. Bangladesh National Health Accounts 1996/97. Heath Economics Unit and Data International, Dhaka. Geographic Resource Allocation in Bangladesh · 127 ------. 1998b. Project Implementation Plan. Health and Population Sector Program, Dhaka. ------. 2000. Public Expenditure Review of the Health and Population Sector Program, 1999­2000. Research Paper 19. Health Economics Unit and Management Accounting Unit, Dhaka. Peacock, S., and P. Smith. 1995. The Resource Allocation Consequences of the New NHS Needs Formula. Discussion Paper 134, Centre for Health Economics. York, U.K.: University of York. Quayyum, Z., and S. R. Howlader. 2000. Output and Costs of Health Care Provided at District Hospitals: A Case Study in Two Districts of Bangladesh. Dhaka: Dhaka University, Institute of Health Economics. Streatfield, P. K. 2000. Status of Performance Indicators. International Center for Diarrheal Disease, Bangladesh. Report prepared for Program Co-ordination Cell and World Bank for the mid-term review of the Bangladesh Health and Population Sector Program, Dhaka. UNDP (United Nations Development Programme). 1996. Report on Human Devel- opment in Bangladesh. Dhaka. SECTION III Expenditure Analysis CHAPTER 6 Public Expenditure Review of the Health and Population Sector Program in Bangladesh Tim Ensor, University of York Atia Hossain, Priti Dave Sen, Liaquat Ali, Shamin Ara Begum, and Hamid Moral, Ministry of Health and Family Welfare, Bangladesh Abstract The chapter summarizes findings from a public expendi- ture review of the health sector for the 1999/2000 financial year, which marked the second year of the Health and Pop- ulation Sector Program (HPSP) for Bangladesh. The review examines the extent to which the program was con- sistent in size and breakdowns with planned budgets and priorities. During the first two years of the HPSP, spending on health fell significantly short of approved budgets. Restoring real per capita funding to 1996 levels would require a 12.4 percent increase in spending during 2000/01. Spending on the Essential Services Package (ESP) in 1999/2000 was between 60 and 70 percent of total spend- ing--28 percent on family planning, 35.5 percent on child health, 12.5 percent on limited curative care, 3.4 percent on communicable disease control, and 13.2 percent on 131 132 · Health Policy Research in South Asia maternal care. The distribution of public spending differed significantly by geographic region, but these differences do not seem to reflect differences in need. In fact, there appears to be a negative correlation between health and socioeconomic need and public spending allocations. ESP services are used predominantly by low-income groups. This reflects the fact that most of the HPSP's services are provided in rural areas in health complexes and family wel- fare centers. Men and boys appear to use nonreproductive primary level services more than women and girls. Esti- mates suggest that over the next five years, resources for the health sector will continue to come mainly from taxa- tion and development agencies. User charges may become an important source of revenue at the local level, but their overall contribution will remain small. Insurance may have some potential to provide additional funding, mostly through the gradual coverage of the informal sector. Introduction The Health and Population Sector Program (HPSP) for Bangla- desh completed its second year of a five-year program at the end of June 2000. This chapter looks back at the financial allocations for core HPSP activities over that fiscal year. The HPSP places a strong emphasis on the Essential Services Package (ESP) as a way of delivering cost-effective health care, par- ticularly to vulnerable groups. For convenience, the ESP has been defined as all primary care interventions delivered at thana levels and below. On this basis, between 60 and 70 percent of funding under the HPSP is now provided for ESP services (see table 6.1). However, as has been observed in previous public expenditure reviews (PERs), it is important to be aware that ESP services are also delivered in hospitals, particularly district hospitals.1 It is also likely that many of the resources spent at the thana level and below Public Expenditure Review of the Health and Population Sector Program · 133 Table 6.1 Financial Indicators of the HPSP INDICATOR BASE LEVEL 1997 (%) FINAL LEVEL 2003 (%) 1998/99 (%) 1999/2000 (%) Total spending on 60 65 65 60­70 the ESP (delivery and support) as a proportion of total health sector spending Proportion of health 23 30 43 51 sector recurrent expenditure going to important nonsalary components (especially medicine, maintenance) versus to salary component Proportion of health 75 80 85 91 sector expenditure for recurrent rather than capital expenditure Note: The definitions of the second and third indicators have both changed since the origi- nal baselines were set. This means that comparison of the current indicator with the base- line is somewhat misleading. The indicators may, however, be useful in their own right in monitoring the input composition of HPSP spending. are not used effectively. The financial definition of ESP must, therefore, be treated with caution. In addition to ESP services, the HPSP provides funding for a range of other activities, including the development of better approaches to the management of hospitals and changes to the way in which medical staff are trained. Although it is not yet possible to measure the impact of these interventions, in the future they should increase the efficiency of the public sector through quality and cost enhancements. This is the fifth PER conducted by the Health Economics Unit, and for the year 2000 it is produced jointly with the Management Accounting Unit. In addition to carrying out a national review of spending under the HPSP, the report focuses in depth on the ques- tion of equity, particularly the allocation of funding by geography, 134 · Health Policy Research in South Asia gender, and income. The final section of the report examines the question of what future resources will be available for the health sector in Bangladesh through some projections of funding from the public budget, health insurance, and user charges. National Expenditure Review of the HPSP For the 1999/2000 financial year, the Ministry of Health was origi- nally allocated 2,441 crore2 taka as the revenue and development budget for the HPSP. This represented 6.6 percent of the total government budget, nearly equal to the total allocation to the health sector of just over 7 percent, as specified by the fifth five- year plan. Total spending on the HPSP was 1,984 crore taka, according to the statements of expenditure of line directors and project reports from the Project Financing Cell. While this represents an increase of 3.7 percent over the previous year's spending in real terms, it is only 85 percent of the original approved budget (figure 6.1). During the two years of the HPSP, total annual spending has declined slightly in real terms by about 0.1 percent, compared to an increase of 18 percent in real terms in the two years prior to the start of the HPSP. In per capita terms, spending rose between 1998 and 1999; in 1999/2000, it has risen slightly from 135 taka (Tk.) to Tk. 143 per person. This is a small increase in real spending but still means that public (HPSP) health care spending per capita remains below 1996 levels. Public spending in 1999/2000 accounted for 1.1 percent of gross domestic product (GDP), the lowest share since 1992, although this rises to 1.22 percent if an estimate of health spending by other ministries is included.3 Two main reasons have been cited for the significant shortfalls in spending this year. The first reason is a general and continuing lack of understanding about procedures involved in procurement. This is exacerbated by a lack of procurement capacity, caused by an acute shortage of trained procurement experts. The second is that Public Expenditure Review of the Health and Population Sector Program · 135 Figure 6.1 Revenue and Development Spending on the HPSP 1995­2000, Original and Revised Budgets Crore taka 2,400 2,200 2,000 1,800 1,600 1,400 1,200 1995/96 1996/97 1997/98 1998/99 1999/2000 Original approved budget Revised budget Actual expenditure Real expenditure (constant 1995 prices) Note: The original approved budget is the budget approved at the start of the financial year. The budget is revised halfway through the year based on expenditures during the first six months. This new budget is referred to as the revised budget. Actual expenditure is based on reports of line directors for the year ending in June 2000. the guidelines for procurement of supplies and services have been criticized as being cumbersome and time-consuming. Both of these factors appear to have persisted more than two years into the HPSP, despite significant investment in procurement training pro- vided through the program. Most of the shortfall in actual spending relative to the budget is due to lower-than-anticipated development expenditure. Direct development spending by development partners (direct program 136 · Health Policy Research in South Asia assistance) was more than 86 percent of planned spending, while reimbursable program aid was more than 76 percent of planned spending. Development spending through the government was around 70 percent of the budgeted amount. The main shortfall was in the other reimbursable program assis- tance provided through the donor pool administered by the World Bank. This was just over a quarter of what was planned. This underspending on pool funding accounts for more than 47 percent of the total shortfall in spending for the year. The underspending of reimbursable program assistance is chiefly due to the procure- ment procedures, which continue to be misunderstood, are under- resourced, and lead to long delays in obtaining services. It has been suggested that some of the shortfall, particularly on the direct program assistance funding, may be due to underreport- ing by line directors because bilateral funding is not always prop- erly reflected in operational plans. The problem merits further investigation. Total program aid accounted for just under 37 percent of total HPSP spending. This is the highest level since the beginning of the 1990s, although the ratio has varied little during this time. It should be noted, however, that the planned proportion of spending from program aid was 42 percent, and the lower proportion mostly reflects the underspending on the development budget. There is an additional argument that much of the donor funding that is currently reflected in total health spending was not fully included before the advent of the HPSP. Therefore, the apparent increase in donor share may simply reflect better reporting. The conclusion, perhaps, is that while it is too early to start worrying much about the government being too dependent on donor fund- ing, the situation needs to be monitored in future years, particu- larly as the goal of financial sustainability requires that more recurrent items currently funded by donors be transferred to the government's revenue budget. Spending on nonsalary items was 51 percent of total spending in 1999/2000, an increase over the previous year's share of 43 percent (figure 6.2). This is explained by the greater share of development Public Expenditure Review of the Health and Population Sector Program · 137 Figure 6.2 The Distribution of Recurrent Expenditure by Salary and Nonsalary Items (by percentage and crore taka) Percent 100 302.26 608.29 910.55 90 80 70 60 628.25 50 870.15 40 30 241.90 20 10 0 Revenue Development Total Nonsalary Salary spending in 1999/2000, which was largely used for nonsalary items such as commodities and equipment. It is difficult to say what share of spending constitutes a reasonable level for salaries. Internation- ally, ratios for salary spending to total spending vary widely, from 20 to 30 percent in some Central Asian countries to more than 70 percent in some Organisation for Economic Co-operation and Development (OECD) and African countries (Barnum and Kutzin 1993).4 The main question, at least at the macroeconomic level, is whether the proportion devoted to supplies is sufficient for the existing staff to do their jobs satisfactorily, while at the same time providing them with adequate remuneration to ensure that they work effectively. Presently, around 0.35 percent of GDP (the salary component of the revenue budget) finances approximately 75,000 health workers, who constitute more than 1.2 percent of the formal workforce. This raises important human resource questions, includ- ing whether this funding is sufficient to motivate the workforce and 138 · Health Policy Research in South Asia whether the workforce has the right balance of skills to deliver the planned HPSP services. An additional point is that the plan to transfer the family plan- ning staff from the development budget to the revenue budget is not budget-neutral, because staff paid from the revenue side enjoy pension benefits and enhanced pay scales. These increases must be factored into the estimates of required revenue funding in future years. Capital spending, much of which was investment in community clinics, accounted for 9.6 percent of total HPSP spending in 1999/2000. Spending was lower than in 1998/99. Most capital spending is financed from the government-funded part of the development budget. Capital funding for ESP was 75 percent financed by the government, and more than half was used to finance the building of the first wave of community clinics. The remaining 25 percent largely financed the construction of the Insti- tute of Mother and Child Health at Azimpur. Given that considerable investment activity is planned during the early stages of the HPSP, the amount spent on capital appears to be low. Spending on the Essential Services Package To allocate spending to ESP and non-ESP items, expenditure on the operational plan subcomponents was allocated into three categories--ESP, non-ESP, and overhead. The overhead category was included because of the need for substantial expenditure on activities and infrastructure that support both ESP and non-ESP activities, such as research and training activities, management information systems (MIS), and procurement. To maintain consistency with the Project Implementation Plan and with the previous year's estimates, ESP continues to be defined as those services delivered at the thana level and below. However, this excludes any ESP services that are provided in hospitals. It also assumes that all spending at the thana level and below is on essen- tial services, which may not necessarily be the case. Public Expenditure Review of the Health and Population Sector Program · 139 When overhead expenditures are excluded and only direct ESP service expenditures are included, then the proportion of total spending that went to ESP services is estimated to be just more than 60 percent (figure 6.3). Assuming that overhead expenditures are about the same proportion as direct service expenditures, this suggests that around 70 percent of HPSP spending is on the ESP (figure 6.4). Therefore, based on the assumptions above, total ESP spending was between 60 and 70 percent for the financial year 1999/2000. Spending by Component Five main operational plans are partly or wholly involved in directly delivering ESP services: ESP-Health, ESP-Reproductive Health, Behavior Change Communications (BCC), Nutrition, and Construction (including the construction of community clinics). Figure 6.3 ESP, Non-ESP, and "Super Overhead" Expenditures, 1999/2000 (proportion of total revenue and development spending and crore taka) Percent 100 177.1 80.09 257.19 90 181.25 505.93 80 324.68 70 763.97 60 1,204.94 50 40 440.97 30 20 10 0 Revenue Development Total Overhead Non-ESP ESP 140 · Health Policy Research in South Asia Figure 6.4 ESP and Non-ESP Spending in the HPSP Assuming Proportionate Overhead Allocation, 1999/2000 (proportion of total spending and crore taka) Percent 100 399.78 196.61 596.39 90 80 828.70 70 1,371.67 60 50 542.97 40 30 20 10 0 Revenue Development Total Non-ESP ESP Other plans are indirectly involved through shared overheads. We analyzed these five plans in order to examine the distribution of expenditure by ESP subcomponent. We excluded Bangladesh Inte- grated Nutrition Project spending because it proved difficult to allocate these expenditures among ESP subcomponents. The level-three national accounting classification codes make disaggregating the main operational plans by ESP subcomponent relatively straightforward. The main difficulty is in allocating the shared direct costs of constructing community clinics and of reno- vating other thana-level facilities, which represent a significant share of the government-financed development budget. Analysis of spending by ESP component in the development budget shows that it is dominated by expenditures on family plan- ning and construction; family planning, for example, accounted for 69 percent of the development budget. In the case of funding by development partners through direct program assistance and other reimbursable program assistance, spending is more evenly distrib- Public Expenditure Review of the Health and Population Sector Program · 141 uted, although it is still dominated by family planning and, to a lesser extent, maternal and child health. There are two main problems with this approach. First, we did not include the revenue budget in our calculations because it is not coded in the same level-three format as development budget activ- ities. Second, the ESP­Reproductive Health Plan (development budget) includes the salaries of staff who work both on family planning and other reproductive health care and ESP compo- nents. In the case of ESP­Reproductive Health, staff costs account for more than 45 percent of development spending, whereas for ESP­Health Services they are less than 5 percent. This imbalance tends to exaggerate the level of funding devoted to family plan- ning services. In order to adjust for both factors, we used survey data from a recent study on the costs of ESP care. The survey obtained infor- mation on the use and costs of the time spent by both clinical and field staff on each ESP subcomponent at the thana level and below. Then we used the proportionate allocation to distribute the staff portion of both the revenue and the development budget among ESP subcomponents. This gives a more accurate picture of how the spending was actually distributed among components (figure 6.5). The allocations indicate a much wider distribution of funding than is suggested by development spending alone. Family planning spending still constitutes a substantial proportion of the total, but our work pattern analysis suggested that, at the clinical (nonfield) level, staff spent most time on child health. Maternal health received the third largest share of spending--13 percent--equal to the share of spending on limited curative care. It is difficult from these figures to judge whether this amount is adequate. The World Development Report suggests that on average, 31 percent of a low-income country's health package (US$3.80 per capita out of US$12 per capita) should be spent on prenatal and delivery care (World Bank 1993). Another way to measure relative spending is to compare actual expenditures with the budget estimates (annualized) provided in 142 · Health Policy Research in South Asia Figure 6.5 The Distribution of Revenue and Development Spending by ESP Component (Salary Spending Allocated According to Work Pattern Analysis)--Provisional Estimates, 1999/2000 Limited curative care BCC 3% Control of 13% communicable Family planning diseases 28% 3% Child health Maternal health 36% 13% Other reproductive health 4% Note: Construction costs are allocated in proportion to time distribution. BCC is also included as a hidden activity in each component, so the figure is probably an underestimate. Sources: Management Accounting Unit and ESP cost survey, Health Economics Unit, and Institute for Economic and Private Sector Development. the Project Implementation Plan (PIP). Although this approach assumes that the pattern of staffing is appropriate for the services being delivered, it provides a reasonable estimate of the equipment and supplies required to provide the ESP, based on quite detailed calculations of need. The comparison suggests that, while spending on child health and family planning is close to target, there was underspending in other categories of expenditure, notably mater- nal health, where there was a shortfall of nearly 60 percent. Many of the reasons for this underspending can be traced back to the procurement issues mentioned earlier. These trends will have to be monitored closely in future years to see whether these shortfalls are being made up. Public Expenditure Review of the Health and Population Sector Program · 143 The Distribution of Public Expenditure A central objective of the HPSP is to target public services to the most vulnerable: women, young children, and the poor. It is impor- tant to assess the extent to which this is actually taking place through the allocation of public subsidies as indicated by these tar- get groups' use of services. Analyzing public expenditures by activity and line item provides an aggregate picture of how public funds are allocated. To examine how funding is used in practice, it is necessary to map the flow of funds through the system to the final beneficiary of service using disaggregated data on spending and users. There are two key equity requirements: · That public services be distributed according to need. In the case of the HPSP, the needy groups are defined as women of reproductive age (particularly around the time of pregnancy), children under the age of five, and the poor. Access to services can be measured by the number of people in each of these groups who are receiv- ing appropriate, high-quality medical attention. · That services be financed according to income (ability to pay). Financ- ing includes both tax revenues and out-of-pocket expenditures. The new management information system makes it possible to dis- aggregate to some extent admission and consultation rates by level of facility and gender. It does not, however, make it possible to disaggre- gate the data by income group for each ESP category. To enable us to further disaggregate utilization rates and expenditures by categories of beneficiary and service, we carried out a benefit incidence analysis (BIA) survey to complement the MIS data. A total of 1,100 patients were surveyed across nine districts at thana and union levels. Thus, with the benefit of these data, in the next three subsections we analyze equity from three perspectives: geography, gender, and poverty. Geographical Analysis of Public Health Spending Our analysis of public expenditures by district indicates that there are significant geographical variations.5 144 · Health Policy Research in South Asia The Dhaka District, with per capita expenditures exceeding Tk. 400, is excluded from the analysis because Dhaka has the largest concentration of public tertiary facilities, which are used by residents from all parts of the country. The same is true, to a lesser extent, for division capital districts such as Chittagong and Rajshahi. In all divisions, with the exception of Chittagong, the capital district has higher per capita expenditures than elsewhere in the division, reflecting its higher concentration of facilities. In Chittagong, the districts of Rangamati, Bhandarban, and Cox's Bazaar all have higher expenditures than Chittagong itself, largely because of the additional allocations given to the hill tract areas. In order to correct for these variations within divisions, we sepa- rated towns and districts into two groups: the main division districts and all others. In the absence of other data, we assume that the inten- tion is to equalize expenditures among all of the districts. Currently, some are above average and some are below average (see figure 6.6). Figure 6.6 The Distance from Equal per Capita Targets (Division and All Other Districts), 1999/2000 Percent 140 120 100 80 60 40 20 0 Dhaka District District Khulna Khulna Sylhet Sylhet of Rajshahi Barishal Barishal of of Chittagong of of Rest of Rest Rest Rajshahi Rest Rest Chittagong Rest Public Expenditure Review of the Health and Population Sector Program · 145 Only in Chittagong are per capita expenditures below target for both the main division district and all other districts. This holds true despite the fact that hill tract districts receive a preferential allocation. Khulna is right on target, while Barisal exceeds the tar- get in both urban and rural areas. These targets are mostly illustrative. They show what would happen if per capita allocations were equalized across the country with all other things being equal. An important qualification is that other things may not be equal. In particular, differential per capita need and patterns of facility use that cross district boundaries may mean that equal per capita allocations are not appropriate. Need. An ideal resource allocation formula will account for the needs of the district. Per capita expenditures account for only one aspect of need--the size of the population. Other indicators of need include health status--more spending for districts with poorer health indicators--and income--more public funding for poorer areas with little capacity to finance care out-of-pocket. One proxy of health need that is robust in many circumstances and is used in several countries is the standardized (adjusted for age and gender) mortality rate. It can be considered to be a proxy for the extent of illness in the population. Data on these rates are not readily available by district, but this approach will be considered in further analysis when data do become available. Some information on income per capita in the 21 former districts was available. Grouping these districts into three categories, exclud- ing the Dhaka (new) District, suggests that current allocations are not inversely related to income and, in fact, that the richest six dis- tricts are allocated more funding per capita than the poorest six districts (see figure 6.7). Cross-Boundary Flows. A key claim, made in many countries, is that more resources per capita are required in urban areas to support the concentration of secondary and tertiary services used by citi- zens of both urban and rural areas. This is an important point because it is clearly not possible or economical to position special- ist services in areas with relatively low population density. 146 · Health Policy Research in South Asia Figure 6.7 Per Capita Public Health Expenditures by Income of District Income per capita (taka, 1996/97 prices) High (more than 7,800) Medium (6,700­7,800) Low (less than 6,700) 0 10 20 30 40 50 60 70 80 90 100 Public health spending per capita (1999/2000) Note: Data exclude Dhaka district. Source: Management Accounts Unit, Ministry of Health and Family Welfare (MOHFW), 1999/2000 In many countries, both rich and poor, evidence has demon- strated that urban citizens use such services more than rural resi- dents do. There are two reasons for this. The first is location, as urban citizens live closer to these facilities than rural residents do. Second, urban citizens often use secondary and tertiary facilities for their primary care needs. This means that patients are using rela- tively expensive facilities when they should be using lower-level facilities. One reason is that urban primary care is often relatively underdeveloped. In Bangladesh, little information is available about where those who use urban facilities and types of services live. It is important that such information be collected in order to examine whether facilities in these areas can justify the relatively higher public fund- ing that they receive. Patient surveys that collect data on what Public Expenditure Review of the Health and Population Sector Program · 147 treatment patients receive and where they live would provide such information. Conclusion. The absence of district-level health status information and information on cross-boundary flows means that the conclu- sions of this analysis must remain tentative. A first analysis suggests that the largely hospital bed­based criterion for allocations among districts means that there are considerable differences in per capita allocations. This holds even when the need for greater expendi- tures in urban areas to finance the supply of secondary and tertiary facilities is taken into account. There are clearly differences in need, as reflected in health status indicators and per capita incomes, that may modify the aim of dis- tributing resources equally. For this exercise, detailed district-level health data were not available. Data on per capita income and the HDI suggest that resources are not inversely distributed according to incomes--the opposite appears to be closer to the truth. More sophisticated analyses will be possible once data on health status, particularly standardized mortality rates, become available. Dhaka City has deliberately been left out of consideration because it dis- torts the analysis so significantly. It could indeed be argued that Dhaka is a special case because it serves the entire country with sophisticated services. This assumption should itself be tested by examining the case mix and district of residence of those using the facilities. Public Health Spending on Vulnerable Groups Services Distributed According to Need: Gender Use of Services. Based on district MIS data, women have a higher rate of consultation and hospital admission than men. Men, however, account for 17 per- cent more bed-days than women. This is probably partly accounted for by the type of admission (case mix). Many of the female admis- sions are probably for childbirth, which usually involves only a short stay. In contrast, many of the male admissions are probably for lifestyle diseases such as cardiovascular disease and trauma. The 148 · Health Policy Research in South Asia difference may also be a product of social-cultural factors that mean that men spend longer as inpatients than women. At the primary level (thana and below), a similar pattern prevails, with women accounting for a higher proportion of admissions and men accounting for a higher proportion of bed-days. In terms of per capita expenditures, there is much variation among districts. In the majority of districts (53 percent), per capita expenditures on women exceed those on men, which can be accounted for by childbirth and other reproductive services. In a significant minority of cases (47 percent), expenditures are actually higher for men despite the greater need for health care among women. Based on our expenditure analysis above together with the bene- fit incidence analysis in this section, it is possible to make a tenta- tive analysis of the distribution of expenditure by gender and by ESP service. It is assumed that family planning services benefit both men and women proportionately to their share of the popula- tion, although women obtain the most services. In a number of cases, it was difficult to assign patients to one of the ESP categories for a variety of reasons. One was that the con- sultation found nothing wrong. Another was that the patient reported having a general symptom such as headache or diarrhea, but the consultation did not yield any definite diagnosis that was shared with the patient. In a large number of these cases, medi- cines were prescribed. Men and boys accounted for a higher per- centage of the use of health facilities in cases where no diagnosis was provided. We found some differences in expenditures by gender. Total attendance at health facilities was slightly higher for women than for men. However, if the use of reproductive services is excluded, then use is actually higher for men than for women. If the patterns of use found in this survey were reflected throughout the country, this would suggest that the majority of nonreproductive health spending is directed toward men and boys (55 percent). Males seem to receive more care in the categories of child health (those under the age of five) and communicable diseases than females did. Public Expenditure Review of the Health and Population Sector Program · 149 For limited curative care where a diagnosis was possible, use was slightly greater for women than for men (table 6.2). Further work is required to determine whether the rates of admission and consultation among women are sufficient to meet their generally greater needs for health care, particularly reproduc- tive health services. We found that around 14 percent of total ESP spending was devoted to maternal health, compared, for example, to 26 percent for family planning. Given that the HPSP assigns particular priority to reducing maternal mortality, further work is required to estimate an adequate level of spending related to needs. The full cost of providing key obstetric services such as essential obstetric care should be investigated. Use of Services by the Poor. We analyzed the use of services by the poor by dividing the BIA sample into income quintiles per figure 6.8 based on national data on rural income distribution. The quin- tiles were based on total household consumption rather than income. This is because estimates of consumption proved to be eas- ier to obtain than estimates of income where there is a multiplicity of overlapping sources.6 Using these quintiles, we discovered that public primary care facilities are primarily used by lower-income groups. Quintile 1 (lowest household consumption) accounts for Table 6.2 Public Expenditure Allocation of Benefits by Gender and Type of Service (crore taka) BENEFIT MALE FEMALE TOTAL % Reproductive health Family planning 165.65 155.61 321.26 29.0 Maternal health n.a. 151.17 151.17 13.6 Other reproductive health 26.15 23.80 49.95 4.5 Child health 226.06 178.83 404.89 36.5 Control of communicable diseases 24.68 14.02 38.70 3.5 Limited curative care 68.81 74.33 143.14 12.9 Total 511.35 597.8 1,109 100.0 Percentage 46.1 53.9 Total (Nonreproductive health) 319.6 267.2 586.73 Percentage 55 45 n.a. = Not applicable. 150 · Health Policy Research in South Asia Figure 6.8 Use of Services by Income Quintile Quintile 5 14% Quintile 1 35% Quintile 4 8% Quintile 3 25% Quintile 2 18% Note: Quintile 1 represents the group with the lowest household consumption, while quintile 5 represents the group with the highest household consumption. Source: BIA, Bangladesh, 2000. more than 35 percent of visits, while quintile 5 (highest household consumption) accounts for only 14 percent (figure 6.8). Once at the health facility, patients have different experiences. For example, those in the richest quartile have to wait a shorter time than others before being seen by a health professional. Paying for Services. There is also some evidence of inequity in pay- ment for services. For outpatient services, the survey suggested that on average, people from different income groups pay about the same--a total of about Tk. 16 per visit. Just under one-third of patients reported making a payment. Equal payments by income group imply unequal proportionate spending. One visit to a public facility costs a poor household about 19 percent of per capita household income, compared with 4 percent for households in the richest quartile. In the case of inpatient services provided by thana health com- plexes, the average payment made by the poorest income group Public Expenditure Review of the Health and Population Sector Program · 151 rises. For all patients using facilities, the amount is nearly Tk. 40 (13 percent of per capita monthly income), compared with about Tk. 17 for the richest group (less than 2 percent of per capita monthly income; see figure 6.9). If only those who made a payment are included, the amount rises to around 49 percent of per capita household income for the poorest households. Some differences in per capita spending were also found between men and women. On average, men paid more for treat- ment than women. As outpatients men paid an average of Tk. 22, compared with Tk. 9.5 for women. As either inpatients or outpa- tients, men paid Tk. 51, compared with Tk. 17 for women. Payments varied considerably by the type of service provided. The largest payments were for communicable diseases (an average of Tk. 73) and for limited curative care (Tk. 41). We found low payment rates for maternal care and family planning (less than Tk. 12 on average). A cause for concern is that the poorest group Figure 6.9 Average Patient Cost per Visit (percentage of per capita monthly income) Average cost per visit (taka) Percentage of monthly income per person 45 16 40 14 35 12 30 10 25 8 20 6 15 10 4 5 2 0 0 Low Lower­middle Upper­middle Upper Average payment % of monthly income per person 152 · Health Policy Research in South Asia paid higher amounts for communicable diseases than any other income group (table 6.3). The reasons for these payment patterns are complex and require more extensive investigation, as does the fact that these patterns probably create barriers to access, especially for the poor. It is clear, however, that most spending (more than 56 percent) is on medi- cines and other medical supplies required for treatment (figure 6.10). Around 10 percent of these payments were "unofficial" pay- ments to doctors and other medical staff. We examined the prescribing practices of providers in some detail. In many cases, we found the medicines that they prescribed were inappropriate for the symptom or illness, and the medicines were prescribed in the wrong dosage or were given to the patient with inappropriate advice. In total, more than 91 percent of patients were prescribed some medicine as a result of their consul- tation, and 7 percent of the prescriptions were antibiotics. These general findings are confirmed by other surveys. A recent survey, for example, found that antibiotics are routinely prescribed for a Table 6.3 Average Payments per User by Income Group and ESP Category (taka) PROGRAM CATEGORY LOW LOW-MIDDLE UPPER-MIDDLE UPPER TOTAL Reproductive health Family planning 23.00 -- 0.33 2.00 12.53 Maternal health 1.44 9.40 3.00 -- 4.20 Other 0.38 125.10 34.75 60.00 63.04 Control of communicable diseases 79.29 109.67 24.75 1.40 73.05 Child health (under five) 2.19 56.77 12.73 3.29 13.77 Limited curative care 59.37 15.30 13.50 6.24 41.28 Symptoms only Medicines prescribed 33.74 12.11 38.64 28.27 29.10 Medicines not prescribed 1.55 -- -- 13.00 1.58 Further tests/hospitalization required 21.29 21.60 9.00 2.00 17.30 -- Not available. Public Expenditure Review of the Health and Population Sector Program · 153 Figure 6.10 Composition of Patients' Payments for Public Medical Treatment Outdoor ticket Admission 1.3% 0.2% Others 30.7% Diagnostic tests 2.7% Medicine and medical supplies Surgery 56.8% 0.9% Doctor's fee 7.4% Source: Bangladesh 2000. high proportion of disorders based on extremely short consulta- tions (Ahmed and others 2000). Key Findings Tentative conclusions that arise from this analysis can be summa- rized as follows: · There appears to be some inequality in the use of nonreproduc- tive health services. Males make up about 51 percent of the popu- lation yet consume about 58 percent of public spending. The difference is particularly marked in the case of communicable dis- ease control, where males constitute 64 percent of users. Whether this difference constitutes inequity depends upon whether males have a greater need for these types of services. · The poor make most use of services at the thana level and below, as measured by the number of patients presenting themselves at public facilities. 154 · Health Policy Research in South Asia There is some evidence that the rich and the poor are treated differently once they arrive at the health facility and during their treatment. The poor wait longer than the rich for treatment. They also pay considerably more for services, both in relative and absolute terms. Payments for communicable diseases are high and, given the large externalities involved, are of particular concern. More investigation of the process of obtaining key services such as tuberculosis care is required to increase access to care. There is little evidence that public subsidies favor poorer areas. Indeed, the reverse appears to be true. Further work is required to devise methods for allocating resources by geographical region. Medium-Term Resource Projections According to the 1996/97 national health accounts, Bangladesh spends almost US$11 per capita on health care (Heath Economics Unit and Data International 1998). More than two-thirds of this spending is out-of-pocket. There is evidence that considerable out- of-pocket spending is on ineffective drugs. In addition, several stud- ies have shown that people pay considerable sums to receive public services on an unofficial basis. Also, a variety of reports have pointed to the need for systems of risk-pooling that protect people from the high costs of unexpected illness. While total spending on health will increase mainly when national income itself increases, it may be pos- sible to redirect some existing spending to make it more effective. Two key additional sources of funding that could increase the effectiveness of the health care system are formal user charges and health insurance. It should be emphasized that both these tools have objectives that are wider than revenue generation. User charges, if retained by health facilities, have the potential to gener- ate significant improvements in the quality of basic services (Khan and Quayyum 2000; Routh and others 2000). Insurance would extend social protection by pooling risk to protect against the costs of catastrophic illness and would act as a catalyst to make providers more efficient. Public Expenditure Review of the Health and Population Sector Program · 155 Assumptions In this section, we make some crude projections of the resources that will be available to the public sector from the four main sources: the government's own revenue, donor financing, user charges, and health insurance. These projections are made on the basis of a series of assumptions, which are set out below. The Government's Own Revenue. Total government revenue is assumed to increase with the growth of the economy (currently 3.5 to 4 percent, rising to 5 percent by 2005). The existing efficiency of tax collection is assumed to improve slightly over the course of five years, with the proportion of GDP (market prices) collected by the government rising from the current level of 8.9 percent to 11 per- cent by 2005. The projections assume that the government deficit rises to just over 6 percent in 2000/01, but declines to 4.5 percent by 2005. The proportion going to the health sector is assumed to remain constant at just over 7 percent of GDP. Donor Funding. Projecting funding by Bangladesh's development part- ners is complicated by the fact that, during the two years of the HPSP, the donor development budget has far exceeded actual spending. This has mostly been due to problems of procurement. It is assumed that during the next financial year these problems will decline and that funding will rise to 75 percent and then 80 percent of the budget. (In 1999/2000, spending was around 65 percent of budget.) User Charges. It is assumed that the current submission to the Min- istry of Finance for the return of user fee revenue in the next finan- cial year will be successful (Dave Sen and others 2000). The simulations assume that a user fee initially will be tried out in about 10 percent of facilities, then extended to the majority of facilities by 2005. We assume that charges initially will be set at the current official admission ticket levels but that, in addition, charges will be introduced for outpatient treatment and inpatient admissions based on existing evidence of what people are willing to pay. Substantial 156 · Health Policy Research in South Asia exemptions for the poor and vulnerable are assumed, ranging from 60 percent of cases in primary care to 20 percent of cases at the ter- tiary level. Also, we assume that no exemptions are given for entrance tickets. Health Insurance. It is assumed that insurance is developed in two ways. For the formal industrial sector, representing about 6 percent of the population, this will be in the form of payroll-based social health insurance (Ensor 2000). Coverage will be extended first to civil servants by transferring part of their monthly medical allowances to a health fund (Killingsworth 1999) and later to the formal private sector. Over a five-year period, coverage is assumed to rise to 50 percent of the formal industrial sector. It is assumed that voluntary community insurance then will be developed through multiple schemes for the 35 percent of the pop- ulation who are employed in the informal sector. Premiums are assumed to be set at a level similar to those of existing nongovern- mental organization­based schemes (Desmet, Chowdhury, and Islam 1999). Coverage is assumed to rise to 5 percent of this sector in five years. Projections Figure 6.11 shows the evolution of the revenue available for the public health sector over a five-year period until 2005. Figures are given in nominal terms and assume a constant 5.5 percent inflation rate. The most notable aspect of the projections is the continued dominance of government and development partner funding even once insurance and user charges begin to be implemented. The projections show that available resources will grow by 92 percent over the five years (50 percent in real terms). Government revenue and development partner funding will remain the domi- nant sources, accounting for 93 percent of financing. Insurance will account for 6.1 percent and user fees for 1.4 percent by 2005. Without the new sources of funding, resources would grow by about 82 percent (40 percent in real terms). Public Expenditure Review of the Health and Population Sector Program · 157 Figure 6.11 Finance Available for Publicly Funded Health Care, 1999­2005 Taka 4,500 4,000 3,500 3,000 2,500 2,000 1,500 1,000 500 0 1999/00 2000/01 2001/02 2002/03 2003/04 2004/05 Goverment revenues (ADP and revenue) User charges Donor ADP Insurance Note: ADP = Annual Development Plan. Alternative Scenarios There are clearly many points of uncertainty in these estimates, and so the scope for sensitivity analysis is potentially wide. Thus, we investigated two further scenarios. In what we will call scenario 2 (taking the projections outlined above as scenario 1), we assume that macroeconomic variables remain the same. However, revenue from insurance would increase as a result of the following: · Wider community insurance coverage--rising to 15 percent of the target population · A larger formal sector--rising to 12 percent of the population · A larger community contribution--increasing from Tk. 50 to 100 per person per year (in constant prices) 158 · Health Policy Research in South Asia In scenario 3, again we assume that the macroeconomic variables remain the same. However, revenue from user fees would increase as a result of the following: · Inpatient fees rising to Tk. 1,000 for tertiary and Tk. 150 for dis- trict hospital admissions by 2005 · Outpatient treatment charges rising to Tk. 70 (average) at terti- ary and Tk. 30 at district hospitals Revenue projections under each scenario are shown in table 6.4. A notable feature is that, even with relatively high user charges, overall revenue generation from this source remains small. Rev- enue could be increased if exemptions were reduced, but much care would have to be exercised to ensure that doing so did not adversely affect access for vulnerable groups. Insurance has the potential to contribute a significant amount of revenue to the health sector. The qualification here is that increas- ing coverage, even to the modest levels suggested in the scenarios, may prove difficult within the time frame of the scenarios. Table 6.4 Revenue Projections under Different Assumptions (crore taka) 2001 (BASELINE) 2005 (S1) 2005 (S2) 2005 (S3) Insurance -- 236 536 236 User charges 12 55 55 76 Donor ADP 837 1,285 1,285 1,285 Government revenues (ADP and revenue) 1,543 2,315 2,315 2,315 Total 2,393 3,891 4,192 3,912 Percentage shares Insurance 0.0 6.1 12.8 6.0 User charges 0.5 1.4 1.3 1.9 Donor ADP 35.0 33.0 30.6 32.8 Government revenues (ADP and revenue) 64.5 59.5 55.2 59.2 Increase in revenue 97.0 112.0 98.0 Increase in real revenue 51.0 62.0 52.0 Note: ADP = Annual Development Plan. S1 = Scenario 1; S2 = Scenario 2; S3 = Scenario 3. -- Not available. Public Expenditure Review of the Health and Population Sector Program · 159 A second important qualification is that obtaining insurance con- tributions, particularly on a voluntary basis in the case of commu- nity insurance, requires that those who are insured receive valued and significant benefits. Often these benefits involve high-cost hospital treatment. Therefore, it may not be possible to use much of this revenue to improve essential care as currently defined by the ESP. An important question is whether the revenue projections are sufficient to cover the long-term sustainable costs of the ESP and other government services. To answer this question comprehen- sively, it is necessary to obtain an accurate, full-cost estimate of high-quality ESP services that take into account the necessary recurrent costs of the package together with the ongoing replace- ment costs of equipment purchased under the HPSP. The ESP study mentioned earlier would have provided some of this informa- tion, but the final analysis was not available at the time of writing this chapter. A more detailed analysis will be produced later. Conclusion Macroeconomic Overview of Spending As much as two-thirds of HPSP financing is being channeled into the Essential Services Package. On the development side, the ESP is dominated by spending on family planning, child health, and general investments in infrastructure. When the revenue budget is included, substantial resources in terms of staff time are spent on family planning, child health, limited curative care, and, to a lesser extent, maternal health. However, in terms of patient load, patients who fit into the child health, limited curative care, and symptoms-only categories make up the majority of users. These two different pictures are not necessar- ily incompatible, though. Most of the symptoms-only and limited curative care patients receive little in terms of staff time and clinic resources, although most are advised to purchase medicines. Never- theless, an interesting aspect of the attempt to prioritize services 160 · Health Policy Research in South Asia through the HPSP is that most patients using the ESP-level facili- ties are not actually demanding the ESP services that receive the most development funding. Equity Analysis Those in the lowest income groups are the ones demanding ESP- level services. Most patients presenting themselves for treatment are from the lower-income groups, a fact that lends credence to the view that reasonably effective pro-poor targeting can be achieved by directing general subsidies to primary care facilities. This judgement must, however, be qualified by the observation that there are inequalities in the process of obtaining care, as indi- cated by factors such as waiting times and patient payments. Geo- graphical targeting of needy areas also appears to be weak, although the lack of district-level data on direct program assistance funding means that this conclusion is only tentative. Although services and expenditure flows appear to be reasonably equally divided by gender, some potential inequities are apparent. When reproductive health care is excluded, men and boys appear to use more child health, communicable disease, and uncategorized primary-level services than women and girls. Resource Envelope The estimates of what future resources will be available for public services in Bangladesh indicate that the main sources of funding will continue to be overwhelmingly taxes and donors. User charges may become an important source of additional revenue for local facilities, but for the country as a whole, the percentage will remain small in the medium term. Insurance does have the potential to provide significant additional funding, mostly through gradually increasing coverage of the formal sector. User charges and insur- ance are also ways of channeling existing out-of-pocket spending in a more effective way. Public Expenditure Review of the Health and Population Sector Program · 161 Notes 1. For example, a recent study carried out by the Institute of Health Economics at Dhaka University found that, in two district hospitals, up to 30 percent of the outpatient caseload could be cate- gorized as ESP services. 2. One crore equals ten million. 3. The 1997 national health accounts suggested that other line ministries, together with local government, account for about 7.5 percent of public spending (around 0.12 percent of GDP). The main ministries that spend money on health services, mostly for their own employees, are Home Affairs, Defence, Railways, and Local Government. For fiscal 1999/2000, these health costs are estimated to increase public health spending to about Tk. 169 per person (Tk. 127 at constant 1993 prices), or 1.22 percent of GDP. 4. The high ratios are for different reasons. In OECD countries, high proportionate spending on staff is the result of relatively high salaries. In contrast, in low-income countries, while salaries are often low relative to average incomes, the level of spending is also so low that spending on medical supplies gets "crowded out." 5. This analysis is based on expenditures through the govern- ment, including the revenue budget, the government-funded development budget, and reimbursable donor aid provided through the government. Figures for donor aid given directly to districts were not available at the time this report was prepared. 6. Note, however, that national data suggest that on average the poorest groups consume up to 50 percent more than their income, while the rich consume around 8 percent less. The effect is to dampen any perceived income effect. References Ahmed, M., S. A. R. Chowdhury, C. F. Hossain, X. Quayyum, and O. F. Khan. 2000. "Drug Utilisation and Cost Efficiency in Thana Health Complexes of Bangla- desh." Draft report. University of Dhaka, Institute of Health Economics. 162 · Health Policy Research in South Asia Bangladesh, Government of. 2000. Benefit Incidence Analysis Survey. Health Eco- nomics Working Paper. Dhaka, Bangladesh. Barnum, H., and J. Kutzin. 1993. Public Hospitals in Developing Countries. Baltimore, Md.: The Johns Hopkins University Press. Dave Sen, P., E. Karim, J. Martin, and N. Abedin. 2000. Proposal to Ministry of Finance for Local Utilisation of User Fee Revenue on a Pilot Basis within HPSP. Occasional paper. Dhaka: Ministry of Health and Family Welfare, Health Economics Unit. Desmet, M., A. Q. Chowdhury, and M. K. Islam. 1999. "The Potential for Social Mobilisation in Bangladesh: The Organisation and Functioning of Two Health Insurance Schemes." Social Science & Medicine 48(7): 925­38. Ensor, T. 2000. Covering the Population: Extending Health Insurance in Bangladesh. Research Note 18. Dhaka: Ministry of Health and Family Welfare, Policy and Research Unit, Health Economics Unit. Heath Economics Unit and Data International. 1998. Bangladesh National Health Accounts 1996/97. Dhaka: Ministry of Health and Family Welfare. Khan, M., and Z. Quayyum. 2000. "Role of Private Providers in Healthcare: Lessons from Bangladesh on Public-Private Mix." Annual Scientific Confer- ence, International Centre for Diarrhoeal Disease Research, Dhaka, Bangladesh, February 11­13, 2000. Killingsworth, J. 1999. Bangladesh Government Employees' Health Insurance. Policy Options Paper 4. Dhaka: Ministry of Health and Family Welfare, Policy and Research Unit, Health Economics Unit. Routh, S., A. Hossain, M. Alam, M. N. Saha, Z. Quayyum, and Barkat-e-Kuda. 2000. "Current Expenditure for Health Care and Willingness-to-Pay for Health Services in Public-Sector Facilities: Evidence from Two Rural Thanas in Bangladesh." Paper presented at a conference on Health Sector Reform: Equity, Efficiency, Sustainability? Dhaka, Bangladesh, July 4­6. World Bank. 1993. World Development Report 1993: Investing in Health. Washington, D.C.: Oxford University Press. CHAPTER 7 Sri Lanka's National Health Accounts: National Health Expenditures 1990­1999 Institute of Policy Studies and Ministry of Health, Sri Lanka Abstract The chapter provides estimates of national health expendi- tures by source, by functional use, and by provider on an aggregate basis for 1990 to 1997 and on a preliminary basis for 1998 to 1999. Total health expenditures were equivalent to 3.2 percent of gross domestic product (GDP) in 1997, while total health expenditures per capita were 1,530 rupees (Rs.). These expenditures were equivalent to US$26 in 1997, up from US$16 in 1990. Government and private sources each accounted for approximately 50 percent of total financing throughout the decade. Government expen- ditures came almost exclusively from the central govern- ment's general revenues or from donor assistance. Private financing was mostly out-of-pocket spending by house- holds, with employer spending accounting for one-tenth and commercial insurance and expenditures of nongovern- mental organizations accounting for only 1 to 3 percent each. Expenditures were in the range of 3.1 to 3.5 percent 163 164 · Health Policy Research in South Asia of GDP throughout the 1990s. Government expenditures initially fell as a percentage of GDP and then rose, but at the end of the decade they were no higher than the 1.7 per- cent of GDP they had reached in 1990. Government expen- ditures increased less than was expected given the growth in the economy. Private expenditures remained relatively con- stant as a share of GDP. The three largest components of spending by function were inpatient care services, outpa- tient care services, and purchases of medicines in the private sector. Inpatient care services rose from 19 percent to 23 percent of total spending by function. Inpatient care was mostly funded by government sources, and as a share of government spending, it increased from 29 percent to 37 percent. This percentage is lower than in most demographi- cally advanced, industrial democracies. Most funding for outpatient care was from private sources. However, public facilities delivered a much greater proportion of actual ser- vices. Preventive and public health services fell from 11 per- cent to 6 percent of total spending as government allocations to these services fell. Health administration accounted for only 2 to 3 percent of total spending--very low by international standards. The central government share of public spending increased throughout the decade. The Western province had the highest level of expenditures, driven largely by its concentration of centrally funded facili- ties, and the highest level of private spending. Introduction This chapter summarizes the first estimates of the Sri Lanka national health accounts (SLNHAs), developed by the Institute of Policy Studies (IPS) working under the direction of the Depart- ment of Planning, the Ministry of Health (MOH), and the MOH Health Expenditure Survey Committee. It provides estimates of Sri Lanka's National Health Accounts: National Health Expenditures · 165 national health expenditures by source, by functional use, and by provider on an aggregate basis for 1990 to 1997 and on a prelimi- nary basis for 1998 to 1999. It also presents estimates of aggregate spending on a per-person basis and by province. The conceptual basis and definitions used to measure health expenditures in this chapter are based on the SLNHA Conceptual Framework.1 This framework is based on the System of Health Accounts (SHA) published by the Organisation for Economic Co- operation and Development (OECD 2000), modified to meet national requirements. The definition of national health spending we adopted corresponds to the OECD definition but with the explicit inclusion of expenditures on medical services and goods provided by unregistered and non-Western medical providers.2 To make international comparison possible, we also provide a dupli- cate set of estimates of aggregate spending using only the OECD classifications. The tables in this chapter express expenditure in nominal terms and, where indicated, in constant prices. We have used the Central Bank gross domestic product (GDP) deflator throughout for deriv- ing expenditures in constant terms. In the absence of a recent national census covering the whole island, the population statistics we used were IPS staff estimates, and they refer to the de facto resi- dent population of the country and its provinces. The estimates of total government spending for 1990 to 1997 are based on the audited financial accounts of the government and other public bod- ies and can be considered definitive. The estimates for 1998 to 1999 are preliminary, as they are based on the nonaudited statistics of final government spending and so remain subject to revision. Therefore, our the estimates of health expenditures in 1998 and 1999 are provisional. Our figures for private expenditures are nec- essarily estimates, compiled mostly from survey sources. Like any such estimates, they are subject to error. However, the estimates of the level of private-source spending are accurate to within plus or minus 0.35 percent of GDP, so the estimates of total national expenditures must be subject to the same degree of absolute error. 166 · Health Policy Research in South Asia The SLNHAs are the product of a collaborative effort over two years by many government agencies and private sector institutions in the country. While the SLNHAs meet the latest international standards, they were developed solely by Sri Lankan institutions and experts without any foreign technical input at any stage. Indeed, Sri Lanka is the first non-OECD country to produce esti- mates compatible with the OECD SHA 2000 standard. The SLNHA system and the information in this chapter are a good basis from which to assess and understand trends and levels of health spending in the country. Such data provide important infor- mation to the public, to policymakers, and to researchers for evalu- ating health expenditure­related policies and the performance of the national health system over time. This chapter presents only the main findings of the SLNHA findings. Detailed tables, a description of the conceptual frame- work, and a description of methods and data sources can be found in the full report, Sri Lanka National Health Accounts (Sri Lanka MOH and IPS 2001). Total Expenditures on Health Sri Lanka's total expenditures on health were estimated to be Rs. 28.3 billion in 1997, with per capita spending at Rs. 1,530. In real terms, total expenditures on health increased by 31 percent, from Rs. 21.7 billion in 1990 to Rs. 28.4 billion in 1997. Nominal and real trends in total expenditures on health are shown in figure 7.1. Health spending as a share of GDP ranged between 3.1 and 3.5 percent during the 1990s. There was no definite trend during this period (table 7.1). The ratio fell to 3.2 percent by 1997, but provi- sional estimates for 1998 and 1999 suggest that there were larger increases in those two years. Per capita spending was US$26 in 1997, up from US$16 in 1990. Overall, per capita spending on health increased at an average annual rate of 5 percent between 1990 and 1997, compared with an average annual increase in real GDP per capita of 5 percent (table 7.2). Sri Lanka's National Health Accounts: National Health Expenditures · 167 Figure 7.1 Trends in Total Expenditures on Health (TEH) for Sri Lanka, 1990­99 (in million rupees) 45,000 40,000 35,000 30,000 25,000 20,000 15,000 10,000 5,000 0 1990 1991 1992 1993 1994 1995 1996 1997 1998 1999 Nominal Constant 1997 prices Source: Sri Lanka MOH and IPS 2001. Expenditures by Sources of Financing The responsibility for funding health services in Sri Lanka is shared among all levels of government as well as the nongovern- ment sectors. Public expenditures at current market prices rose from Rs. 5.6 billion in 1990 to Rs. 14.0 billion in 1997. Private expenditures rose from an estimated Rs. 5.6 billion to Rs. 14.3 bil- lion during the same period (table 7.3). Government and private sources each accounted for approxi- mately 50 percent of total financing throughout the decade, or about 1.7 percent of GDP each (figure 7.2). The differences between them in any one year are within the margin of error, so neither can be said to be greater than the other with any certainty. During our study period, government expenditures initially fell as a percentage of GDP and then rose. At the end of the decade, 1999 31 9 8 4 3.53 39,177 34,293 1,110,653 972,196 1998 22 14 21 3.40 34,608 31,798 1,017,986 935,315 1997 31 16 4 3.19 28,389 890,272 28,389 890,272 1996 21 15 0 3.26 25,068 768,128 27,307 836,722 1995 22 15 31 3.34 22,288 667,772 27,215 815,395 1994 91 16 9 3.14 18,194 579,084 24,090 766,707 5 1993 17 ­4 3.06 15,276 499,565 22,116 723,275 1992 42 14 13 3.43 14,591 425,283 23,126 674,049 1990­99 5 1991 16 ­6 44766365 3.15 (TEH), 11,742 372,345 20,467 649,018 Health on 1990 3.48 11,196 321,784 21,659 622,488 2001. (%) IPS (%) (%) (%) (%) GDP of and Expenditures TEH GDP otalT prices in in rupees TEH GDP tion in in MOH 1997 propor market Lanka 7.1 million) million) increase increase million) million) a increase increase as Sri (Rs. (Rs. (Rs. (Rs. current constant ableT EXPENDITURE At TEH Annual GDP Annual At TEH Annual GDP Annual Health Source: 168 7 1999 29 2,068 1,810 1999 19.2 20.0 39.2 51,322 1998 11 29 1,843 1,694 1998 17.7 16.9 34.6 49,820 3 1997 26 1,530 1,530 1997 47,988 14.0 14.3 28.4 1996 1­ 25 1,369 1,491 1996 12.5 12.6 25.1 45,685 1995 11 24 1,231 1,503 1995 45,020 10.8 11.5 22.3 7 1994 21 1,018 1,348 1994 8.4 9.8 42,917 18.2 1993 867 ­6 18 1,255 1993 6.9 8.3 41,051 15.3 1990­99 1992 838 12 19 1,329 1992 7.1 7.5 38,727 14.6 Prices, 1991 681 ­7 336551543 16 1,187 Market 1991 5.5 6.3 37,631 11.7 1990­99 Current 1990 659 16 1,275 36,632 at 1990 5.6 5.6 11.2 Health Expenditures, (%) on (%) 2001. billion) 2001. Health capita IPS IPS capita (Rs. per and per billion) billion) and Expenditures Capita GDP (Rs. (Rs. prices rupees TEH Per (Rs.) (Rs.) in (Rs.) in MOH MOH (US$) otalT expenditures 1997 market Lanka sources sources Lanka 7.2 capita capita capita change increase capita Sri 7.3 Sri per per per per public private national current constant ableT EXPENDITURE At TEH At TEH Annual GDP Annual TEH Source: ableT EXPENDITURE otalT otalT otalT Source: 169 170 · Health Policy Research in South Asia Figure 7.2 National Health Expenditures by Source as Percentage of GDP, 1990­99 Percentage of GDP 4.0 3.5 3.0 1.7% 2.5 1.8% 1.7% 1.6% 1.9% 1.7% 1.6% 1.7% 1.6% 1.7% 2.0 1.5 1.0 1.7% 1.7% 1.6% 1.7% 1.7% 1.5% 1.6% 1.4% 1.6% 0.5 1.4% 0.0 1990 1991 1992 1993 1994 1995 1996 1997 1998 1999 (provisional)(provisional) Total government Total private Source: Sri Lanka MOH and IPS 2001. they were no higher than the 1.7 percent of GDP they had reached in 1990. Real incomes rose significantly during the decade, and the general tendency in any country is for government health expendi- tures to rise faster than income. Therefore, government expendi- tures increased less than would be expected given the growth in the Sri Lankan economy. Private expenditures remained relatively constant as a share of GDP despite a substantial increase in some areas of private spend- ing, such as spending at private hospitals and by commercial insur- ance companies.3 The constant private expenditure share is explained by a compensating decrease in household spending on traditional medicine, continuing a trend that had been observed in the Central Bank's Consumer Finance Surveys since the 1960s. Government expenditures come almost exclusively from the cen- tral government's general revenues or from donor assistance, which amounts to less than 10 percent of total public sector spending. Sri Lanka's National Health Accounts: National Health Expenditures · 171 Other sources, such as the Employees Provident Fund and provin- cial councils' own revenues, fund a very small amount. Private financing mostly consists of out-of-pocket spending by households, with employer spending accounting for one-tenth and commercial insurance and nongovernmental organization (NGO) expenditures accounting for only 1 to 3 percent each. Household spending accounted for 43 percent of total expenditures in 1997 (figure 7.3). Central government ministries and departments accounted for a growing share of total public sector expenditures during the decade, with the provincial councils' share declining to 31 percent Figure 7.3 National Health Expenditure by Source, 1997 Insurance Nonprofits Employers 1% 2% 4% Central MOH 30% National Health Expenditures Rs. 28,389 million 3.2% of GDP Public share 50% Private share 50% Households 43% Provincial Departments of Health Other government Local 16% ministries, departments, governments agencies 1% 3% Source: Sri Lanka MOH and IPS 2001. 172 · Health Policy Research in South Asia (figure 7.4). The bulk of central government expenditures were from the Ministry of Health, with a small, relatively constant pro- portion from other ministries, such as the Ministries of Social Ser- vices, Labour, and Defence. Local governments consist of both municipal councils and urban councils. Local government expendi- tures as a share of total government expenditures fell from 2.9 per- cent in 1991 to 2.3 percent in 1997. Expenditures by Function Expenditures in the SLNHAs are classified according to the five- digit SLNHA Functional Classification system, which is based on Figure 7.4 Government Spending by Administrative Level, 1990­99 Percent 100 90 80 39.4% 35.7% 37.1% 35.4% 34.2% 31.9% 31.3% 31.2% 39.5% 70 42.6% 60 50 40 30 55.9% 57.1% 56.6% 57.2% 57.8% 60.7% 62.1% 62.2% 53.0% 20 51.3% 10 0 1990 1991 1992 1993 1994 1995 1996 1997 1998 1999 (provisional)(provisional) Other government ministries, departments, agencies Local governments Provincial DOHs Central MOH Source: Sri Lanka MOH and IPS 2001. Sri Lanka's National Health Accounts: National Health Expenditures · 173 the OECD International Classification of Health Accounts (ICHA) system. Developing estimates of the functional breakdown of expenditures using the ICHA system required considerable effort and represented the major part of the work involved in estab- lishing the SLNHA system. For policymaking use in Sri Lanka, the ICHA system was modified into the SLNHA Functional Classifi- cation system, principally by adding more detailed subcategories to several of the ICHA categories. This adaptation was done in such a way that all SLNHA results can be easily mapped back into the ICHA categories, thus facilitating easy reporting of all results using the international standard. The overall experience is that the ICHA represents a feasible and sufficiently adaptable classification system for making functional disaggregations in developing country NHA systems. The largest share of total national spending, averaging about 78 percent annually, went to personal health services. Preventive and public health services accounted for about 2 percent of the total. Capital investment ranged from 9 percent to 15 percent, with the higher figures being driven by investments by private hospitals in recent years (table 7.4). An important feature of expenditures in Sri Lanka is that inpa- tient services account for a relatively high and growing share of national expenditures. Inpatient care services rose from 10 percent to 19 percent of total spending by function during the 1990s. This share is probably high by developing country standards but is low compared with most OECD economies, where inpatient spending accounts for about 35 to 45 percent of total spending. There is a clear distinction between how public and private expen- ditures are allocated. Table 7.5 compares public and private expen- ditures by function for 1997, when total public and private expenditures were approximately equal. As is evident, private spend- ing was used mostly for ambulatory care and purchases of medicines, while government spending was used mostly to fund inpatient services, public health services, and capital investment. Figures 7.5 and 7.6 show the components of public and private expenditures in 1997. 74 32 42 0 5 6 2 5 26 13 1999 2,068 39,177 41,072 74 32 42 0 5 6 2 5 24 17 1998 1,843 34,608 36,239 15 52 62 25 10 1997 1,530 28,389 29,861 05 42 62 25 12 1996 1,369 25,068 26,398 94 42 52 27 10 1995 1,231 22,288 23,473 84 32 62 8 28 1994 1,018 18,194 19,183 54 02 52 28 087761 9 1993 867 15,276 16,011 24 91 32 9 28 14 1992 838 14,591 15,220 44 02 42 28 10 10 1991 99­ 681 11,742 12,321 1990 14 91 22 00000000 55555555 22222222 35455555 26 11 15 1990 659 11,196 11,507 Function, (%) by (%) (%) vices million) 2001. (%) care outpatients ser (Rs. IPS (%) care to including and Expenditures care (%) medical health functions otalT (%) to (%) MOH dispensed public administration (%) (Rs.) items expenditures curative care carey rehabilitative vices related expenditures Lanka 7.4 of of sery and mation expenditures) goods program for million) capita insurance health all Sri health vices Inpatient Ambulator vices (Rs. per of ableT FUNCTION and (% Ser Ser Ancillar Medical Preventive Health Capital TEH TEH Memorandum Other otalT health-related Source: 174 Sri Lanka's National Health Accounts: National Health Expenditures · 175 Table 7.5 Relative Share of Funding by Public and Private Expenditures to Selected Functional Categories in 1997 FUNCTION PUBLIC (%) PRIVATE (%) Hospital services 81 19 Ambulatory care services 39 61 Medical goods dispensed 5 95 Preventive and public health services 87 13 Capital expenditures 99 1 Source: Sri Lanka MOH and IPS 2001. Figure 7.5 Total Government Expenditures by Function, 1997 Capital formation Inpatient care 20% 42% Health program administration and insurance 4% Total Government Expenditures Rs. 14,047 million 1.6% of GDP 29% Preventive and public health services 11% Medical goods dispensed to outpatients 2% Ambulatory care 21% Note: Total government expenditures for services of rehabilitative care and ancillary services to medical care were 0 percent. Source: Sri Lanka MOH and IPS 2001. 176 · Health Policy Research in South Asia Figure 7.6 Total Nongovernment Expenditures by Function, 1997 Preventive and Health program public health administration and services insurance 2% 1% Inpatient care 10% Medical goods dispensed to outpatients 47% Total Nongovernment Expenditures Rs. 14,342 million 1.6% of GDP Ambulatory care 31% Ancillary services to medical care 9% Note: Total nongovernment expenditures for services of rehabilitative care were 0 percent. Source: Sri Lanka MOH and IPS 2001. The three administrative levels of government--central, provin- cial, and local--also had distinctive expenditure patterns. More than two-thirds of hospital and ambulatory care expenditures and nearly all expenditures on medical goods were incurred at the cen- tral government level (table 7.6). Provincial councils were respon- sible for about one-third of hospital and ambulatory care expenditures. Preventive and public health services were distrib- uted across the three levels, with provincial councils spending more than half and local governments the least. Note that the largest Sri Lanka's National Health Accounts: National Health Expenditures · 177 Table 7.6 Relative Share of Funding by Central, Provincial, and Local Governments to Selected Functional Categories in 1997 CENTRAL PROVINCIAL LOCAL FUNCTION GOVERNMENT (%) COUNCIL (%) GOVERNMENT (%) Hospital services 69 31 0 Ambulatory care services 64 33 3 Medical goods dispensed 98 0 2 Preventive and public health services 37 56 7 Capital expenditures 74 24 2 Source: Sri Lanka MOH and IPS 2001. proportion of local government expenditures in this area was allo- cated to health-related expenditures such as environmental health, especially sanitation, which are not included in the SLNHA esti- mate of total expenditures on health. Inpatient care as a share of government spending increased from 13 percent in 1990 to 28 percent in 1997 (table 7.7). This percent- age is lower than in most demographically advanced, industrial democracies. Most funding for outpatient care was from private sources (table 7.8). However, public facilities delivered a much higher proportion of actual services. Spending on preventive and public health services fell from 11 percent to 6 percent of total spending as government allocations to these services fell. Health administration accounted for only 2 to 3 percent of total spending-- very low by international standards (in the United States, this pro- portion is 10 percent). Expenditures by Provider The largest proportion of government expenditures was incurred at hospitals, while almost all private expenditures (96 percent in 1997) were spent at private health care facilities that include both hospital and outpatient centers, as well as retail distributors of medicines (tables 7.9 to 7.11 and figures 7.7 and 7.8). Government and private hospitals accounted for 38 percent of total expenditures on health. Central government hospitals (text continues on p. 183) 1999 65 73 91 11 27 19,191 1,013 20,909 1998 35 53 81 10 32 945 17,739 19,207 1997 26 14 12 11 20 757 14,047 15,385 1996 95 93 02 13 23 683 12,507 13,721 1995 16 04 12 13 20 99988 596 10,800 11,885 1994 16 04 12 15 18 8,359 468 01 9,260 1993 45 53 91 20 20 6,935 394 7,595 99­ 1992 94 23 71 17 28 7,063 406 7,628 1990 1991 25 43 81 21 22 5,450 316 5,973 Function, by 1990 54 92 61 0000000000 0000000000 1122222222 20 4443434434 30 4979 5,619 331 5,882 (%) Expenditures (%) (%) million) vices 2001. (%) and nment care outpatients (Rs. ser IPS (%) care to including and Gover care medical health functions otalT (%) to (%) MOH dispensed (%) public administration (Rs.) expenditures items curative rehabilitative vices (%) and mation expenditures) expenditures Lanka 7.7 of care carey of sery goods program for million) capita all Sri health-related vices vices (Rs. per of health insurance (% health-related ableT FUNCTION Ser Inpatient Ambulator Ser Ancillar Medical Preventive Health Capital TEH TEH Memorandum Other otalT Source: 178 1999 73 9 82 01 50 19,986 1,055 20,164 1998 04 01 13 01 47 899 16,870 17,033 1997 14 13 9 47 773 14,342 14,476 1996 14 13 01 47 686 12,562 12,678 1995 83 92 01 50 634 11,488 11,588 1994 83 03 01 50 9,836 551 9,923 1993 83 03 01 50 8,340 473 8,416 99­ 1992 63 82 01 52 1990 7,528 432 7,592 9 Function, 1991 83 03 51 6,291 365 6,348 by 1990 63 88898999 82 0000000000 01 52 1112112222 0011111111 0000000000 1111111111 5,577 328 5,625 Expenditures (%) (%) nment (%) million) vices 2001. (%) and care outpatients (Rs. ser IPS (%) care to including and Nongover care medical health functions otalT (%) to (%) MOH dispensed (%) public administration (Rs.) expenditures items curative rehabilitative vices (%) and mation expenditures) expenditures Lanka 7.8 of care carey of sery goods program for million) capita all Sri health-related vices vices (Rs. per of health insurance (% health-related ableT FUNCTION Ser Inpatient Ambulator Ser Ancillar Medical Preventive Health Capital TEH TEH Memorandum Other otalT Source: 179 1999 21 12 49 39,177 2,068 1998 21 14 47 34,608 1,843 1997 21 13 49 28,389 1,530 1996 20 13 48 25,068 1,369 1995 19 14 50 22,288 1,231 1994 20 12 52 18,194 1,018 1993 19 11 53 867 15,276 99­ 1990, 1992 21 12 50 838 14,591 Provider 1991 16 13 52 681 of 11,742 ypeT 0000000000 0000000000 0000000000 3323322222 1111111111 7877566664 4543345578 2111112121 1111121111 0000000000 0011111111 0000000000 1990 21 13 48 by 659 11,196 (%) (%) (%) care vice (%) (%) Expenditures 2001. (%) facilities ser facilities facilities health (%) vices (%) IPS vice ser and National hospitals ser medical (%) (%) (%) residential health administration providing hospitals nonhospital providers otalT MOH nment and carey (%) (Rs.) medical care (%) hospitals nment nonhospital (%) health entities nment institutions Lanka 7.9 gover home (%) health entities million) capita Sri gover gover nment nment secondar entities (Rs. per facilities as ableT PROVIDER Central Provincial Local Nursing Nonhospital Provincial Local Public/community Gover Gover Nonprofit Insurers Private Other Foreign TEH TEH Source: 180 1999 41 24 8 61 19,191 1,013 1998 39 27 11 31 945 17,739 1997 41 27 12 9 757 14,047 1996 40 27 11 01 683 12,507 1995 38 29 12 596 10,800 1994 42 26 12 8,359 468 99­ 1993 42 24 15 6,935 394 1990, 1992 43 24 14 8779 7,063 406 Provider of 1991 34 28 16 01 5,450 316 ypeT by 1990 41 26 0000000000 0000000000 1111100101 5656654334 1211122111 13 8 4211223333 0000010000 0000000000 0000011100 1122222211 0001100000 5,619 331 Expenditures (%) (%) (%) care vice (%) (%) 2001. nment (%) facilities ser facilities facilities health (%) vices (%) IPS Gover vice ser and hospitals ser medical (%) (%) otalT (%) hospitals residential providing nonhospital health administration providers MOH nment and carey (%) (Rs.) medical care (%) hospitals nment nonhospital (%) health entities nment institutions Lanka 7.10 gover home (%) health entities million) capita Sri gover gover nment nment secondar entities (Rs. per facilities as ableT PROVIDER Central Provincial Local Nursing Nonhospital Provincial Local Public/community Gover Gover Nonprofit Insurers Private Other Foreign TEH TEH Source: 181 1999 1.42 96 19,986 1,055 1998 1.26 96 899 16,870 1997 1.06 96 773 14,342 1996 0.95 96 686 12,562 1995 0.91 96 634 11,488 1994 0.86 96 9,836 551 99­ 1993 0.78 96 1990, 8,340 473 1992 Provider 0.85 96 7,528 432 of ypeT 1991 0.88 97 6,291 365 by 0000000000 0000000000 0000000000 0000000000 0000000000 0000000000 0000000000 0000000000 0000000000 2222222222 0011111111 1990 0.77 97 0000000000 0000000000 5,577 328 Expenditures (%) (%) (%) care nment vice (%) (%) 2001. (%) facilities ser facilities facilities health (%) vices (%) IPS Nongover vice ser and hospitals ser medical (%) (%) otalT (%) hospitals residential providing nonhospital health administration providers MOH nment and carey (%) (Rs.) medical care (%) hospitals nment nonhospital (%) health entities nment institutions Lanka 7.11 gover home (%) health entities million) capita Sri gover gover nment nment secondar entities (Rs. per facilities as ableT PROVIDER Central Provincial Local Nursing Nonhospital Provincial Local Public/community Gover Gover Nonprofit Insurers Private Other Foreign TEH TEH Source: 182 Sri Lanka's National Health Accounts: National Health Expenditures · 183 Figure 7.7 Total Government Expenditures by Provider, 1997 Government entities providing health care as secondary care Private health Other entities 3% care providers 2% 1% Government health administration Central goverment 9% hospitals Public/community 41% health services 12% Local government nonhospital facilities Total Government 1% Expenditures Rs. 14,047 million 1.6% of GDP Provincial nonhospital medical service facilities 3% Nonhospital medical service facilities 1% Provincial hospitals 27% Note: Total government expenditures at local government hospitals and nursing home residential facilities were 0 percent. Source: Sri Lanka MOH and IPS 2001. accounted for two-thirds of government expenditures and private hospitals for more than 95 percent of private expenditures on hos- pitals. Of all expenditures on hospitals, private hospitals accounted for 13 percent, in contrast with the approximately 5 percent of total inpatient admissions that they provide nationally. Nonprofit insti- tutions accounted for less than 1 percent of all expenditures. Most private expenditures were made at pharmacies and other retail outlets. Purchases of medicines and medical supplies accounted for most of these expenditures. The other significant 184 · Health Policy Research in South Asia Figure 7.8 Total Nongovernment Expenditures by Provider, 1997 Nonprofit institutions 2% Central goverment hospitals Insurers 1% 1% Private hospitals 10% Pharmacies and other retail outlets 48% Total Nongovernment Expenditures Rs. 14,342 million 1.6% of GDP Private practitioners 29% Private medical and diagnostic laboratories 9% Source: Sri Lanka MOH and IPS 2001. element of private expenditures was on private practitioners, who accounted for 29 percent of all private spending on health. Expenditures by Province Tables 7.12 to 7.15 summarize trends in spending by province. The figures refer to all expenditures that can be directly attributed to a province by location, whether the spending was by provincial councils, local governments, central ministries, or households. 1999 342 266 346 249 289 373 190 596 1998 298 270 322 184 255 279 179 529 AL 940 930 TOT 1,205 1,278 1,008 1,199 1,215 2,157 83 1997 289 122 208 181 245 255 593 643 732 379 578 724 672 484 1,337 18 90 rupees) 1996 286 107 199 148 225 591 NONGOVERNMENT 1997 67 84 96 rupees) 1995 303 107 202 216 623 AL TOT 562 546 560 430 474 544 446 820 constant million GOVERNMENT (in 23 81 13 18 (in 1994 237 116 140 526 99­ 1997 7 0 6 0 2 0 0 30 26 27 12 14 20 OTHER 1990 1993 180 129 423 GOVERNMENT Source, nment, by & 25 95 02 13 29 18 1992 167 399 266 424 346 249 228 289 364 196 LOCAL Gover PROVINCIAL COUNCIL GOVERNMENT 2001. 39 66 92 22 69 32 Expenditures Central 1991 149 336 IPS 38 by 289 122 208 181 245 255 593 and Capita CENTRAL 21 64 71 11 28 15 GOVERNMENT 1990 121 312 Per MOH 2001. n n Lanka Expenditures IPS 7.12 ester n n Sri and th-Central th-Easter th-W Capita ableT ester PROVINCE Central Nor Nor Nor Sabragamuwa Souther Uva W Source: Per MOH n n Lanka 7.13 ester n n Sri th-Central th-Easter th-W ableT ester PROVINCE Central Nor Nor Nor Sabragamuwa Souther Uva W Source: 185 1999 275 614 382 314 242 278 432 219 93 1999 119 1998 276 567 380 298 250 295 431 209 1998 107 152 84 1997 266 424 346 249 228 289 364 196 1997 114 99­ 1990 98 1996 271 483 329 309 305 263 386 202 1996 104 99­ 1990 Province, 93 89 1995 296 492 262 313 386 259 392 204 1995 Any nment, to 75 81 1994 264 447 230 305 292 268 391 194 1994 Gover Local Attributable 84 99 1993 209 356 219 244 210 213 303 183 1993 and Directly Council Not 85 1992 268 417 249 304 255 260 368 183 1992 155 Capita 81 79 Provincial 1991 265 429 222 290 251 262 356 177 per 1991 by Health rupees) 87 1990 282 469 233 317 274 273 388 176 on rupees) 1990 163 2001. 2001. Expenditures 1997 1997 IPS IPS and Expenditure and Capita constant constant vices vices ser Per (in MOH otalT ser (in MOH n n Lanka medical Lanka 7.14 ester n n Sri collective 7.15 Sri th-Central th-Easter th-W ableT ester VICE PROVINCE Central Nor Nor Nor Sabragamuwa Souther Uva W Source: ableT SER National Islandwide Source: 186 Sri Lanka's National Health Accounts: National Health Expenditures · 187 Expenditures that are excluded under this definition include coun- trywide medical services that cannot be attributed directly to any specific province, such as medical services for prisoners or the armed forces, and national collective services that are of national benefit, such as HIV/AIDS prevention, malaria control, and the regulation of pharmaceutical products. The share of central government in public spending increased throughout the decade. From 1990 to 1994, provincial council spending fluctuated between 36 percent and 42 percent of total gov- ernment health spending. From 1994 onward, provincial councils' share of spending dropped sharply, from 40 percent to 31 percent. Government expenditures by province vary almost twofold in per capita terms. The range of expenditure levels has typically been 75 percent to 150 percent of the median province (figure 7.9). The Figure 7.9 Total Government Expenditures per Capita by Province (in constant 1997 rupees) Rupees 1,000 900 800 700 600 500 400 300 200 100 0 1990 1991 1992 1993 1994 1995 1996 1997 1998 1999 Central North-Central North-Eastern North-Western Western Southern Sabragamuwa Uva Source: Sri Lanka MOH and IPS 2001. 188 · Health Policy Research in South Asia Western province had the highest level of expenditures, driven largely by its concentration of centrally funded facilities, and the highest level of private spending. International Comparisons Much of the policy contribution of reliable national health expenditure data in OECD countries is related to the implica- tions arising from comparisons between countries. Similarly, much of the interest of policymakers in developing countries in such data is related to knowing how their expenditures compare with those of similar countries. This section discusses some of the technical issues that have bedeviled such comparisons and how this approach addresses these issues, before turning to some comparisons. Data Comparison Issues International comparison of health expenditures should be under- taken with caution because there was no uniform standard for measuring and reporting health expenditures prior to the develop- ment of the OECD System of Health Accounts (SHA) (OECD 2000). Statistics for national expenditures on health, particularly for developing countries, are rarely comparable, owing to differ- ences in the concepts and boundaries that define health spending, in accounting methodologies, and in classifications. These prob- lems are greatest when comparing private expenditures among countries. This situation should gradually improve in the coming years thanks to the new OECD SHA and other efforts to enhance the international comparability of health expenditure statistics both in developing countries and in the Asia-Pacific region. Since the SLNHAs are based on the OECD SHA framework, estimates will be comparable with the new SHA-compatible estimates of national health expenditures that should become available for both indus- trial and developing countries in the next few years. Sri Lanka's National Health Accounts: National Health Expenditures · 189 There is currently no reliable database of national health expen- ditures covering both industrial and developing countries. The only database of any reliability is OECD Health Data, compiled and maintained by the OECD Secretariat for member states. Other compendiums of health expenditure data published by the World Bank (1993) and by the World Health Organization (WHO 2000) are not useful for making international comparisons, as most of the national data are either imputed or of questionable reliabil- ity. In the absence of reliable international statistics, we have based our comparisons in this section on OECD Health Data and on data reported by members of the Asia-Pacific National Health Accounts Network (APNHAN). No comparison of health expenditures among countries can show whether a particular level of expenditure is either appropriate or effective. However, comparisons are useful in that they point to certain general patterns in spending and because they may point up differences between one country and another that may be impor- tant under specific circumstances. Levels of Expenditure From a global perspective, health expenditures in Sri Lanka are low both as a percentage of GDP and on a per capita basis (table 7.16). Health spending increases as a country's per capita income increases, so expenditure levels in Sri Lanka are not very different from those of countries at a similar level of economic development. However, as a share of GDP, expenditures in Sri Lanka (at 3.2 per- cent) are lower than those of most low-middle income or low- income developing countries for which estimates are available. Table 7.16 presents some comparisons. Expenditures by Source Unlike the advanced capitalist economies of the OECD, Sri Lanka funds a large share of its overall health expenditures from private sources. A 50 percent share for spending by the public sector is sig- nificantly less than the typical 85 to 95 percent in most OECD 190 · Health Policy Research in South Asia Table 7.16 National Health Expenditures in Selected Asia-Pacific and Organisation for Economic Co-operation and Development Countries and Economies HEALTH HEALTH PER CAPITA EXPENDITURE PER EXPENDITURE AS % COUNTRY/ECONOMY YEAR INCOME (PPP$) CAPITA (PPP$) GDP Bangladesh 1997 1,090 43 3.9 Sri Lanka 1997 2,460 79 3.2 Indonesia 1995 3,050 59 1.9 China 1997 3,070 138 4.5 Philippines 1997 3,670 131 3.6 Thailand 1996 6,650 247 3.7 Korea, Republic of 1997 13,430 674 5.0 New Zealand 1997 15,780 1,199 7.6 Taiwan, China 1997 16,000 843 5.3 Australia 1997 19,510 1,619 8.3 Sweden 1997 20,429 1,757 8.6 United Kingdom 1997 20,430 1,389 6.8 France 1997 21,294 2,044 9.6 Germany 1997 22,081 2,363 10.7 Netherlands 1997 22,639 1,924 8.5 Canada 1997 23,745 2,185 9.2 Hong Kong, China 1996 23,950 1,102 4.6 Japan 1997 24,400 1,830 7.5 Switzerland 1997 26,007 2,601 10.0 United States 1997 29,401 4,087 13.9 Sources: OECD health data, APNHAN sources, and national publications. economies. (Switzerland and the United States are the only two exceptions.) Low levels of public sector spending on health are in fact characteristic of developing countries in general. However, compared with most Asian countries, the public sector in Sri Lanka finances a below-average share of total national health expendi- tures, while households finance an above-average share (table 7.17 and figure 7.10). The limited data that exist suggest that this is true partly because of an increasing level of government involvement in health care expenditures during the 1990s in many Asian economies (including economies as diverse as Bangladesh; Hong Kong, China; Japan; the Republic of Korea; the Philippines; Taiwan, China; and Thailand), while public sector expenditures in Sri Lanka sustained a level of only 1.7 percent of GDP. The Sri Lanka's National Health Accounts: National Health Expenditures · 191 Table 7.17 International Comparison of Expenditures by Source (percentage) HONG KONG, CHINA SRI LANKA TAIWAN, CHINA UNITED STATES SOURCE 1996­97 1997 1998 1998 Government 54 48 7 33 Social insurance 0 0 52 19 Private sector 7 1 4 4 Private insurance 2 1 10 33 Households 37 51 27 11 Total 100 100 100 100 Sources: OECD health data; APNHAN sources. Figure 7.10 National Health Expenditures by Source as Percentage of Gross Domestic Product Bangladesh China Arab Republic of Egypt Hong Kong, China India Japan Jordan Nepal Sri Lanka Thailand United Kingdom United States 0 10 20 30 40 50 60 70 Percentage of GDP Public Private Sources: OECD health data; APNHAN sources. 192 · Health Policy Research in South Asia increasing dominance of governments in financing health care in Asian market economies does not necessarily mean that Sri Lanka should increase public spending, but it does suggest that this area needs further analysis. Notes 1. The SLNHA system and estimates were developed at the direction of Dr. K. C. S. Dalpatadu, deputy director general plan- ning, Department of Health Services, Ministry of Health, and the MOH Health Expenditure Survey Committee. The Health Expen- diture Survey Committee consists of representatives of the Min- istry of Health (Department of Health Services, Accounting Office), Ministry of Finance (Departments of National Planning and Census and Statistics), and the Central Bank of Sri Lanka. The Health Expenditure Survey Committee was responsible for super- vising and monitoring the development of the SLNHAs. The Institute of Policy Studies (IPS) Health Policy Programme was primarily responsible for the technical design and compilation of the SLNHAs. The team at IPS consisted of Ravi P. Rannan- Eliya, Aparnaa Somanathan, G. D. Dayaratne, Varuni Sumathi- ratne, and Shermal Karunaratne. The study team is indebted to the large number of individuals and agencies in both the government and private sectors who provided information and advice and gave generously of their time to assist with the compilation of the SLNHAs. The names of people and organizations who were con- sulted or who provided assistance are numerous. We would like to thank them all for their assistance and contributions. In particular, we thank Dr. S. M. Samarage at the Department of Health Services (Planning), MOH; Dr. W. Karandagoda, formerly of the Depart- ment of Health Services (Planning), MOH; T. G. Jayasinghe and A. Chandrasiri of the Finance Commission; A. G. W. Nanayakkara, director general, and Yasantha Fernando of the Department of Census and Statistics; Soma Mahaweva of the Finance Ministry; Sri Lanka's National Health Accounts: National Health Expenditures · 193 D. F. C. Hanwella of the University Grants Commission; M. Bala- subramaniam of IMS-Health, Sri Lanka; M. Kanapathipillai of the Insurance Controllers Division, Ministry of Finance and Planning; and Hema Wijeratne of Insurance Services International Ltd. The International Development Agency/World Bank Health Services Project provided funding support for the compilation of the SLNHAs. The work of the IPS Health Policy Programme was also funded by a number of other research grants and projects that the Institute holds, including grants from the South-East Asia Regional Office of the World Health Organization and the Policy Project of the Futures Group. 2. Although OECD (1997, 1998, 2000) definitions implicitly exclude such expenditures, this was not deliberate. 3. The official publication gives full details of the methods used to estimate private expenditures. However, it should be noted that in contrast to many health expenditure studies in developing countries, SLNHA estimates of household out-of-pocket spending that makes up the bulk of private expenditures were not derived primarily from household survey data. The primary method of estimation was mod- ification of production-side data; this is similar to the way these expenditures are estimated in U.S. national health accounts and in those of many other industrialized economies, where national income accountants typically estimate private consumption. References OECD (Organisation for Economic Co-operation and Development). 1997. "Draft Introduction to Functional Classifications." OECD-UNECE-EUROSTAT Meeting of National Accounts Experts, June 3­6, 1997. OECD technical docu- ment, number STD/NA (97)18/REV1. Paris: OECD Health Policy Unit. ------. 1998. A System of Health Accounts for International Data Collection. OECD technical document, number STD/NA/(98). Paris: OECD Health Policy Unit. ------. 2000. A System of Health Accounts. Paris: OECD Health Policy Unit. 194 · Health Policy Research in South Asia Sri Lanka MOH and IPS (Ministry of Health and Institute of Policy Studies). 2001. Sri Lanka National Health Accounts: Sri Lanka National Health Expendi- tures 1990­1991. Colombo. WHO (World Health Organization). 2000. World Health Report. Geneva. World Bank. 1993. World Development Report 1993: Investing in Health. New York: Oxford University Press. CHAPTER 8 The Bangladesh Health Facility Efficiency Study Ravi P. Rannan-Eliya, Institute of Policy Studies, Sri Lanka Aparnaa Somanathan, Institute of Policy Studies, Sri Lanka Abstract The Bangladesh Health Facility Efficiency Study surveyed a nationally representative, stratified sample of 122 Ministry of Health and Family Welfare (MOHFW) facilities. From these data covering 1997, we estimated service indicators and recurrent unit costs for outpatient and inpatient ser- vices in four kinds of facilities: (a) thana health complexes (THCs), (b) district and general hospitals (DH/GHs), (c) medical college hospitals (MCHs), and (d) specialized hospitals. We found that all facilities have generally high levels of utilization. Occupancy rates are high, close to an optimal level of 80 to 85 percent or even higher, and stays are generally short. Because facility budgets are generally fixed according to norms, high utilization rates translate into low unit costs for services. DH/GHs have the lowest unit costs of all facilities, and THCs had unit costs as high as those of MCHs, even though MCHs offer more 195 196 · Health Policy Research in South Asia sophisticated services and treat more severe cases than do THCs. The high costs of THCs are due to greater staffing levels than at higher-level facilities, coupled with lower utilization rates. In most countries, doctor-bed ratios are lowest in basic-level facilities, but Bangladesh is unusual in having its highest doctor-bed ratios in its low- est, primary-level facilities. We present evidence that strongly indicates that the current pattern of staffing and infrastructure at lower-level facilities in Bangladesh is suboptimal. We conclude that increasing the number of beds at THCs would make them more efficient, as the main problem seems to be undercapacity rather than oversupply. Also, increasing the ratio of nurses to doctors and reducing the numbers of Class 4 employees in THCs might lower average costs in delivering health services in Bangladesh. There is little evidence of systematic differ- ences in unit costs between facilities in different divisions, except for some evidence that facilities in Barisal and Syl- het have below-average levels of equipment and staffing, particularly doctors, which may partly explain the lower levels of utilization at facilities in these divisions. Introduction The government of Bangladesh faces significant resource con- straints in funding the proposed Essential Services Package (ESP). Previous reports have found that the potential for mobilizing addi- tional resources is limited and have suggested that improving the internal efficiency of the health services delivered by the Ministry of Health and Family Welfare (MOHFW) is an essential compo- nent of the effort to provide the ESP to the whole population. This study was conceived to provide the basic data required to develop a strategy for increasing the efficiency of all facilities--particularly the thana health complexes (THCs) and district hospitals (DHs) of The Bangladesh Health Facility Efficiency Study · 197 Bangladesh--to provide baseline data on the performance of MOHFW facilities before the start of the Fifth Population and Health Project, and to demonstrate the feasibility of survey meth- ods to collect the necessary information to calculate the unit costs and assess the efficiency of health facilities. At the beginning of this study, data on actual unit costs of deliv- ering services at the thana and division level were extremely lim- ited. This made it difficult to estimate the likely cost of the ESP and impossible to quantify the likely costs of existing inefficiencies. The absence of detailed facility cost data made it impossible to assess the scope for improving the efficiency of facilities. In the first phase of the Facility Efficiency Study, survey ques- tionnaires were developed and used to gather data on a sample of facilities. This phase has already demonstrated that rapid collection of data on unit costs is feasible and inexpensive. A survey of 100 facilities or more can produce division-level cost information in addition to national aggregates. It may be useful to carry out repeat surveys to monitor changes in the efficiency of the facilities. More sophisticated methods of analysis are required to fully examine the determinants of efficiency in these facilities, but these were beyond the scope of this initial study. In this report, we present our findings for this sample of facilities and discuss some preliminary implications, which must, however, be subjected to further analysis and investigation. Approach and Methods Phase I of the Bangladesh Facility Efficiency Study (BGFES98) collected data from a representative national sample of MOHFW inpatient health facilities. The study collected data on expendi- tures, levels of staffing, the availability of drugs and equipment, structural quality indicators, service volumes, and other indicators for calendar year 1997. The data set was designed to permit the estimation of recurrent unit costs in delivering services. The total 198 · Health Policy Research in South Asia sample consisted of 80 THCs, 18 DHs, 12 medical college hospi- tals (MCHs), and 12 specialized facilities. Development of the Methodology The methodology used was based on that developed for the Sri Lanka Public Facility Study 1998 (SLPFS98) by the Institute of Policy Studies of Sri Lanka, which in turn was based on the methodology developed for the Health Facility Survey conducted for the 1992 Sri Lanka Health Strategy and Financing Study by the Ministry of Health and the International Development Association (Akin and Samarasinghe 1994). The survey questionnaire was based closely on the initial draft questionnaire prepared for the SLPFS98, with appropriate modifications for Bangladesh. Data The BGFES98 collected data from a national sample of MOHFW inpatient facilities. A survey team collected the data by administer- ing a paper questionnaire at each facility. After the questionnaire had been adapted from the one then being developed for SLPFS98 and reviewed by a group of MOHFW hospital directors, it was pilot-tested at six THCs and DHs that were not to be included in the final sample. Based on feedback from this pilot and on the results of the simultaneous piloting of the SLPFS98 instrument in Sri Lanka, the authors made appropriate revisions to the questionnaire in consultation with the Health Economics Unit (HEU). The SLPFS98 instru- ment was later revised to keep it close in structure to that of BGFES98. This was done to enable comparison of the results of the two surveys at a later date. While they were developing the Bangladesh survey, the design- ers decided that cost and activity data would be collected only on the health activities of inpatient facilities. Many facilities also pro- vide reproductive services that have a separate budget and are administered by the Family Planning Division of the Ministry of The Bangladesh Health Facility Efficiency Study · 199 Health. They were excluded from the data collection and analysis for reasons of simplicity and cost. However, it was also decided to expand the survey to cover MCHs and specialized hospitals. The questionnaire was modified accordingly and was piloted at two MCHs, then revised to produce a second version to be used in MCHs and specialized hospitals. The final questionnaire was printed in English and was adminis- tered by two-person field survey teams from Data International. The teams collected the data by interviewing the staff of the facility and by extracting information from administrative records. In some cases, the field collection of data was supplemented by data from central MOHFW records. The fieldwork was conducted in two rounds: the first covering THCs and DHs, and the second cover- ing MCHs and specialized hospitals. Sampling The sampling frame consisted of all MOHFW health facilities with inpatient beds. The sample was selected using a stratified multi- stage probability design. The population was divided into two strata: (a) district and general hospitals (N = 60), and (b) thana health complexes (N = 395). Each stratum was then divided into six groups according to administrative divisions (N = 6). District Hospitals. It was decided that a minimum of two facilities would be drawn from each division and that sampling would be proportionate to the share of the overall MOHFW budget allo- cated to each division. In some cases, this would have led to the selection of one DH in a division. Given the available budget, the survey team decided to increase the sample size in the smallest divi- sion (Sylhet) by one DH to guarantee a minimum of two DHs per division. This yielded a desired sample size of 20 DHs (step 1). For this study, we reviewed the BGFES98 data for expenditures, admissions, and sanctioned beds for each facility for 1996. Total facility expenditures are driven by total sanctioned bed numbers because of budgeting norms, and they vary little among facilities. 200 · Health Policy Research in South Asia Expenditures per admission are, therefore, largely a function of admission rates and will approximate the final unit costs for admis- sions to be calculated in the survey. The survey team calculated the ratio of total expenditures per admission for all DHs and then ranked all hospitals in each division according to the level of this ratio. After this ranking, they divided each divisional list into equally sized strata; the number of these strata was based on the number of facilities determined in step 1. The survey team then selected one facility randomly from each stratum (step 2). It made sense to use budget data to order the sample because the ultimate objective was to obtain nationally representative cost estimates, and in Bangladesh, where hospital nonbudgetary revenues are limited, costs are driven by budgets. Thana Health Complexes. There is only a limited amount of infor- mation on the use of services at THCs. The number of beds per THC is fixed, and budgets are tied closely to sanctioned bed num- bers. Given the absence of recent comprehensive data on THCs in a usable format, the survey team chose the THCs for our sample randomly (random sampling without replacement) from the thanas that were also listed in the Bangladesh Bureau of Statistics (BBS) sampling frame for the Health and Demographics Survey (HDS). They chose two THCs in every district where the DH selected in step 2 was being surveyed and one THC each from every other dis- trict. The BBS HDS was a household survey that had collected population-level data on households by thanas. The survey team did this because they believed it was necessary to have household data to match against the data from each facility during subsequent analyses of efficiency and performance. There were three reasons why they chose two THCs from each district where a DH was being sampled: (a) it reduced their travel costs, (b) having a sample of two facilities would make it possible for analysts to estimate standard deviations in the future, and (c) they hypothesized that how lower-level THCs refer patients to other facilities may influ- ence demand at DHs. This procedure yielded a sample of 85 THCs. The Bangladesh Health Facility Efficiency Study · 201 Medical College Hospitals and Specialized Hospitals. The team ran- domly selected two separate samples of MCHs and specialized hos- pitals from the lists of such facilities. In total, they selected 8 MCHs from a national total of 13 facilities and 9 specialized hospi- tals from a national total of 28 facilities. The distribution of hospi- tals included in each sample was as follows (the actual number in each division is given in parentheses): · Medical college hospitals Barisal: 1 (1) Chittagong: 1 (2) Dhaka: 3 (4) Khulna: 0 (1) Rajshahi: 2 (4) Sylhet: 1 (1) · Specialized hospitals Barisal: 0 (1) Chittagong: 0 (4) Dhaka: 8 (10) Khulna: 0 (3) Rajshahi: 1 (7) Sylhet: 0 (3) Response Rates To ensure full cooperation, the survey team sent all of the facilities copies of the questionnaire in advance. The MOHFW in Dhaka also wrote officially to all facilities, seeking their cooperation. If staff were not available in a facility to complete questionnaires, field investigators were required to return to the facility at a later date. Two THCs were dropped from the survey and counted as non- responses because a flood made transport to the facilities unavail- able. All other facilities responded satisfactorily. The response rates were 100 percent for DHs, MCHs, and specialized hospitals and 98 percent for THCs. 202 · Health Policy Research in South Asia Estimations The survey had categorized the facilities into four types: · Thana health complexes · District/general hospitals · Medical college hospitals · Specialized hospitals The survey team put general hospitals (GHs) into the same cate- gory as DHs because, in practice, there is little to distinguish them and they are similar in both scale and function. General and district hospitals are essentially facilities that offer only basic services and are, therefore, also similar to THCs. However, they differ from THCs by virtue of their size and staffing norms. Therefore, we treated them as a separate category for the purposes of our initial analysis. Also, DH/GHs are regarded officially as secondary-level facilities, while THCs are regarded as primary-level facilities. We calculated the average unit costs of services for the inpatient and outpatient services each facility provided. The data set contains information on the total recurrent expenditures of each facility in 1997 by major line items, such as personnel, supplies, utilities, and drugs. We allocated all recurrent costs to either inpatient or outpa- tient services using a step-down procedure. For each facility, we allocated personnel costs, consisting of salaries and all allowances, to either outpatient or inpatient use. We used facility-specific data on the allocation of staff time to inpatient and outpatient duties by different grades of nurses and doctors to allocate their personnel costs by grade (see table 8.1). We distributed drug costs into the inpatient and outpatient cate- gories based on an estimate of the value of drugs actually distrib- uted from facilities' stores. We acquired information on the allocation of drugs to wards and outpatient departments by exam- ining the records kept at facilities' pharmacies for a sample of months during 1997. We allocated other medical supply costs as indirect costs, using the distribution of staff and drug costs as the The Bangladesh Health Facility Efficiency Study · 203 Table 8.1 Allocation of Recurrent Costs to Inpatient and Outpatient Services STAFF CATEGORY BASIS OF ESTIMATION Doctors According to reported allocation of time between outpatient and inpatient duties Nurses According to reported allocation of time between outpatient and inpatient duties Pharmacists, medical technologists Prorated according to percentage value (pharmacy), storekeepers of drugs used by inpatient and out- patient services Physiotherapists, occupational therapists 30% to inpatient (ratio estimated by Begum 1998) Pathologists 32% to inpatient (ratio estimated by Begum 1998) Radiology technicians 48% to inpatient (ratio estimated by Begum 1998) Rent controllers, ward masters, ward boys, laundry staff, cooks, stretcher boys 100% to inpatient Sweepers 75% to inpatient (ratio estimated by Begum 1998) Other staff Allocated as overhead cost using distribution of all other salary costs allocation ratio. Laundry and diet costs were allocated 100 percent to inpatient use. All other costs (excluding laundry and diet costs) were treated as overheads and allocated on a pro-rata basis accord- ing to the distribution of other costs. We replaced selected missing data with imputed values. Missing data on staff time allocations to inpatient and outpatient use were imputed using the observed averages for the relevant type of facil- ity (in other words, by THCs, DH/GHs, or MCHs). We used a similar procedure for missing data on the size of the MSR budget (for medical supplies), staff numbers, and laundry costs. Where data were imputed, the missing data accounted for less than 10 percent of all records with respect to the variable con- cerned. We analyzed the data using the computer software pack- age Stata (version 5.0). We calculated unit costs by dividing total estimated recurrent inpatient or outpatient costs by the number of inpatient services 204 · Health Policy Research in South Asia delivered. These unit costs were calculated for outpatient visits, admissions, bed-days, available bed-days, and beds. Lack of addi- tional data prevented us from doing a more detailed desegregation of unit costs by type of ward or medical department. We did not analyze or report on the parts of the data set relating to manage- ment and other structural quality indicators. Results In this section, we present the results of our analysis of the data from the Bangladesh Health Facility Efficiency Survey. Distribution of Facilities No facilities refused to cooperate with the survey, but two facilities were not surveyed because of logistical difficulties. Overall comple- tion rates were high for all of the items in the questionnaire. The geographical distribution of facilities in the final sample is shown in table 8.2. Hospital Characteristics The facilities in the THC and DH/GH categories seem to be very similar. Table 8.3 summarizes key statistics as reported by each cat- egory of facility. Table 8.2 Distribution of Sampled Facilities in Survey by Type and Division DISTRICT/ MEDICAL THANA HEALTH GENERAL COLLEGE SPECIALIZED DIVISION COMPLEXES HOSPITALS HOSPITALS HOSPITALS TOTAL Barisal 8 2 1 0 11 Chittagong 12 3 1 0 16 Dhaka 17 6 3 8 34 Khulna 12 4 0 0 16 Rajshahi 28 4 2 1 35 Sylhet 6 2 1 0 9 Total 83 21 8 9 121 The Bangladesh Health Facility Efficiency Study · 205 Table 8.3 Key Statistics by Category of Facility DISTRICT/ MEDICAL THANA HEALTH GENERAL COLLEGE SPECIALIZED DIVISION COMPLEXES HOSPITALS HOSPITALS HOSPITALS Beds 31.2 90.5 781.2 258.9 (2.9) (29.3) (216.6) (283.4) Outpatients (000s) per year 50.0 68.7 296.6 34.5 (68.0) (25.7) (109.2) (21.5) Admissions (000s) per year 2.3 7.6 34.3 3.1 (1.0) (3.8) (14.5) (4.0) Bed occupancy (%) 74.8 94.6 109.9 76.0 (28.7) (47.3) (28.3) (21.1) Operations performed per year 200.0 1,296.8 9,827.0 809.4 (525.3) (2,541.6) (3,385.5) (975.1) Number of doctors 5.5 10.0 60.7 9.1 (1.3) (2.6) (13.8) (5.0) Number of nurses 5.9 26.2 203.5 60.0 (1.2) (17.7) (68.8) (54.8) Number of Class 3/Class 4 31.0 33.2 480.5 95.2 employees (9.2) (19.4) (308.0) (97.5) Recurrent expenditures 6.2 8.1 115.8 25.2 (taka millions) (1.9) (3.1) (64.5) (16.3) Note: Mean values in sample with standard deviation in parentheses. The typical thana health complex has an average of 31 beds (in a range of 15 to 50) and is staffed by 5 doctors (in a range of 2 to 9), 6 nurses (range 2 to 8), and 31 other staff. With an average recur- rent budget of 6.2 million taka (Tk.), the typical THC has 50,000 outpatient visits, 2,300 inpatient admissions, and 200 operations a year. THCs deliver only very basic medical services and carry out few operative interventions. They seem to be quite homogeneous in their basic characteristics, reflecting the fact that they operate according to fixed norms. District and general hospitals (24 and 48 percent of the sample, respectively) are larger facilities, typically with 50 or 100 beds. A few district and general hospitals have more beds, up to a maximum of 150. The typical 100-bed DH is staffed by 10 doctors (in a range of 5 to 14), 26 nurses, and 33 other staff. With an average recurrent budget of Tk. 8.1 million (range Tk. 6 million to 14 million), DHs 206 · Health Policy Research in South Asia provide an average of 68,000 outpatient visits, 7,000 inpatient admissions, and 1,200 operations a year. Medical college hospitals are larger inpatient medical facilities that provide a range of different services, including specialties. They have from 540 to 1,100 beds, 40 to 90 doctors, and 140 to 370 nurses. Their budgets are much larger than those of the other kinds of facilities, an average of Tk. 115 million. General Facilities, Equipment, Hours of Operation, and Services Offered Utilities and Equipment. As we expected, the amount and range of equipment provided increase with the level of facility (table 8.4). All facilities have laboratories and operating theaters, although the laboratories in 12 percent of THCs are not functional. Only half of THCs have been provided with functioning X-ray machines. All DH/GHs and MCHs have functional X-ray machines. Electro- cardiogram (ECG) equipment is not available in THCs and is available in only 43 percent of DH/GHs. Cardiac monitors, ultra- sound scanners, and intensive care facilities are found only in MCHs. Only two THCs and just over half of all DH/GHs reported having blood banks, while all MCHs had them. Generally, all facilities have basic utilities, such as electricity, piped or deep-tube well water, and refrigerators (table 8.5). Four percent of THCs reported having no telephone. Surprisingly, 98 percent of THCs reported having freezers, but only 24 percent of DH/GHs did so. The reason for this is unclear but may be related to the distribution of freezers to THCs through the expanded pro- gram of immunization. Availability of Services. The regular hours of operation are similar at all levels. Facilities offer routine outpatient services for eight hours a day, five days a week, and are open for emergencies on a 24-hour, seven-day-a-week basis. MCHs provide all major types of services, such as obstetric, gynecological, pediatric, medical, and major surgical care. THCs (%) (%) 57 57 52 78 63 100 100 100 100 100 100 100 100 100 100 ALS ALS COLLEGE FUNCTIONAL COLLEGE FUNCTIONAL HOSPIT HOSPIT MEDICAL (%) (%) 57 57 52 100 100 100 100 100 001 MEDICAL 75 100 100 100 100 100 AILABLEVA AILABLEVA (%) (%) 57 43 95 24 91 ALS 100 100 100 ALS 100 100 100 AND FUNCTIONAL AND FUNCTIONAL HOSPIT HOSPIT DISTRICT (%) DISTRICT (%) GENERAL GENERAL 57 43 95 29 100 100 100 100 100 100 100 AILABLEVA AILABLEVA (%) (%) 2 0 Facilities 88 99 52 93 98 96 95 TH TH 100 100 at HEAL FUNCTIONAL Facilities HEAL FUNCTIONAL at COMPLEXES COMPLEXES THANA (%) THANA (%) Equipment 1199 2 0000 0000 0 0000 53 Utilities 96 98 96 ILABLEAVA 100 100 100 100 100 AILABLEVA vailableA unit vailableA theater scanner water y 8.4 care monitor y 8.5 bank well machine equipment water/deep- tube ableT EQUIPMENT Laundr Laborator Operating Blood Intensive X-ray Ultrasound ECG Cardiac ableT UTILITY Refrigerator Freezer oiletsT Piped Electricity/generator elephoneT 207 208 · Health Policy Research in South Asia and DH/GHs are similar in the services that they actually provide, with more than 85 percent of facilities in each category providing obstetric, gynecological, pediatric, and minor surgical services. This is notable because only 80 percent of THCs are designated to provide obstetric services and 24 percent of THCs are designated to provide pediatric services. Major surgery is generally available only at the DH/GH level and above. A large proportion (24 per- cent) of THCs is designated to provide dental services but do not actually provide them. Staffing and Allocation of Staff Time MCHs have more staff than DH/GHs, which have more staff than THCs (table 8.6). DH/GHs have twice as many doctors as THCs and two to four times the number of nurses. However, both cate- gories have similar numbers of Class 3 and Class 4 employees. DH/GHs have fewer Class 3 employees than THCs and have cor- respondingly more Class 4 employees. The staff mix varies across different categories of facility. The nurse-doctor ratio is higher in higher-level facilities, while the ratio of Class 3/Class 4 staff to doctors and nurses is lower. Although the number of skilled staff (doctors and nurses) in relation to beds is more or less similar at facilities of all levels, the number of total staff per bed is higher in THCs than in other facilities. The higher Table 8.6 Staffing Indicators and Ratios DISTRICT/ MEDICAL THANA HEALTH GENERAL COLLEGE SPECIALIZED STAFF COMPLEXES HOSPITALS HOSPITALS HOSPITALS Doctors 5.5 10.1 60.7 9.1 Nurses 5.9 26.2 203.5 60.0 Class 3 15.0 9.6 96.4 27.2 Class 4 16.0 23.6 384.1 68.0 Nurse/doctor ratio 1.2 2.8 3.3 10.2 Class 3/4 doctor/nurse ratio 2.8 1.1 1.8 1.1 Bed/doctor ratio 6.2 9.3 13.1 55.9 (Nurses + doctors)/bed ratio 0.37 0.43 0.35 0.39 Staff/bed ratio 1.4 0.9 0.9 0.9 The Bangladesh Health Facility Efficiency Study · 209 ratio of staff to beds at THCs is due to their relatively higher num- bers of Class 3/Class 4 staff. The reason why THCs have relatively more staff than any other level of facility is not apparent. In the case of doctors, the ratio of doctors per bed is actually lower in the more sophisticated facilities than in lower-level facilities. Whether this counterintuitive finding reflects an optimal staffing pattern is worth exploring. Generally, doctors allocate 40 to 50 percent of their time to inpatient duties in all types of facilities, while other categories of staff allocate higher proportions of their time to inpatient duties (table 8.7). This is consistent with the fact that the duties of some staff (such as ward boys, laundry staff, and cooks) are limited to providing services for patients in inpatient wards. Utilization and Performance General Patient Load. All facilities provide both outpatient and inpa- tient services. The service mix at THCs is geared more toward out- patients than the mix at higher-level facilities. The ratio of outpatient visits to admissions at the THC level is 22, compared with approximately 9 at higher-level facilities. The overall patient load at MCHs is approximately five times greater than at DH/GHs. The type of care provided is more sophisticated at the MCH level--proportionately more patients at that level are given laboratory, radiology, and other tests. However, the number of immunizations provided is lower in the more sophisticated types of facility than in the less sophisticated types (table 8.8). Table 8.7 Allocation of Staff Time to Inpatient Care DISTRICT/ MEDICAL THANA HEALTH GENERAL COLLEGE SPECIALIZED STAFF COMPLEXES (%) HOSPITALS (%) HOSPITALS (%) HOSPITALS (%) Doctors 41 41 43 49 Nursing staff 94 86 95 99 Class 3 71 56 54 66 Class 4 76 78 56 93 210 · Health Policy Research in South Asia Table 8.8 Average Annual Number of Outpatient Services and Tests by Category and Type of Facility DISTRICT/ MEDICAL THANA HEALTH GENERAL COLLEGE SPECIALIZED SERVICE COMPLEXES HOSPITALS HOSPITALS HOSPITALS Outpatient department visitsa 50,024 68,744 296,619 34,557 Dental visits 451 4,712 17,689 0 Laboratory tests 3,736 7,039 53,987 23,972 Radiology examinations 580 3,217 30,781 9,855 Immunizations 51,096 22,842 6,135 0 aIncludes dental visits. Inpatient Services. Most facilities reported high levels of occupancy, admission rates, and turnover rates. The highest occupancy rates are at MCHs (110 percent). DH/GHs were 95 percent occupied during 1997, compared with 75 percent occupancy at THCs. The higher rate at higher-level facilities is comparable with admission patterns in many other developing countries, including those in the South Asia region. It probably reflects patients' preference for the better care provided by higher-level facilities. The average length of stay is quite short--3.9 days at THCs and 4.5 days at DH/GHs. This fact, coupled with the high occupancy rates, suggests that most of these primary-level facilities are operating close to their capacity. The longer average stay at MCHs (11.0 days) is consistent with the fact that they have more seriously ill patients than other facilities. All facilities other than specialized hospitals have a broad mix of inpatients. They are roughly equally distributed across surgical and medical specialties at both DH/GH and MCH levels. THCs main- tain only general wards, but responses by THC personnel to the survey questions about which services they provide suggest that THCs probably have a diagnostic mix of patients similar to that of DH/GHs. Cabin inpatients represent 3 to 4 percent of all patients at DH/GHs and MCHs; THCs do not have cabins. The number of inpatients admitted increases as the facility level becomes more sophisticated (table 8.9). At the MCH level, 32 per- cent of inpatients undergo surgical interventions, compared with The Bangladesh Health Facility Efficiency Study · 211 Table 8.9 Inpatient Service Statistics DISTRICT/ MEDICAL THANA HEALTH GENERAL COLLEGE SPECIALIZED CATEGORY COMPLEXES HOSPITALS HOSPITALS HOSPITALS Admissions per year 2,301 7,656 34,288 3,119 Operative intervention rate (%) 8.7 15.8 31.6 29.9 Mortality rate (%) 1.8 4.6 10.2 6.2 Deliveries per year 95 488 5,105 0 Cesarean section rate (%) 0.9 12.9 35.9 n.a. Ratio of outpatient visits to admissions 21.7 9.0 8.7 11.1 n.a. = Not applicable. 16 percent at the DH/GH level and 9 percent at the THC level. The fact that patients in higher-level facilities stay longer and are more seriously ill is also consistent with the fact that mortality rates are higher at higher levels, ranging from 2 percent at the THC level to 10 percent at the MCH level. The proportion of babies delivered by cesarean section is signifi- cantly higher at the MCH level than at the DH/GH level. Respon- dents for the THCs reported doing very few cesarean sections. Under normal circumstances with optimal care, one would expect fewer than 10 percent of babies to be delivered by cesarean section; for example, teaching hospital units in Sri Lanka generally report cesarean section rates of less than 12 percent of all births. Whether the much higher rates of 36 percent reported at MCHs and 13 per- cent at DH/GHs in Bangladesh reflect the admission of women with high-risk pregnancies or a high rate of unnecessary cesarean sections cannot be determined from these data. As a rate of 36 per- cent can be considered high from a clinical perspective, this ques- tion should be explored further. Comparing Facilities' Performance Using Service Indicators. Perfor- mance indicators can be used to conduct a preliminary assessment of facilities' relative performance (Barnum and Kutzin 1993). We used Lasso's (1986) method to summarize data on bed occupancy and turnover rate (and, therefore, implicitly average length of stay 212 · Health Policy Research in South Asia [ALOS]) in a large sample of facilities. Figure 8.1 presents the data on bed occupancy and turnover rates for the sample of THCs. The solid vertical and horizontal lines indicate the mean values for bed occupancy and turnover rates respectively, while the dotted lines are one standard deviation each from the respective means. The rays from the origin represent points whose ALOS is either one standard deviation below the mean or one standard deviation above the mean. The four quadrants represent different groups of facilities. Those in quadrant I have below-average turnover rates and bed occupancy. These facilities have the capacity to admit more cases without reducing ALOS. There are a large number of facilities in this quad- rant, suggesting that many have the capacity to admit more patients. Figure 8.1 Performance Indicators for THCs by Division, Bangladesh, 1997 Turnover rate (percent) ALOS = 2.7 180 160 140 II III 120 ALOS = 5.4 100 80 60 40 I IV 20 0 20 40 60 80 100 120 140 160 Bed occupancy rate (percent) Barisal Khulna Chittagong Rajshahi Dhaka Sylhet Note: ALOS = Average length of stay. The Bangladesh Health Facility Efficiency Study · 213 Facilities in quadrant II have below-average occupancy rates and above-average turnover rates. These facilities have an ALOS below the mean, perhaps because they admit predominantly minor cases. Facilities in quadrant III have above-average turnover rates and bed occupancy. These facilities have occupancy rates close to 100 per- cent or higher, indicating considerable overcrowding. A large per- centage of THCs fall into this quadrant. Since for most THCs, ALOS is less than five days, there seems to be little room for improving output by reducing ALOS, which confirms that these facilities have insufficient capacity to meet demand. Our examination of the distribution of facilities by division revealed no systematic pattern. There are many outliers in each quadrant, and their exceptional performance may warrant further detailed examination. Costs Detailed information was collected on costs at each facility. We used these data, as described above, to estimate unit costs for ser- vices. These cost estimations are for recurrent costs only, and therefore, they underestimate full costs. In addition, we did not consider the costs of the services administered and funded by the Family Planning Division. Table 8.10 gives the overall distribution of costs by category in each group of facilities. Personnel costs account for 84 percent of total recurrent costs at THCs. The proportion is lower at DH/GHs and MCHs, where spending on drugs and other medical supplies is relatively higher. Table 8.10 Distribution of Recurrent Costs by Category of Cost DISTRICT/ MEDICAL THANA HEALTH GENERAL COLLEGE SPECIALIZED COST CATEGORY COMPLEXES (%) HOSPITALS (%) HOSPITALS (%) HOSPITALS (%) Personnel 84 61 54 57 Drugs 5 14 16 16 Medical supplies 3 8 13 8 Other 8 17 17 19 214 · Health Policy Research in South Asia Within facilities, inpatient services account for the greatest share of all costs (table 8.11). Surprisingly, despite the relatively greater outpatient load at THCs, the proportion of overall costs accounted for by inpatient services is similar at THCs (63 percent) and at DH/GHs (62 percent). The cost mix for outpatient services is sim- ilar to that for inpatient services in all facilities, except that drug costs are higher for outpatient services (tables 8.12 and 8.13). Table 8.11 Share of Recurrent Costs Accounted for by Inpatient Use DISTRICT/ MEDICAL THANA HEALTH GENERAL COLLEGE SPECIALIZED COST CATEGORY COMPLEXES (%) HOSPITALS (%) HOSPITALS (%) HOSPITALS (%) Personnel 64 66 78 80 Drugs 26 32 65 64 All inpatient costs 63 62 77 77 Table 8.12 Breakdown of Recurrent Costs in Providing Inpatient Services DISTRICT/ MEDICAL THANA HEALTH GENERAL COLLEGE SPECIALIZED CATEGORY COMPLEXES HOSPITALS HOSPITALS HOSPITALS Share of facility costs (%) 63 62 77 77 Cost per admission (taka) 1,957 843 3,249 11,872 Percentage of costs Staff 85 63 54 60 Drugs 3 7 13 11 Medical supplies 3 8 13 8 Other 9 22 20 21 Table 8.13 Breakdown of Recurrent Costs in Providing Outpatient Services DISTRICT/ MEDICAL THANA HEALTH GENERAL COLLEGE SPECIALIZED CATEGORY COMPLEXES HOSPITALS HOSPITALS HOSPITALS Share of facility costs (%) 37 38 23 23 Cost per outpatient visit (taka) 66 55 102 283 Percentage of costs Staff 82 52 53 58 Drugs 10 24 24 22 Medical supplies 3 7 13 8 Other 5 17 10 12 The Bangladesh Health Facility Efficiency Study · 215 Inpatient Unit Costs. We estimated three indicators of inpatient costs: · Annual cost per available bed · Cost per bed-day occupied · Cost per admission We also estimated the average cost of an outpatient visit. The results are summarized in table 8.14. THCs appear to be the most costly facilities for the delivery of inpatient services. The cost per available bed and per bed-day occupied is lowest in DH/GHs and highest in THCs. Although the cost per available bed in THCs (Tk. 111,397) is only double that in DH/GHs (Tk. 56,119), the cost per bed-day occupied is almost three times higher (Tk. 521 versus Tk. 188), owing to the higher utilization of DH/GHs. There are several possible explanations for the higher unit costs at THCs. First, THCs have higher staff-to- bed ratios than DH/GHs and MCHs. Second, the staff mix at THCs is more expensive than at DH/GHs because THCs use more nurses per doctor and fewer Class 3/Class 4 employees than DH/GHs. Overall, the ratio of administrative and other support staff to doctors and nurses is highest at THCs, which must add to the cost of delivering their services. Finally, patient demand is higher for the kinds of services offered by DH/GHs than for those Table 8.14 Gross Unit Costs for Inpatient and Outpatient Services (taka) DISTRICT/ MEDICAL THANA HEALTH GENERAL COLLEGE SPECIALIZED ITEM COMPLEXES HOSPITALS HOSPITALS HOSPITALS Beds available per year 111,397 56,119 110,565 117,830 (46,515) (14,924) (31,820) (71,419) Bed-days occupied 521 188 277 441 (325) (68) (45) (260) Admissions 1,957 843 3,249 11,872 (1,232) (603) (2,896) (7,673) Outpatient visits 66 55 102 283 (45) (44) (68) (516) Note: Mean values in sample with standard deviation in parentheses. 216 · Health Policy Research in South Asia offered by THCs. An unavoidable conclusion is that THCs are too small to achieve economies of scale. Although the cost per available bed is similar at MCHs and THCs, the unit cost of an occupied bed-day is almost twice as high at THCs. This must be the result of the almost 50 percent higher occupancy rate at MCHs than at THCs. Cost per admis- sion is lowest again at DH/GHs (Tk. 843) and highest at MCHs (Tk. 3,249), with THC admission costs falling in between at Tk. 1,957. The high admission costs at MCHs reflect the much longer average stay at these facilities and, presumably, the more severely ill patients they admit and the more sophisticated services they provide. Outpatient Unit Costs. Outpatient unit costs are highest at the highest- level facilities, MCHs (Tk. 102). However, surprisingly, they are lowest at DH/GHs (Tk. 55) rather than at THCs, which might have been expected to be the least costly in delivering outpatient services. The high costs of THC outpatient visits again primarily reflect their higher staffing levels with respect to the volume of services delivered. Geographical Variations in Unit Costs. There is little systematic dif- ference in budgets and costs per available bed between facilities in different geographical divisions. This may reflect the standard norms used in allocating budgetary resources and staff to different facilities. In contrast, there are significant differences in the utiliza- tion of facilities among divisions. Facilities in Barisal and Sylhet reported significantly lower rates of inpatient and outpatient uti- lization than facilities in other areas. Facilities in Dhaka and Chit- tagong had the highest utilization rates. In combination with essentially fixed and relatively equal budgets for each facility, this leaves facilities in Barisal and Sylhet with the highest unit costs, while facilities in Dhaka and Chittagong have the lowest unit costs (tables 8.15 to 8.17). The variation in unit costs is largely driven by differences in uti- lization, which might be due to underlying differences in demand The Bangladesh Health Facility Efficiency Study · 217 Table 8.15 Cost per Bed-Day Occupied by Type of Facility and Division (taka) DISTRICT/ MEDICAL THANA HEALTH GENERAL COLLEGE SPECIALIZED DIVISION COMPLEXES HOSPITALS HOSPITALS HOSPITALS Barisal 439 195 266 n.a. (112) (26) (a) Chittagong 635 149 252 n.a. (396) (73) (0) Dhaka 465 149 325 479 (161) (36) (35) (256) Khulna 491 245 238 n.a. (137) (136) (11) Rajshahi 523 211 238 178 (407) (2) (11) (a) Sylhet 657 236 245 n.a. (558) (5) (a) Country 521 188 277 441 (325) (68) (45) (260) n.a. = Not applicable. Note: Mean values in sample with standard deviation in parentheses. a Only one facility in cell. Table 8.16 Cost per Admission by Type of Facility and Division (taka) DISTRICT/ MEDICAL THANA HEALTH GENERAL COLLEGE SPECIALIZED DIVISION COMPLEXES HOSPITALS HOSPITALS HOSPITALS Barisal 2,256 895 2,390 n.a. (405) (336) (a) Chittagong 2,564 506 2,082 n.a. (2,845) (71) (a) Dhaka 1,960 733 5,430 9,738 (975) (247) (4,201) (5,119) Khulna 1,834 1,066 n.a. n.a. (559) (735) Rajshahi 1,684 1,281 1,603 26,806 (623) (1,375) (355) (a) Sylhet 1,968 631 2,023 n.a. (1,232) (100) (a) Country 1,957 843 3,249 11,872 (1,232) (603) (2,896) (7,673) n.a. = Not applicable. Note: Mean values in sample with standard deviation in parentheses. a Only one facility in cell. 218 · Health Policy Research in South Asia Table 8.17 Cost per Outpatient Visit by Type of Facility (taka) DISTRICT/ MEDICAL THANA HEALTH GENERAL COLLEGE SPECIALIZED DIVISION COMPLEXES HOSPITALS HOSPITALS HOSPITALS Barisal 89 81 49 n.a. (28) (38) (a) Chittagong 85 95 130 n.a. (40) (103) (a) Dhaka 58 47 121 383 (30) (18) (104) (708) Khulna 52 36 n.a. n.a. (19) (4) Rajshahi 57 43 114 198 (34) (21) (47) (a) Sylhet 61 24 49 n.a. (50) (6) (a) Country 66 55 102 283 (45) (44) (68) (516) n.a. = Not applicable. Note: Mean values in sample with standard deviation in parentheses. aOnly one facility in cell. for facility services, differences in the quality of facilities, or a com- bination of both. Table 8.18 summarizes the differences in the budgeting, staffing, and equipping of THCs by division. Although facilities in all divisions receive similar budgets, facilities in Sylhet have the fewest doctors and nurses in place and have smallest amount of X-ray machines and other basic equipment in function- ing order. These differences may partly explain the differences in Table 8.18 Indicators of Resource Availability at THCs by Division (means per facility) TOTAL RECURRENT NUMBER OF NUMBER OF X-RAY EXPENDITURE DOCTORS NURSES MACHINES DIVISION (TK. MILLIONS) IN PLACE IN PLACE FUNCTIONAL Barisal 5.4 5.4 6.7 0.38 Chittagong 6.9 6.3 5.4 0.58 Dhaka 6.2 5.6 5.7 0.68 Khulna 5.5 4.9 6.6 0.42 Rajshahi 5.7 5.6 6.0 0.52 Sylhet 5.3 4.3 4.5 0.33 The Bangladesh Health Facility Efficiency Study · 219 utilization, but differences in the propensity of people to seek care at MOHFW facilities cannot be excluded. A useful area of further research might be to analyze the BBS HDS data to investigate this question. International Comparison of Costs and Performance Indicators Tables 8.19 to 8.21 compare the performance of the sampled MOHFW facilities in 1997 with facilities in various other develop- ing countries for which comparable data are available. The tables distinguish among three levels of hospitals in these countries (Bar- num and Kutzin 1993): · Level I: Tertiary-level facilities with the most specialized staff and technical equipment, with highly differentiated clinical service functions · Level II: Facilities lacking the most technically sophisticated ser- vices available in Level I hospitals but with some functional dif- ferentiation by clinical specialty · Level III: Most basic facilities with few specialists and limited laboratory services; generally referred to as "district" or "first- level referral" hospitals. MCHs in Bangladesh are comparable with Level I hospitals in other countries, while DH/GHs and THCs are comparable with Level II and Level III hospitals. All tables rank the countries according to the specific indicator being tabulated. Bangladesh's facilities have high occupancy rates in comparison with those of most other countries, with MCHs having among the highest observed occupancy rates for hospitals of their type. This is because of the relatively long patient stays in MCHs and because of their average bed turnover rates. In the case of THCs and DH/GHs, the high occupancy rates are due to their very high turnover rates and short lengths of stay. Why lower-level MOHFW facilities admit so many short-stay cases is unclear. However, it cannot be explained on the basis of a high per capita admission rate, since these are quite low in Bangladesh compared 6.9 7.2 7.2 7.8 8.2 9.0 9.4