CROSS-COUNTRY ANALYSIS OF CASE-MIX IN SELECTED HOSPITALS DISCUSSION PAPER JUNE 2023 Olena Doroshenko Marta Kuzmyn Kristiina Kahur Sarah Bales Solomiya Kasyanchuk CROSS-COUNTRY ANALYSIS OF CASE-MIX IN SELECTED HOSPITALS Olena Doroshenko, Marta Kuzmyn, Kristiina Kahur, Sarah Bales, Solomiya Kasyanchuk June 2023 Health, Nutrition, and Population Discussion Paper This series is produced by the Health, Nutrition, and Population (HNP) Global Practice of the World Bank. The papers in this series aim to provide a vehicle for publishing preliminary results on HNP topics to encourage discussion and debate. The findings, interpretations, and conclusions expressed in this paper are entirely those of the author(s) and should not be attributed in any manner to the World Bank, to its affiliated organizations, or to members of its Board of Executive Directors or the countries they represent. Citation and the use of materials presented in this series should take into account this provisional character. 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. For information regarding the HNP Discussion Paper Series, please contact the Editor, Jung-Hwan Choi at jchoi@worldbank.org or Erika Yanick at eyanick@worldbank.org. Rights and Permissions The material in this work is subject to copyright. Because the World Bank encourages dissemination of its knowledge, this work may be reproduced, in whole or in part, for noncommercial purposes as long as full attribution to this work is given. Any queries on rights and licenses, including subsidiary rights, should be addressed to World Bank Publications, The World Bank Group, 1818 H Street, NW, Washington, DC 20433, USA; fax: 202- 522-2625; e-mail: pubrights@worldbank.org. Copyediting: Anne Himmelfarb Cover design: Bogdana Fomina © © 2023 The International Bank for Reconstruction and Development / The World Bank 1818 H Street, NW, Washington, DC 20433 All rights reserved. 1 Health, Nutrition, and Population (HNP) Discussion Paper CROSS-COUNTRY ANALYSIS OF CASE-MIX IN SELECTED HOSPITALS Olena Doroshenko,a Marta Kuzmyn,a Kristiina Kahur,a,b Sarah Bales,a,c and Solomiya Kasyanchuka a Health, Nutrition, and Population Global Practice, The World Bank, Washington, DC, United States b HC Management Consulting Ltd, Estonia c Hanoi University of Public Health, Vietnam Paper prepared with the support of Korea–World Bank Partnership Facility. Abstract: This study presents a comparative analysis of inpatient care data from six countries (Ukraine, Moldova, Romania, Croatia, Estonia, and the Republic of Korea). It uses diagnosis- related groups (DRGs) as a framework for benchmarking. The primary aim of the study is to propose methods for comparing hospital outputs across countries that use different systems for coding hospital activities. Additionally, this study seeks to identify best practices in coding and hospital performance indicators, which could be of interest to countries and hospitals included in this analysis and beyond. It compares hospital outputs from 2021 at both country and hospital levels, using indicators such as the frequency of cases within different adjacent DRGs (A-DRGs) and major diagnostic categories, length of stay, coding activities, and case complexity. It also employs country-level analysis to determine the scope of hospital-level data collection and analysis. The study draws upon data from 30 hospitals of different types across all six countries. Its findings provide insights and identify areas for further investigation that can orient and guide hospital reforms. Such reforms may aim at ensuring better quality and cost-effectiveness of care, as well as patient access to hospital services. Key words: Diagnosis-related groups, hospitals, benchmarking, comparative study Disclaimer: The findings, interpretations, and conclusions expressed in the paper are entirely those of the authors, and do not represent the views of the World Bank, its Executive Directors, or the countries they represent. Correspondence Details: Olena Doroshenko, 1 Dniprovskiy Uzviz, Kyiv, Ukraine; Tel.: +38 044 490-6671 Email: odoroshenko@worldbank.org; Website http://worldbank.org 2 Table of Contents Acknowledgments 5 Abbreviations 6 Executive Summary 7 1 Introduction 12 2 Methodology and Data Description 14 2.1 Methodology 14 2.2 Data sources and content 16 2.3 Country-level data 17 2.4 Hospital-level data 18 3 Findings at the Country Level 20 3.1 Country contexts 20 3.2 MDC-level analysis 21 3.3 Analysis of Surgical Cases versus Medical Cases 24 3.4 A-DRG-level analysis 26 3.5 Analysis by main diagnosis 38 3.6 Analysis by procedure codes 40 3.7 Use of country-level findings to define the scope of the hospital-level analysis 42 4 Findings at the Hospital Level 44 4.1 Coding activity 44 4.2 Performance indicators 47 4.3 Specific cases of interest 51 5 Discussion of Study Limitations 54 6 Summary and Conclusions 55 7 Recommendations 57 References 59 Annexes 62 Annex 1. Hospital-Level Data Processing and Cleaning 62 Annex 2. Templates for Collection of Country- and Hospital-Level Data 65 Annex 3. Questionnaire for Collecting the Country-Specific DRG System Data 67 Annex 4. Summary of DRG systems in countries 69 Annex 5. Overview of Countries’ Health Care Systems 70 Annex 6. Overview of ALOS differences 79 Annex 7. Definition of the Case Types 83 Boxes Box 1. Main Steps for Processing the Country-Level Data ................................................................. 15 Figures Figure 1. Overview of the DRG Algorithm 15 Figure 2. Total Population and Number of Hospital Cases per Country, 2021 18 Figure 3. Overall Admissions per 100 People, 2021 18 Figure 4. Share of Hospital Cases by Select MDCs in Study Countries 23 Figure 5. Admissions by Top MDCs per 100 People, 2021 24 Figure 6. Medical vs. Surgical Cases in Study Countries: Number and as Share of All Cases, 2021 25 Figure 7. Medical vs Surgical Admission per 100 People, 2021 25 Figure 8. Medical A-DRGs with Similar Prevalence in All Countries (Group 1) 29 Figure 9. Medical A-DRGs More Prevalent in Ukraine and/or Moldova (Group 2) 30 Figure 10. Medical A-DRGs Less Widely Provided in Ukraine and/or Moldova (Group 3) 31 Figure 11. Medical A-DRGs Widely Provided in Countries Other Than Ukraine and/or Moldova (Group 4) 31 3 Figure 12. Surgical A-DRGs with Similar Prevalence in All Countries (Group 1) 32 Figure 13. Surgical A-DRGs More Prevalent in Ukraine and/or Moldova (Group 2) 33 Figure 14. Surgical A-DRGs Less Widely Provided in Ukraine and/or Moldova (Group 3) 34 Figure 15. Surgical A-DRGs Widely Provided in Countries Other Than Ukraine and/or Moldova (Group 4) 35 Figure 16. Average Length of Stay for Top Medical A-DRGs 36 Figure 17. Average Length of Stay for Top Surgical A-DRGs 37 Figure 18. Difference in ALOS across Complexity Levels within Selected A-DRGs 38 Figure 19. Average Number of Secondary Diagnosis Codes and Procedure Codes per Case 44 Figure 20. Appendectomy Cases: Coding Activity 45 Figure 21. C-Section Cases: Coding Activity 46 Figure 22. Vaginal Delivery: Coding Activity 46 Figure 23. AMI: Coding Activity 47 Figure 24. Appendectomy Performance Indicators 48 Figure 25. C-Section Performance Indicators 49 Figure 26. Vaginal Delivery Performance Indicators 49 Figure 27. AMI Performance Indicators 50 Figure 28. Coding Activity and ALOS for Vaginal Delivery in Ukrainian Maternity Hospitals 51 Figure 29. Histogram of ALOS in Ukraine Maternity Hospital 5 (Total of 2,229 Cases) 52 Figure 30. Number of Different DRGs Used to Group Appendectomy Cases, by Hospital 52 Tables Table 1. Distribution of Cases by Country, Case Type, and Hospital Type 19 Table 2. Distribution of Hospital-Level Data by Country and Case Type 20 Table 3. Basic Indicators for Study Countries, 2019 20 Table 4. Distribution of Hospital Cases in Study Countries by Top-10 MDCs, 2021 21 Table 5. COVID-19 DRGs: Number of Cases and Cases as Percentage of Total Cases, 2021 27 Table 6. Delivery-Related DRGs: Number of Cases and Cases as Percentage of Total Cases, 2021 28 Table 7. Top-Three Main Diagnosis Codes in Study Countries 39 Table 8. Top-Ten Main Diagnosis Codes and Share of Cases in Study Countries 39 Table 9. Example of C-Section Procedure Codes in Study Countries 40 Table 10. Top-Three Surgical Procedure Codes in Study Countries 41 Table 11. Top-Ten Surgical Procedure Codes as Share of Cases in Study Countries 41 Table 12. Comparison of DRGs Used for Grouping Appendectomy Cases in Croatia and Ukraine 53 Table 1A-1. Reasons for Excluding the Data from Analysis 64 Table 2A-1. Country-Level Data 66 Table 2A-2. Hospital-Level Data 67 Table 4A-1. Summary of Countries’ DRG System 71 Table 6A-1. Difference in ALOS across Complexity Levels 81 4 Acknowledgments This study was prepared by a World Bank team led by Olena Doroshenko (Senior Economist, Health) and comprising Kristiina Kahur (Consultant), Marta Kuzmyn (Consultant), Sarah Bales (Consultant), and Solomiya Kasyanchuk (Consultant). The study was prepared under the oversight and management of Tania Dmytraczenko (Practice Manager) and with support from Caryn Bredenkamp (Lead Economist, Program Leader). The authors are grateful to the World Bank for publishing this report as an HNP Discussion Paper. The team benefited from excellent comments by the peer reviewers: Owen Smiths (Lead Economist, Health) and Christophe Lemiere (Practice Leader, Human Development). This study was made possible with the help and support of country experts, who provided access to data and supported analysis and interpretation of data: • Tatjana Trupec (Croatia, Consultant) • Riho Peek and Malle Avarsoo (Estonia, Health Insurance Fund) • Laszlo Lorenzovici and team (Romania, Hospital Consulting SRL) • Adrian Barba and Larisa Solomon (Moldova, National Health Insurance Company) • Serhiy Voinalovych (Ukraine, Consultant) and team of the National Health Service of Ukraine led • Hyeseung Wee, Sejin Bae, Yoonkyung Choi (Republic of Korea, National Health Insurance Service), and Younghee Kim (Institutes of Green Bio Science & Technology Center, Seoul National University) This study was implemented with the support of the Korea–World Bank Partnership Facility. 5 Abbreviations ACHI Australian Classification of Health Interventions A-DRG Adjacent diagnosis-related group ALOS Average length of stay AMI Acute myocardial infarction AR-DRG Australian Refined diagnosis-related group CHIF Croatian Health Insurance Fund CNAM National Health Insurance Company (Moldova) DRG Diagnosis-related group EHIF Estonian Health Insurance Fund EU European Union FFS Fee-for-service GDP Gross domestic product GI General intervention ICD-10 International Classification of Diseases-10th Revision ICP Interventional Coronary Procedures LOS Length of stay MDC Major diagnostic category NHIS National Health Insurance Service (Korea) NHSU National Health Service of Ukraine NOMESCO Nordic Medico-Statistical Committee NordDRG Nordic diagnosis-related group OECD Organisation for Economic Co-operation and Development OOP Out-of-pocket OR Operating room PPP Purchasing power parity RO-DRG Romania-diagnosis-related group VHI Voluntary health insurance 6 Executive Summary Understanding hospital performance variation through benchmarking motivated this study to compare hospital performance across six countries. Diagnosis-related groups (DRGs) provided a useful framework for comparison across these countries despite differences in their DRG classifications and health care and provider payment systems. Cross-country and cross-facility comparisons of inpatient case-mix can help identify relative over- or under- provision of share of cases and length of stay, which can orient and guide hospital reforms to ensure efficiency and patient access to hospital services. Identification of the most prevalent case types can also help to focus attention on a small number of conditions affecting large shares of the population. This study contributes to the literature on cross-country studies of DRG systems and of hospital performance. Methodology and data This study performs benchmark comparison across countries and across facilities within countries using DRG classification as a tool for measuring hospital output. DRG classification allows comparisons of the share of cases in major diagnostic categories (MDCs), which correspond to body systems or etiology. Within MDCs, surgical versus medical cases can be compared, and analysis can also be carried out at the more granular adjacent DRG (A- DRG) level, which corresponds to major subcategories within MDCs. Cross-country analysis compares MDCs, A-DRGs, and diagnosis and procedure codes. Hospital-level analysis examines hospital coding (diagnosis and procedures) and length of stay in a subset of the most treated case types in the six countries (appendectomy, C-section, vaginal delivery, and medical treatment of acute myocardial infarction [AMI]). Ukraine and Moldova are compared with Croatia, Romania, Estonia, and the Republic of Korea to draw lessons on hospital performance. Ukraine, Moldova, Croatia, and Romania all use a DRG classification based on the Australian Refined DRG (AR-DRG), which differs from Korea’s and Estonia’s DRG classifications. 1 Analysis of results acknowledges cross-country differences in DRG systems and important socioeconomic, demographic, and health system features. The work involved country teams in study design to achieve comparability of case and indicator definitions and appropriate interpretation of results. Country teams submitted data following a standard template and instructions, with analysis highlighting the 2021 results. Due to data protection issues and time constraints, Korea could only share MDC-level data for the cross-country analysis. Findings Analysis of MDCs revealed that 75–79 percent of inpatient cases are concentrated within the top-10 MDCs. The same six MDCs (diseases and disorders of the circulatory system; musculoskeletal system; respiratory system; pregnancy, childbirth, and the puerperium; digestive system; and nervous system) are among the top 10 in almost all the countries, although with slightly different rankings across countries. The admission rates within top MDCs show significant variations across countries, with some notable patterns. While circulatory diseases have consistent admission rates, other MDCs exhibit inconsistencies. Korea has higher admission rates for digestive, 1 Korea developed its own DRG system; Estonia uses the Nordic DRG system. 7 musculoskeletal, and nervous system disorders, while Moldova has a higher admission rate for respiratory diseases, possibly influenced by COVID-19 cases. Surprisingly, Croatia and Korea, both having a higher proportion of older population, show lower admission rates for nervous system disorders. Estonia demonstrates a higher admission rate for musculoskeletal disorders, possibly because of better access to orthopedic surgeries. The analysis of medical and surgical admission rates reveals interesting patterns in Korea, Ukraine, and Moldova. Korea and Moldova have exceptionally high rates of medical admissions, the rate of surgical admissions was found to be very high in Korea and low in Ukraine. These findings may reflect disparities in condition prevalence, differences in clinical practices, or variances in patient preferences among these countries. The share of inpatient admissions for surgery ranges from nearly 40 in Croatia to just over 20 percent in Ukraine, Moldova, and Korea. Surgeries are concentrated in a limited number of types. For example, in Ukraine, each of top-20 surgery A-DRGs accounts for more than 1 percent of all surgery episodes, and top-32 surgery A-DRGs account for nearly 60 of all surgery episodes. In Moldova, each of top-21 surgery A-DRGs accounts for more than 1 percent of all surgery episodes, and the top-31 surgery A-DRGs collectively account for nearly 62 percent of all surgery episodes. The most prevalent A-DRGs for surgery in Ukraine are C- section, diagnostic curettage or hysteroscopy, and hernia procedures, while in Moldova the most prevalent are vaginal delivery with operating room procedure, C-section, and eye procedures (glaucoma and complex cataract procedures or penetrating eye injury). The top-20 medical A-DRGs account for 42–65 percent of all medical admissions, depending on the country. Medical A-DRGs are also concentrated in a relatively small number of case types. In Ukraine and Moldova, 20 medical A-DRGs each account for more than 1 percent of medical cases. In Ukraine the most prevalent medical A-DRGs are respiratory infections and inflammations, vaginal deliveries, and coronary atherosclerosis. In Moldova the most prevalent medical A-DRGs are respiratory infections and inflammations, neonate with normal weight, and other disorders of the nervous system. COVID-19 affected the structure of hospital admissions in all countries during 2021, but differences in DRG assignment rules confounded comparisons. The share of admissions in DRGs related to COVID-19 (primarily respiratory disorders) ranges from 7.7 percent in Estonia to 15.2 percent in Moldova. Ukraine is the only country with a specific COVID-19 DRG. Nordic DRG (NordDRG) classification of respiratory diseases used in Estonia was very different from classification by countries using AR-DRG. Deliveries are among the most prevalent DRGs in all countries, ranging from 5 percent of all admissions in Ukraine and Moldova to 8 percent in Croatia and Estonia. C-sections account for 1–2 percent of all admissions in most countries. Romania stands out as an outlier on structure of deliveries, with surgical deliveries (C-sections and vaginal delivery with operating room procedure) accounting for 85 percent of all deliveries, and with the share of C-sections out of all deliveries two to four times higher than in the other countries. Cross-country comparison reveals certain A-DRGs that account for a higher share of all cases in Ukraine and/or Moldova compared to the other countries. These include surgical A-DRGs like diagnostic curettage or hysteroscopy, abortion with operating room procedure, anal and stomal procedures, and hernia procedures. Some surgical A-DRGs have a greater prevalence in Ukraine but not in Moldova; examples include lens procedures; hysterectomy 8 for non malignancy; gastroscopy; and other vagina, cervix, and vulva procedures. On the other hand, glaucoma and complex cataract procedures are most prevalent in Moldova but not in Ukraine. Similar results are found for medical A-DRGs. In contrast, some A-DRGs are less prevalent in Ukraine and Moldova, such as interventional coronary procedures for AMI, major small and large bowel procedures, other back and neck procedures, hip replacement, and other hip and femur procedures. Higher than normal shares can be further investigated to determine whether these anomalies may be due to excess admission for cases treated as outpatient in other countries. Similarly, lower than normal shares may indicate underprovision of certain services in these countries. Similar analysis of anomalies in average length of stay (ALOS) reveals specific medical or surgical A-DRGs where length of stay may be excessively long. Long length of stay is particularly problematic in Ukraine and Moldova, such as in the case of coronary atherosclerosis or hypertension (on average over 10 days in Ukraine, and 8 days in Moldova). Similar results are found with length of stay for surgical A-DRGs; ALOS for major procedures for breast disorders reaches 17 days in Ukraine and 10 days in Moldova, much higher than in the other countries. Cross-facility comparisons for four case types found distinct coding patterns across countries. A higher average number of secondary diagnoses per case was coded in hospitals in Croatia and Romania compared to the other countries. Croatia also coded a consistently higher number of procedure codes per case for most conditions studied (except AMI), while Ukraine had the highest average number of procedure codes for AMI. Explanations for these patterns reflected data submission and utilization rules for DRG grouping, such as database size limits (Romania), exclusion of procedure codes that do not influence DRG grouping (Estonia), or hospital payments that do not reflect procedure codes (Ukraine). In contrast, Croatia’s high number of clinical codes reflects a provider payment system that pays more for cases with more secondary diagnoses, patient documentation practices encouraging completeness, and data entry interfaces that start with fully loaded sets of procedures and require removal of unprovided items. Cross-facility comparisons also revealed important coding variation across facility types for the selected four case types. Substantial cross-facility variation in secondary diagnosis and procedure coding in the same country for the same case type was apparent for some conditions and countries, while uniform coding across facilities was evident in other countries for the same case types. For example, appendectomy cases in Ukraine had on average about one secondary diagnosis and one procedure code across all facilities. However, for appendectomy cases in Croatia, the average number of secondary diagnoses varied from 3.5 to 4.6, and the average number of procedures varied from 5.7 to 10.0, depending on the facility. In contrast, C-section coding practices in Ukraine varied substantially across facilities, with an average of 1.9 to 5.4 secondary diagnosis codes and 2.0 to 12.7 procedure codes per case. Once this variation is recognized, it can be further investigated, better understood, and potentially reduced. Consistency between severity (measured by the number of secondary diagnosis codes) and intensity of treatment (measured by ALOS) was also assessed in cross-country comparisons for each case type. For some conditions and some countries, these two measures are correlated with each other. For example, in Moldova, AMI medical cases had little variation in ALOS or in number of secondary diagnoses, while for C-sections, facilities with a higher 9 number of secondary diagnoses also tended to have a longer ALOS. Ukraine almost universally had long length of stay regardless of the number of secondary diagnoses coded. Discussion of study limitations The reliability and quality of data are crucial for cross-country comparisons but collecting data from multiple countries presented challenges. The countries included in this study collected data using different scopes (all included acute inpatient care, but some also included same-day cases) and classifications (four countries use a version of Australian DRG, but Ukraine groups only to A-DRGs, while Korea and Estonia used entirely different DRG classifications). Despite use of common data extraction templates and definitions, comparability was still limited due to differences in coding accuracy and classification systems. Coding accuracy could not be assessed with the information collected in the study. The use of different procedure classifications (Korea and Estonia) and use of different versions of the Australian classification of health interventions in the remaining four countries also inhibited comparison of procedure coding patterns in the six countries. Cross-facility comparisons within each country were hindered by limited information on the characteristics of the different hospitals. Future case- mix comparisons can target these specific areas to improve validity of future comparisons. Conclusions The cross-country and cross-hospital benchmarking conducted in this study was able to identify positive and negative performance deviations for hospital cases organized by diagnoses and procedures that account for a high share of the total hospital outputs. In addition to identifying performance deviations, the study also used DRGs to describe the structure of case-mix in each country. This information is particularly useful for health care systems in planning their service delivery capacity. An international comparison allows for the benchmarking and comparison of countries or individual hospitals, as well as the investigation of factors contributing to variations in performance. The study found that the rate of surgical admissions was relatively uniform across four countries, but varied significantly in Korea and Ukraine. In addition to the variation in surgical admission rates across countries, the study also found that the share of cases undergoing surgery varied across MDCs and across countries within each MDC. The data analysis for different case types revealed significant variations in coding activity as well as in ALOS. Analysis of hospital cases should always relate to the organization of service delivery in countries. Recommendations Based on the study findings and conclusions, the following steps are recommended: • Strengthen comprehensiveness and accuracy of clinical coding for improved DRG classification, particularly in Ukraine, to enable moving from classification of care in A-DRGs toward a full DRG classification system. • Compare patterns of primary clinical coding in addition to comparing case-mix to identify potential anomalies that are due not to substantial differences in practice, but rather to coding practices. • Conduct routine cross-facility and cross-country comparisons of hospital case-mix to gain insights into health care delivery strengthening and to improve health system management. 10 • Establish a regional case-mix observatory to enable routine cross-country case-mix comparison at the global or regional level. • Devote adequate effort and resources to international collaboration to ensure valid results if cross-country hospital activity benchmarking is institutionalized. 11 1 Introduction This report presents the findings of a cross-country comparative study aimed at assessing and comparing the performance and coding activity of hospitals across six different countries. The study was motivated by the need to understand the reasons for variation in hospital performance and to benchmark the performance of hospitals against their counterparts in other countries. In addition to providing an overview of the participating countries’ health systems, with a focus on hospitals and the use of diagnosis-related groups (DRGs) for classifying and coding hospital outputs, the study also compares country-level and hospital- level performance in each country using a range of performance indicators such as length of stay, complexity of cases, and coding of procedures. While the study does not include an analysis of costs and tariffs, it sheds light on the factors that contribute to differences in hospital performance across countries. It also highlights the implications of these findings for policy makers and health care professionals. The study aims to gain a deeper understanding of the differences and similarities in hospital outputs and performance across a selection of countries, namely Croatia, Estonia, Moldova, Republic of Korea, Romania, and Ukraine. It focuses on Ukraine, which has recently adopted the use of DRGs, and Moldova, which has a longer history of DRG implementation, and on their efforts to optimize hospital performance. But it also provides useful insights from comparisons across other countries. The study aims to propose methods for comparing hospital outputs across countries that use different systems for coding hospital activities. It also helps identify areas for further investigation and helps determine how the experiences of different countries may be useful for the advancement of the hospital sector. Through this cross-country and cross-facility comparison, the study seeks to offer insights into hospital performance and identify best practices that can be shared across borders. The study is inherently exploratory in nature. Through this study, authors aim to generate insights into the distribution, variation, and complexity of cases, laying the groundwork for further investigations and informed decision-making in health care management. Hospitals play a critical role in the health care system, but excess reliance on hospitals can be inefficient and lead to suboptimal use of resources. Hospitals provide needed specialized and inpatient care to patients with acute and chronic conditions. They serve as first responders, preserving health and offering interventions when they matter most. Moreover, hospitals facilitate medical research and serve as a vital platform for students to gain practical skills. While they set the standards for national health systems, excess reliance on hospitals can lead to inefficiencies and waste scarce resources. The high cost of hospital care necessitates reforms that can achieve optimal health outcomes and outputs within budget constraints. There is thus a growing consensus that hospital reforms are essential to improve the efficiency of health care systems and provide high-quality, cost-effective care to patients. Cross-country and cross-facility comparison of inpatient case-mix has the potential to identify over- or under-provided case types in hospitals and can offer useful insights to orient and guide hospital reforms. Similarities in DRG classification, such as the classification of cases into chapters based on major body systems and similarities in the major surgical and nonsurgical first-level classifications within each body system, make it possible to compare case-mix across countries and facilities. By analyzing variation in patient case-mix across countries or across facilities within a country, in comparison to epidemiological, demographic, health system, and other factors, useful insights can be gained into the health 12 needs of the population and the efficiency of health care delivery. These insights can inform policy decisions and help health care providers to improve the quality and cost-effectiveness of care. The proposed study complements several studies that have been conducted in the past to explore the design features and performance of DRG systems across different countries. One such study, conducted by Busse and others in 2011, compared the DRG systems of 11 European countries, specifically Austria, England, Estonia, Finland, France, Germany, Ireland, the Netherlands, Poland, Sweden, and Spain. Another study (Street et al. 2012) focused on comparing DRG classification algorithms and indicators of resource consumption, such as cost or length of stay, for different clinical issues, including cholecystectomy, hip and knee replacement, childbirth, acute myocardial infarction (AMI), appendectomy, and stroke. The researchers aimed to identify variations in resource use across hospitals and patients with different conditions and proposed an analytical strategy to examine these variations. This study seeks to add to the knowledge available on comparison of hospital performance across countries using similar systems of coding hospital cases. For example, in 2013, Medin and others conducted a hospital productivity comparison among Nordic countries by using DRGs to measure outputs. This study focused on describing the methodological challenges intrinsic in international comparative studies of hospital productivity and how these challenges have been addressed within the context of hospital comparisons in the Nordic countries. The hospital sectors in the Nordic countries provide good examples for international comparison, as the characteristics of Nordic hospitals are in many ways harmonized; they have administrative data at the hospital level, they all apply a similar secondary patient classification system, and they exhibit a rather similar structure of the organization for hospital care. This study aims to analyze country-level and hospital-level data from several countries that have different systems. The countries included in the study are different in terms of size and structure of hospital networks, and they use different DRG systems. The study seeks to shed light on the potential relevance of cross-country comparisons in supporting decision- making related to improving hospital-level performance. Following this introduction, which lays out the context for and purpose of the study, the report is organized into five sections that present the methodology, findings on the country level, findings on the hospital level, summary and conclusions, and recommendations, followed by annexes. The methodology section describes the process and methods used in the study, including data collection and analysis. The section on country-level findings presents the results of the cross-country comparison, while the section on hospital- level findings presents the results of the cross-facility comparison within countries. The summary and conclusions section provides an overview of the main findings and draws conclusions from the study. The recommendations section has suggestions for conducting further analysis, organizing cross-country case-mix benchmarking as a routine activity that benefits participating countries, and implementing cross-facility benchmarking within countries. Annexes are an integral part of this report. They provide detailed information about health care in the countries included in the study, templates for data collection, additional details about the analyses conducted, and much more. 13 2 Methodology and Data Description 2.1 Methodology The main methodology used in this study is benchmark comparison across countries and across facilities within countries using DRG classification as a tool for measuring hospital output. Aspects of hospital care covered in this study include structure of inpatient episodes and admission rates across specialties (corresponding closely to major body systems), surgical versus nonsurgical admissions, and coding of inpatient cases. To help understand variation in case-mix across countries, the study also examines differences in DRG classifications used and other health system features, specific payment arrangements that may influence behavior of health care provider coding, and scope of care hospitals provide to patients, but the study does not examine the payments for hospital care or hospital efficiency. 2.1.1 DRG classification The analysis of the hospital outputs was organized using the following principles and concepts of DRG classification (see also Figure 1): • Major diagnostic categories (MDCs). Hospital cases are assigned to major diagnostic categories corresponding to organ systems (neurological disorders, eye disorders, etc.) or etiology (e.g., burns, human immunodeficiency virus [HIV]) based on the primary diagnosis codes. In some cases, pre-MDC is used for cases with high-cost procedures such as organ transplantations without considering the principal diagnosis. • Medical/surgical split. Within each MDC, a split on “surgical,” “medical,” and “other” partitions is performed. The surgical partition does not necessarily exist in every country. • Adjacent DRGs (A-DRGs). The cases are further organized into intermediate categories with similar clinical features based on operating room procedures performed or the main diagnosis. In this study they are called adjacent DRGs. This terminology is used because four countries have adopted a DRG system originated from the Australian Refined AR-DRG system, where the A-DRGs are used. Depending on case-mix system, intermediate categories can be called base/basic DRGs, disease cluster, or may not have a specific name. • Diagnosis-related group (DRG). DRGs are organized within an A-DRG on the basis of clinical complexity, in most cases through assessment of secondary diagnosis codes. Not all secondary diagnosis codes have an impact on cost. Those with no impact on costs are therefore not considered in determining complexity level. Using complexity levels allows the organization of clinically meaningful and economically homogenous groups. The number of complexity levels varies across different DRG systems. • Case-mix. Case-mix is a term used to describe hospital output, in this study for inpatient admissions. It consists of information on total cases or shares of cases in each specialty, surgical/medical partitions, and more detailed case groups (A-DRGs or DRGs). 14 Figure 1. Overview of the DRG Algorithm . Demographic and clinical edits Pa�ent record: (Demographic and assess the validity of cases. clinical informa�on). Cases are assigned to MDC based on the principal diagnosis. Pre Pre-MDC MDC MDC MDC combines high-cost cases. Surgical Medical Other Surgical Medical Other Cases are divided into par��ons. Surgical Medical Other Cases are assigned to A-DRGs A-DRG A-DRG A-DRG A-DRG A-DRG mainly based on diagnoses and A-DRG A-DRG interven�ons. Cases are assigned to DRGs mainly DRG DRG DRG DRG based on the case clinical complexity. DRG Source: Prepared by authors based on AR-DRG scheme (https://www.ihacpa.gov.au/sites/default/files/2023-03/ar- drg_overview_and_version_11.0_development.pdf). Notes: MDC = Major diagnostic category; A-DRGs = Adjacent diagnosis-related groups. To ensure comparability, some minor adjustments were needed in certain aspects of DRG classification, which are laid out in Box 1. Box 1. Main Steps for Processing the Country-Level Data MDCs • National MDCs were linked to so-called adjusted MDCs to be able to compare the country-level data based on MDCs. Despite the fact that countries have different DRG systems (Nordic DRG or NordDRG vs. Australian Refined DRG [AR-DRG] or its modifications), the MDCs are to a large extent the same across the countries, and the adjustments were minor. o For example, in Croatia MDC 15 text is “Newborns and Other Neonates (Perinatal Period)”; in Ukraine and Romania it is “Newborns and Other Neonates”; and in Estonia it is “Newborns and Other Neonates with Conditions Originating in the Perinatal Period.” The adjusted MDC text used in the analysis was “Newborns and Other Neonates.” • Ukraine has a separate MDC for COVID-19 that was not present in other countries, and therefore this MDC was not considered in the comparison. • The top-10 MDCs for each country were identified and used for further analysis. • Out of the top-10 MDCs, MDCs common for all countries were identified. DRGs • Adjusted MDC information was linked to each country's DRGs to make the DRG data comparable on the MDC level. • In the Croatian, Moldovan, and Romanian data, in addition to “surgical” and “medical” DRGs, there were “other” DRGs. For the analysis, the latter category was included in “medical” DRGs. • The top-20 medical and surgical DRGs were identified for each country. 15 • Based on the above information, the most widespread and common clinical areas were identified for each country and used in determining the scope of hospital-level analysis. Main diagnoses • International Classification of Diseases—10th Revision (ICD-10) codes with more than three digits were converted to the three character–level diagnosis codes (e.g., instead of code I50.0, code I50 was used) to decrease the granularity of diagnosis information and thus to facilitate cross-country analysis. Procedures For identification of significant surgical procedure codes, the following approach was used: • Estonian data: Only the codes from the main chapters (A–Q) of the classification (Nordic Medico-Statistical Committee Classification of Surgical Procedures [NOMESCO]) were used. The approach excluded codes of subsidiary chapters (T, U, X, Y) containing therapeutic and investigative procedures associated with surgery, and supplementary chapter (Z) containing general procedure qualifiers. Other countries’ data: Only the codes corresponding to general intervention (GI) from the Australian Classification of Health Interventions (ACHI) were included. GI codes are intervention codes that are considered significant in the AR-DRG system. Source: Prepared by authors. Notes: MDC = Major diagnostic category; DRG = Diagnostic-related group. 2.1.2 Levels of analysis Two levels of analysis were included in the study: country-level and hospital-level analysis. The country-level analysis aimed to identify common hospital cases across all countries and analyze hospital outputs at an aggregated level, using country-level statistics and information on MDCs, A-DRGs/DRGs, diagnosis codes, and procedure codes. The hospital- level analysis included a subset of the most frequently treated cases in hospitals across the comparison countries, namely AMI without surgery, vaginal delivery, C-section, and appendectomy. The hospital-level analysis examined the relationship between hospital coding activities and hospital performance. It required extensive data processing and cleaning to ensure comparability of results (see Annex 1 for details). The study team excluded COVID- 19-related cases to minimize the pandemic's impact on the results. The included cases were defined using the EURODRG Project vignettes (Busse 2012) as a reference, with adjustments made to fit the context of the current study, including data collection and timing constraints. The findings of the study were discussed with the teams of the participating countries. To validate and contextualize the findings, the research team shared them with the participating countries and incorporated their feedback. Additionally, a workshop was held to present the results and provide recommendations for next steps. This workshop helped identify areas where hospital performance was similar to or diverged from that of other countries, and it enabled discussion of areas that may need further research. Country teams also provided additional insights to help interpret the findings. 2.2 Data sources and content All data included in the analysis were requested from country teams using a standardized approach. Data were collected from national health systems by the country experts. Data were obtained from each country's respective national database or register using two separate templates, one for country-level data and one for hospital-level data (templates are provided in Annex 2). In addition, information on DRG systems in each country was gathered using 16 specially designed questionnaire (Annex 3). The templates and questionnaire parameters were discussed with country representatives in a virtual meeting, and any question related to the templates was discussed and clarified. The data analyzed in the study were extracts from routine data collection in each country; the analysis did not require additional data collection. Although data for this study were collected from 2019 to 2021, only the full year of 2021 is presented in the analysis. The consistent use of data for one full year allowed the study team to compare the frequencies of different hospital cases and assume similar potential impact of COVID-19 on hospital care utilization. In addition, Ukraine introduced its DRG system in the middle of 2020, so full-year DRG-coded hospital data for Ukraine were available only for 2021. The study involves data from six countries, selected based on various considerations. Ukraine and Moldova were chosen as the primary beneficiaries of the study. Ukraine had recently implemented a major health care reform that included the adoption of a DRG system, while Moldova has a long history of using DRGs and was seeking ways to optimize hospital performance. Romania and Croatia were selected as relevant benchmark countries because their hospital system organization is similar to that of Ukraine and Moldova; they use AR- DRGs, like Ukraine and Moldova; and they likewise seek to improve hospital efficiency and quality. Estonia, which has a different DRG system (the Nordic system), was included to demonstrate the robustness of case-mix comparisons even when using different DRG classifications. Finally, Korea was included as a country facing similar challenges in reforming its hospital system but offering experience from a different region. 2.3 Country-level data Countries included in the study differ in the size of their population and their hospital activity. The largest countries are Korea and Ukraine, followed by Romania. Croatia, Moldova, and Estonia are smaller and similar to one another in size. The number of hospital cases in each country corresponds to the population size (Figure 2). A total of 25 million cases were included in the country-level analysis. In terms of hospital activity in 2021, Korea provides the highest rate of hospital admissions per year at 31 per 100 people, followed by Moldova with 18 per 100 people. Other countries included in the study are quite homogeneous in their admissions rates (Figure 3). Country-level data contained information on MDC, A- DRG, DRG, diagnosis and procedure codes, and length of stay. Due to time constraints and the particularity of the Korean DRG system, only limited country-level data were available for Korea. Information was made available for 1,352,906 cases assigned to seven A-DRGs (out of 16,46,348 cases in 2021 in Korea). So, whenever possible, Korean data were compared with data from other counties, but there are certain limitations to providing comparisons consistently. 17 Figure 2. Total Population and Number of Hospital Cases per Country, 2021 Source: UN data for population; country teams for number of hospital cases per country. Figure 3. Overall Admissions per 100 People, 2021 Source: Authors’ calculations based on data provided by country teams. 2.4 Hospital-level data The study also analyzed data at the level of individual hospitals. The selection of hospitals was made in consultation with country experts and aimed to include hospitals that would be considered “typical” for each country. The intention of selecting hospitals that exceeded the threshold number of cases for selected disease types resulted in a skewed sample, predominantly consisting of large multiprofile hospitals. Nevertheless, the wide variation in length of stay (LOS) and secondary diagnoses across DRGs even within the small sample of hospitals in this study highlights the potential usefulness of benchmark comparisons for health system managers to detect outliers if all facilities were included in the sample, allowing them to identify hospitals that may require interventions to move the system toward a national standard of care. Hospitals provided data on all cases in 2021 that corresponded to the four case types: AMI (nonsurgical), appendectomy, C-section, and vaginal delivery (as detailed in Section 3.7). To ensure adequate numbers of admissions in the four case types, the initial criteria for inclusion required hospitals to have a minimum of 200 cases each of AMI and appendectomy and 400 cases each of C-section and vaginal delivery per year. However, in some countries, it proved difficult to find hospitals that met these criteria for all four case types, mainly due to the hospital sector organization and the reliance on mono-profile (specialized) hospitals. For 18 example, specialized hospitals handle childbirth and pregnancy in Ukraine, and separate cardiology centers often treat AMI patients in Ukraine and Moldova. As a result, the study included the most relevant hospitals proposed by experts that met the threshold criteria for at least one type of case. A total of 30 hospitals from six countries were included in the study. The number of hospitals per country ranged from two multiprofile hospitals in Romania, which treated all types of cases included in the study, to 12 in Ukraine, none of which included all types of cases. The Korean team analyzed the data on their own and supplied only the resulting summary’s tables and charts due to the Korean data confidentiality protection policy. Only three case types were available for the analysis from Korea (appendectomy, C-section, and vaginal delivery). Table 1 provides an overview of the number of hospitals by country, case type, and hospital type. The hospital data were anonymized, and the hospitals were divided into four categories: (1) mono-profile: cardiology, (2) mono-profile: maternity, (3) multiprofile: university, and (4) multiprofile: general (nonuniversity). Table 1. Distribution of Cases by Country, Case Type, and Hospital Type Country Hospital type Appendectomy C-section Vaginal AMI (medical Total delivery DRG) Ukraine Cardiology hospital 1 494 494 Maternity hospital 1 6 1,616 3,329 4,951 Maternity hospital 2 600 2,727 3,327 Maternity hospital 3 1 1,273 4,039 5,313 Maternity hospital 4 1 573 2,198 2,772 Maternity hospital 5 1,306 2,229 3,535 Maternity hospital 6 1,739 3,413 5,152 Multiprofile 1 856 112 968 Multiprofile 2 284 1,308 1,592 Multiprofile 3 779 779 Multiprofile 4 57 213 270 Multiprofile 5 474 474 Moldova Cardiology hospital 1 61 61 Cardiology hospital 2 12 12 Multiprofile 1 63 63 Multiprofile 2 1,342 4,908 6,250 Multiprofile 3 1,860 2,789 4,649 Multiprofile 4 387 387 Multiprofile 5 228 656 1,862 2,746 Multiprofile 6 249 249 Romania Multiprofile 1 357 2,020 1,736 81 4194 Multiprofile university 151 1,837 1,546 49 3,583 Croatia Multiprofile 1 69 163 735 44 1,011 Multiprofile 2 120 391 1,115 109 1,735 Multiprofile university 229 577 1,366 43 2,215 Estonia Multiprofile 1 783 3,018 3,801 Multiprofile 2 383 68 451 Multiprofile university 347 567 1,938 15 2,867 Korea, Rep. Multiprofile 1 172 216 203 n.a. 591 Multiprofile university 105 560 689 n.a. 1,354 Total 5,255 18,079 39,840 2,672 65,846 Source: Authors’ calculations based on data provided by country teams. Notes: AMI = Acute myocardial infarction; DRG = Diagnosis-related group; n.a. = Not available. The submitted data were examined and cleaned to remove cases that did not meet the agreed-upon data collection criteria for the specified case types. After this cleaning process, 81 percent of the originally provided data were included in the hospital-level analysis. 19 A total of 63,901 unique inpatient cases from five different countries were included in this study, supplemented by information provided by Korea team on 1,945 cases. Table 2 shows the distribution of these cases across the countries and case types. Table 2. Distribution of Hospital-Level Data by Country and Case Type Country AMI Appendectomy C-section Vaginal delivery Total Ukraine 2,127 2,458 7,107 17,935 29,627 Moldova 136 864 3,858 9,559 14,417 Romania 130 508 3,857 3,282 7,777 Croatia 196 418 1,131 3,216 4,961 Estonia 83 730 1,350 4,956 7,119 Korea, Rep. n.a. 277 776 892 1,945 Total 2,672 5,255 18,079 39,840 65,846 Source: Authors’ calculations based on data provided by country teams. Notes: AMI = Acute myocardial infarction; n.a. = Not available. The number of nonsurgical (or conservative) AMI cases was significantly higher in Ukraine than in other participating countries. The low number of AMI cases in other countries may be due to more restrictive exclusion criteria designed to filter out surgical cases. As a result, some cases may have been classified as surgical (e.g., cases involving artificial lung ventilation) and excluded from the study. Findings related to AMI cases should therefore be interpreted with caution. 3 Findings at the Country Level 3.1 Country contexts The countries included in this study differ in size, economic development, and health status (see Table 3 and Annex 5 for further details). Korea boasts the highest life expectancy— 83.3 years at birth—while life expectancy in Ukraine and Moldova is the lowest, at approximately 10 years less than in Korea. Croatia and Estonia have slightly above the European average life expectancies at 78.6 and 78.8 years, respectively. Korea spends the most on health at US$3,500 per capita in purchasing power parity (PPP) terms, followed by Croatia and Estonia at US$2,200 PPP and US$2,600 PPP per capita, respectively. In contrast, Moldova and Ukraine have the lowest per capita health expenditures, approximately US$900 PPP. Romania's life expectancy and per capita health expenditures fall between those of Croatia and Ukraine/Moldova. Table 3. Basic Indicators for Study Countries, 2019 Europe Korea, Indicator Ukraine Moldova Romania Croatia Estonia (WHO Rep. region) Total population (thousands) 43,994 2,682 19,414 4,076 1,325 51,225 Life expectancy at birth (years) 73.0 73.3 75.6 78.6 78.9 83.3 78.2 Current health expenditure, US$ per 907 860 1 907 2 168 2,617 3,521 3,215.00 capita, in PPP Current health expenditure as % GDP 7.1 6.38 5.74 6.98 6.73 8.16 7.63 Share of total current health expenditure (%) Domestic general government health 44.79 59.68 80.14 81.54 74.42 59.53 65.14 expenditure Out-of-pocket 51.12 35.69 18.88 11.46 24.05 30.25 28.75 Voluntary health insurance 3.41 4.63 0.98 6.99 1.51 10.22 6.11 Source: World Health Organization, Global Health Observatory database, https://www.who.int/data/gho/data/indicators. Notes: GDP = Gross domestic product; PPP = Purchasing power parity; WHO = World Health Organization. Countries also channel different shares of their gross domestic product (GDP) to finance health expenditures, and the composition of sources for financing of health expenditures 20 varies. In most of the countries analyzed, health expenditures as a percentage of GDP are slightly lower than the European average of 7.6 percent. Romania’s expenditure of 5.7 percent of GDP is lowest among the countries; Korea’s expenditure of 8.2 percent of GDP is the highest, and higher than the European average. Health insurance and government funds cover approximately 80 percent of all health expenditures in Croatia and Romania, resulting in low levels of out-of-pocket (OOP) spending. In contrast, OOP payments make up more than 50 percent of total health spending in Ukraine, while the government covers only 44 percent. Mandatory health insurance is the primary funding source for health care in all the countries except for Ukraine. Health expenditures are financed through payroll taxes and supplemented by state budget transfers (Croatia, Korea, Moldova) and tobacco surcharges (Croatia, Korea). In Ukraine, health care is funded through general taxation from the state budget and receives additional financing from local budgets, which directly finance utilities and capital investments at health facilities owned by local governments. In all countries, a single institution, such as the National Health Insurance in Croatia, Estonia, Korea, Moldova, and Romania or National Health Service in Ukraine, acts as the main purchaser of health services and pools funding for health care. The setup of health systems is crucial for interpreting country-level data. The study team aimed to understand hospital outputs at both an aggregated level, using MDC and DRG codes, and a more detailed level, using single diagnosis and procedure codes. The country-level analysis results were used to identify case types for further hospital-level data analysis. But country-level analysis already highlights areas of difference among the countries, which could trigger additional interest in understanding the routes of such differences. 3.2 MDC-level analysis To compare hospital outputs, the study looked at shares of cases in each country that fall within top MDCs. Such analysis helped identify the high-level distribution of cases to anatomical or clinical areas. Table 4 presents the results of the top-10 MDCs by country. The top-10 MDCs on average covered 76 percent of all cases, ranging between 72 percent and 79 percent in the individual countries. Table 4. Distribution of Hospital Cases in Study Countries by Top-10 MDCs, 2021 % of % of MDC text cases MDC text cases Diseases and Disorders of the Circulatory System 11 Diseases and Disorders of the Respiratory System 20 Diseases and Disorders of the Nervous System 11 Diseases and Disorders of the Nervous System 10 COVID-19 9 Diseases and Disorders of the Circulatory System 9 Diseases and Disorders of the Respiratory System 9 Pregnancy, Childbirth, and Puerperium 7 Pregnancy, Childbirth, and Puerperium 9 Diseases and Disorders of the Digestive System 7 Diseases and Disorders of the Musculoskeletal Diseases and Disorders of the Digestive System 8 7 System and Connective Tissue Moldova Ukraine Diseases and Disorders of the Musculoskeletal Newborns and Other Neonates 7 5 System and Connective Tissue Diseases and Disorders of the Ear, Nose, Mouth, Diseases and Disorders of the Ear, Nose, Mouth, 5 5 and Throat and Throat Diseases and Disorders of the Female Diseases and Disorders of the Hepatobiliary System 5 4 Reproductive System and Pancreas Diseases and Disorders of the Skin, Subcutaneous Diseases and Disorders of the Skin, Subcutaneous Tissue, and Breast 4 Tissue, and Breast 4 21 Diseases and Disorders of the Respiratory System 14 Diseases and Disorders of the Circulatory System 12 Diseases and Disorders of the Circulatory System 10 Diseases and Disorders of the Respiratory System 12 Diseases and Disorders of the Musculoskeletal Pregnancy, Childbirth, and Puerperium 9 10 System and Connective Tissue Diseases and Disorders of the Digestive System 9 Pregnancy, Childbirth, and Puerperium 10 Diseases and Disorders of the Musculoskeletal 8 Diseases and Disorders of the Digestive System 9 System and Connective Tissue Romania Croatia Diseases and Disorders of the Nervous System 7 Diseases and Disorders of the Nervous System 7 Diseases and Disorders of the Kidney and Urinary Newborns and Other Neonates 7 5 Tract Diseases and Disorders of the Hepatobiliary Diseases and Disorders of the Hepatobiliary 5 4 System and Pancreas System and Pancreas Diseases and Disorders of the Kidney and Urinary Diseases and Disorders of the Female Reproductive 4 3 Tract System Diseases and Disorders of the Skin, Subcutaneous Mental Diseases and Disorders 4 3 Tissue, and Breast Diseases and Disorders of the Musculoskeletal Diseases and Disorders of the Circulatory System 11 System and Connective Tissue 14 Diseases and Disorders of the Musculoskeletal 11 System and Connective Tissue Diseases and Disorders of the Nervous System 13 Diseases and Disorders of the Respiratory System 11 Mental Diseases and Disorders 12 Pregnancy, Childbirth, and Puerperium 9 Diseases and Disorders of the Digestive System 10 Korea, Rep. Diseases and Disorders of the Nervous System 8 Diseases and Disorders of the Respiratory System 6 Estonia Diseases and Disorders of the Digestive System 8 Diseases and Disorders of the Eye 5 Mental Diseases and Disorders 6 Infectious and Parasitic Disease 5 Myeloproliferative Diseases and Disorders, Poorly 5 Differentiated Neoplasms Diseases and Disorders of the Circulatory System 4 Diseases and Disorders of the Ear, Nose, Mouth, Factors Influencing Health Status and Other 5 and Throat Contacts with Health Services 4 Diseases and Disorders of the Kidney and Urinary Diseases and Disorders of the Hepatobiliary 5 Tract System and Pancreas 4 Source: Authors’ calculations based on data provided by country teams. Notes: MDC = Major diagnostic category. Colors are used to highlight presence and rank of five most popular MDCs. The six most frequently used MDCs were consistently present among the top-10 MDCs in each country except for Korea, in which the Pregnancy, Childbirth, and Puerperium MDC did not fall in the top-10 MDC categories. 2 In addition, the shares of cases in each of these six most frequently used MDCs was slightly different across the countries (see Figure 4). These similarities in leading MDCs were expected because of similarities in leading causes of morbidity, which are often the reasons for hospitalization and represent a growing health burden among the studied countries. Still, while composition of cases is similar, more similarity is observed for Croatia, Estonia, and Romania on the one hand, and for Moldova and Ukraine on the other (with the exception of a significant difference in Moldova’s share of cases in Diseases and Disorders of the Respiratory System compared to all other countries). 2When MDCs are ranked according to the number of cases falling within them, MDC Pregnancy, Childbirth, and Puerperium in Korea is 17th. The reason relates to the much lower fertility rate in Korea (0.88 versus 1.25 in Ukraine or 1.81 in Moldova). 22 Figure 4. Share of Hospital Cases by Select MDCs in Study Countries Source: Authors’ calculations based on data provided by country teams. Notes: MDCs = Major diagnostic categories. Color coding corresponds to Table 4. The consistency observed in patterns of MDCs across countries reassures us about the relatively good comparability of this case-mix data for meaningful in-depth analysis. This analysis involves a detailed examination, comparing countries at the level of A-DRGs, procedure codes, and other relevant indicators. By exploring these specific parameters, we can derive valuable insights into variation in the case-mix composition within the selected countries. While the frequencies of MDCs are similar across countries, the admission rates within each MDC exhibit significant variation, suggesting different propensities to admit patients (Figure 5). The admission rate for circulatory diseases remains relatively consistent among countries. However, other MDCs do not show such consistency. Korea has substantially higher admission rates for digestive, musculoskeletal, and nervous system disorders, while Moldova has a notably higher admission rate for respiratory diseases, which may be influenced by how COVID cases were treated. It is surprising that Croatia and Korea, both having a high proportion of older population, exhibit lower admission rates for nervous system disorders, which includes stroke. Estonia demonstrates a high admission rate for musculoskeletal disorders, potentially indicating better access to orthopedic surgeries like hip replacement. Conversely, Ukraine may have inadequate care provision for these disorders. Obstetric cases strongly correlate with the pregnancy rate, leading to births or pregnancy termination (miscarriage or abortion). Korea has a low rate of obstetric admissions because of substantially lower crude birthrate. 23 Figure 5. Admissions by Top MDCs per 100 People, 2021 Source: Authors’ calculations based on data provided by country teams. Note: MDCs = Major diagnostic categories. These variations underscore the importance of understanding the specific patterns of health care utilization within each MDC and across countries to identify areas for improvement and optimization. 3.3 Analysis of Surgical Cases versus Medical Cases The composition of medical and surgical admission differs across countries. As shown in Figure 6, there are more medical cases than surgical cases across the six countries. However, an interesting observation is that in Croatia, the share of surgical cases is higher than in the rest of the group, at around 40 percent, with medical cases accounting for slightly more than 60 percent. In contrast, in Ukraine, Moldova, and Korea, medical cases make up almost 80 percent of all cases. This difference in the proportion of surgical cases may suggest variations in inpatient care standards across these countries. 24 Figure 6. Medical vs. Surgical Cases in Study Countries: Number and as Share of All Cases, 2021 Source: Authors’ calculations based on data provided by country teams. The analysis of surgical and medical admissions rate in Korea, Ukraine, and Moldova uncovered intriguing patterns (Figure 7). In terms of medical admission, Korea and Moldova had exceptionally high rates, surpassing the other four counties. In terms of surgical admission, Korea and Ukraine exhibited significant variations, with Korea having a remarkably high rate and Ukraine showing a low rate of surgical admissions, compared with the average. These findings may indicate disparities in the prevalence of certain conditions, differences in clinical practices, or variances in patient preferences among these countries. Figure 7. Medical vs Surgical Admission per 100 People, 2021 Source: Authors’ calculations based on data provided by country teams. 25 3.4 A-DRG-level analysis The countries participating in this study have diverse experiences and backgrounds in using case-mix systems (specifically the AR-DRG, the NordDRG, and the Korea-DRG systems, which originate from the US Health Care Financing Administration DRG system). Estonia and Romania have been using DRGs since 2003, Croatia since 2009, and Moldova since 2010. Ukraine adopted AR-DRGs in April 2020. Annex 4 provides details about each country’s DRG system. Korea started implementing its own case-mix system, which initially covered only seven types of hospital cases, in 2011. The complexity of DRG systems varies. The number of DRGs varies from 372 in Ukraine to 2,721 in Korea; the other countries have approximately 700 DRGs in their systems. Day surgery or day care cases are within the scope of DRGs in Estonia and Moldova. Subacute care is typically outside the scope of DRG-based payment, although some countries use DRGs for performance monitoring of subacute care. DRGs are used for hospital care payment in each of the study countries. All of the countries combine DRG payments with other payment methods. For example, depending on the setting and services, Croatia uses DRGs combined with global budgets and fee-for-service (FFS) payments for expensive drugs and devices. Ukraine, Moldova, and Romania combine DRGs with global budgets. Estonia uses a mix of DRGs and FFS payments. In Korea, the DRG-based payments apply only to seven adjacent DRGs; fee-for-service payments are made in the other cases. However, Korea is piloting a new DRG system for a wider group of diseases, in which essential services are covered under the DRG system, and doctors’ expensive services and procedures are covered under the FFS system. 3.3.1 Selection of top A-DRGs Comparing the frequency and composition of DRGs across different countries posed a challenge for this study because the countries used varying DRG systems. To overcome this obstacle, the Ukrainian A-DRGs were chosen as the base for comparison, since Ukraine has the most aggregate approach to grouping of cases. Specifically, the approach was as follows: 1. The top-20 medical DRGs and top-16 surgical DRGs 3 for Ukraine were identified as the base for comparison with other countries. 2. The DRGs matching the identified top A-DRGs were extracted from the data sets of other countries and reconfigured in a format similar to that used by Ukraine. 3. To identify those DRGs that do not play a major role in Ukraine but might be important in other countries, top-20 medical and top-20 surgical DRGs for the remaining four European countries were analyzed and missing, but important corresponding A-DRGs from other countries were included. The actions described in step 2 were repeated for these A-DRGs. 4. Whenever feasible, similar Korean A-DRGs were included for comparison. 3 The number of analyzed medical DRGs is larger than the number of surgical ones, reflecting the difference in the shares of medical and surgical cases. Outlier DRGs such as COVID in medical DRGs and General Interventions Unrelated to Principal Diagnosis in surgical DRGs were excluded from the analysis despite their high frequency. 26 3.3.2 Impact of COVID-19 on hospital outputs The study analyzed the impacts of COVID-19 on hospital outputs. Table 5 shows the DRGs that are most closely associated with treatment of COVID-19, including the Respiratory Infection and Inflammations DRG. In Ukraine and Estonia, this DRG was not as frequently used as in the other countries; to address this discrepancy, the COVID-19 DRG was substituted in Ukraine and the Simple Pneumonia and Pleurisy DRG was substituted in Estonia. Other COVID-19-related DRGs are Respiratory Diseases with Noninvasive Ventilation, Pulmonary Edema and Respiratory Failure (in the medical category), and Ventilation > 95 Hours (in the surgical category). In 2021, COVID-19-related DRGs accounted for about 15 percent of all inpatient cases in Ukraine and Moldova, about 10 percent in Croatia and Romania, and almost 8 percent in Estonia. The discrepancy in these percentages may be due to the different roles of hospital care in the countries. In Ukraine and Moldova, the health care system is more hospital- centric, with less utilization of primary care services than in the other countries. In addition, inclusion of separate COVID-19 DRGs with attractive tariffs may have created additional incentives to treat COVID-19 in hospitals. Table 5. COVID-19 DRGs: Number of Cases and Cases as Percentage of Total Cases, 2021 DRG Ukraine Moldova Romania Croatia Estonia Cate- MDC gory COVID COVID-19 498,220 9.56% M -19 Respiratory Diseases and disorders of the Infections and 247,582 4.75% 70,211 12.75% 183,929 7.92% 27,468 5.88% 1,128 0.70% M Inflammations respiratory system Simple Pneumonia 10,630 6.58% M and Pleurisy Pulmonary Edema and Respiratory 3,255 0.06% 1,929 0.35% 21,295 0.92% 5,373 1.15% 692 0.43% M Failure Respiratory Disease with Noninvasive 3,419 0.07% 8,905 1.62% 19,730 0.85% 5,190 1.11% M Ventilation MDC Tracheostomy or 00 Pre- Ventilation > 95 1,810 0.03% 2,818 0.51% 23,703 1.02% 6,037 1.29% S MDC Hours TOTAL COVID- 754,286 14.47% 83,863 15.23% 248,657 10.71% 44,068 9.43% 12,450 7.70% 19 cases Source: Authors’ calculations based on data provided by country teams. Notes: MDC = Major diagnostic category; DRG = Diagnostic-related group; M = Medical DRG; S = Surgical DRG. Korean DRGs data related to COVID-19 were not available. 3.3.3 Hospital outputs related to deliveries The study compared the shares of hospital cases related to deliveries, since this group of cases is usually one of the largest. Table 6 presents an analysis of hospital delivery cases from both medical and surgical partitions. Deliveries accounted for 7–8 percent of all cases in Croatia, Estonia, and Romania, while in Moldova and Ukraine they made up only about 5 percent of total cases. It is unlikely that this difference is due to variations in birthrates, but rather may reflect differences in the approach to inpatient care. In Moldova and Ukraine, inpatient care may be provided for a wider range of diagnoses, some of which are likely managed on an outpatient basis in Croatia, Estonia, and Romania. 27 Table 6. Delivery-Related DRGs: Number of Cases and Cases as Percentage of Total Cases, 2021 Share Number MDC Country DRG in total Medical / surgical / total of cases (%) Ukrainea Cesarean Delivery O01 66,875 1 S Vaginal Delivery w OR Procedures O02 754 0 Vaginal Delivery O60 180,392 3 M Total 248,022 5 T Moldova O01 Cesarean Delivery 6,419 1 O02 Vaginal Delivery with an 6,680 1 S Operational Procedure Pregnancy, childbirth, and puerperium O60 Vaginal Delivery 14,252 3 M Total 27,351 5 T Romania O101 Cesarean Delivery 82,855 4 S O102 Vaginal Delivery with an 42,174 2 Operational Procedure O301 Vaginal Delivery 26,296 1 M Total 151,325 7 T Croatia O01 Cesarean Delivery 9,691 2 O02 Vaginal Delivery with an 3,045 1 S Operational Procedure O60 Vaginal Delivery 22,626 5 M Total 35,362 8 T Estonia 370 & 371 Cesarean Delivery 2,630 2 n.a.b Vaginal Delivery with an — — S Operational Procedure 372–375 Vaginal Delivery 9,991 6 M Total 12,621 8 T Korea, Rep. O0160 Cesarean Delivery (First Fetus) 140,623 S O017 Cesarean Delivery (Multiple) 6,527 S Total (only C-section) 147,150 1 T Source: Authors’ calculations based on data provided by country teams. Notes: MDC = Major diagnostic category; DRGs = Diagnostic-related groups; M = Medical; OR = Operating room; S = Surgical, T = Total. a. Data for Ukraine were approximated. There were four “Delivery” DRGs in Ukraine: those shown in the table (O01, O02, and O60), plus O67 Delivery. The former three DRGs existed in Q1 2021 and were substituted with O67 for Q2–Q4 2021. This substitution was aimed at simplifying coding activities, as the payments were identical for all deliveries irrespective of their type. To split O67 between the other three DRGs, their shares in Q1 2021 were applied. b. n.a. = Not applicable. In Estonia there is no separate DRG for Vaginal Delivery with an Operational Procedure. There was also variation in shares of medical and surgical deliveries. While vaginal delivery was the most common type of delivery in most of the countries (accounting for three- quarters of all delivery cases in Estonia, about two-thirds in Croatia and Ukraine, and more than half in Moldova). In Romania, more than 50 percent of all cases were C-sections, followed by vaginal deliveries that required operational procedures (30 percent of all cases); standard vaginal deliveries in the medical partition accounted for only 17 percent of all cases. This discrepancy in Romania may be due to financial incentives for deliveries that require surgery. In addition to Romania, Moldova also had a significant share of deliveries coded as Vaginal Delivery with Operational Procedures (double the number of C-sections). The only data available for Korea were on C-sections. Table 6 shows only C-sections for Korea, as vaginal delivery cases are not included in the DRG system. The share of C-section cases in Korea is relatively low, accounting for 1 percent of total inpatient care, which is comparable to rates in Moldova and Ukraine. This low share is consistent with the low number of cases in the corresponding MDC (Pregnancy, Childbirth, and Puerperium). 28 3.3.4 Analysis of medical A-DRGs The patterns for medical A-DRGs can be organized in four distinct groups based on their frequency across countries. 4 Figure 8, Figure 9, Figure 10, and Figure 11 present the top medical A-DRGs, which represent almost half of all medical cases in comparison countries (ranging from 42 percent in Estonia to 65 percent in Moldova). Higher shares of inpatient cases in certain A-DRGs might be related to higher propensity to admit patients who could be treated on an outpatient basis. Group 1 includes DRGs that are equally important across all countries, including those for stroke, schizophrenia, and diabetes (Figure 8). Low cross-country variation in the share of inpatient cases assigned to certain A-DRGs gives reassurance about access and appropriateness of inpatient services for these case types. Figure 8. Medical A-DRGs with Similar Prevalence in All Countries (Group 1) Source: Authors’ calculations based on data provided by country teams. Notes: A-DRG = Adjacent diagnosis-related group. Korea is not presented in this figure as Korean country-level data could not be split by DRGs (such split is available only for 1,352,906 cases, grouped by seven A-DRGs, out of total 16,46,348 cases in 2021). Group 2 includes DRGs that are relatively more frequent in Ukraine and Moldova than in the other three countries. This is the largest group and includes cases like asthma, hypertension, bone diseases, alcohol intoxication, otitis, etc. (Figure 9). Some cases could be related to inadequate preventive health measures (i.e., hypertension, nonsurgical spinal disorders). Cases in this group in Moldova and Ukraine need further investigation to promote their convergence to the shares of other countries. 4 Some A-DRGs are represented in two groups. 29 Figure 9. Medical A-DRGs More Prevalent in Ukraine and/or Moldova (Group 2) Source: Authors’ calculations based on data provided by country teams. Notes: A-DRG = Adjacent diagnosis-related group; g = Grams. Korea is not presented in this figure as Korean country-level data could not be split by DRGs (such split is available only for 1,352,906 cases, grouped by seven A-DRGs, out of total 16,46,348 cases in 2021). Group 3 includes DRGs that show similar importance across several countries but are missing or have a low share in Ukraine or Moldova (Figure 10). It includes heart failure or septicemia. These lower or missing shares are red flags indicating an issue that should be addressed—or the reasons for these deviations should at least be identified. 30 Figure 10. Medical A-DRGs Less Widely Provided in Ukraine and/or Moldova (Group 3) Source: Authors’ calculations based on data provided by country teams. Notes: A-DRG = Adjacent diagnosis-related group. Korea is not presented in this figure as Korean country-level data could not be split by DRGs (such split is available only for 1,352,906 cases, grouped by seven A-DRGs, out of total 16,46,348 cases in 2021). Group 4 includes DRGs whose patterns varied in countries other than Ukraine and Moldova and that might be candidates for investigation (Figure 11). It includes digestive malignancy and septicemia in Croatia and neonate admission and cirrhosis/alcoholic hepatitis in Romania (shares are similar to those in Moldova). Figure 11. Medical A-DRGs Widely Provided in Countries Other Than Ukraine and/or Moldova (Group 4) Source: Authors’ calculations based on data provided by country teams. Notes: -DRG = Adjacent diagnosis-related group; g = Grams. Korea is not presented in this figure as Korean country-level data could not be split by DRGs (such split is available only for 1,352,906 cases, grouped by seven A-DRGs, out of total 16,46,348 cases in 2021). 3.3.5 Analysis of surgical A-DRGs Surgical A-DRGs can also be organized in four groups. These A-DRGs, presented in Figure 12, Figure 13, Figure 14, and Figure 15, account for approximately 60 percent of all surgical cases (except in Romania, where they account for only half of the cases). They represent different MDCs, but mostly fall into previously identified top-six MDC groups; depending on 31 the country, the share of cases from these MDCs varies from 47 percent to 71 percent of all cases presented in this section. Group 1 includes A-DRGs that are equally important across all countries, suggesting similar and potentially appropriate current levels of service provision. This group includes Appendectomy; Humerus, Tibia, Fibula, and Ankle Procedures; Tonsillectomy and Adenoidectomy; Laparoscopic Cholecystectomy; and a few others (see Figure 12). However, this grouping is somewhat arbitrary. For example, for Appendectomy or Laparoscopic Cholecystectomy, the share of cases is somewhat lower in Romania: about 1 percent of all surgeries compared with an average of 2.8 percent to 3.0 percent in other countries for each of these A-DRGs. Another example is Other Transurethral Procedures, which represent less than 1 percent of all surgeries in Moldova and Romania as opposed to 2.6 percent in Estonia. These discrepancies might reflect natural fluctuation in shares but might also suggest this as an area for attention. Figure 12. Surgical A-DRGs with Similar Prevalence in All Countries (Group 1) Source: Authors’ calculations based on data provided by country teams. Notes: A-DRG = Adjacent diagnosis-related group. Korea is not presented in this figure as Korean country-level data could not be split by DRGs (such split is available only for 1,352,906 cases, grouped by seven A-DRGs, out of total 16,46,348 cases in 2021). Group 2 combines A-DRGs that are more important in Ukraine and Moldova than in the other three countries (Figure 13). The high shares of some A-DRGs do not necessarily reflect a problem in these particular A-DRGs but could instead suggest insufficient total numbers of surgical cases. 5 However, such A-DRGs are candidates for further investigation and, presumably, for efforts to promote convergence to the shares in other countries. Abortions are prevalent both in Ukraine and Moldova, which might reflect insufficient preventive measures and limited sex education in these countries. Hernia is a common surgery in all countries, but in Ukraine it accounts for 4.5 percent of all surgeries, which is a much higher 5 For example, in a specific country the total number of cases might be lower because of an insufficient number of hip replacements, but the number of hernia cases is optimal (same per capita as in other countries). Then, the share of hernia cases in total cases in this country would be higher than it would be in a normal situation. 32 share than in Romania (2 percent), for example. Breast, anal, and stomal surgeries are extra prevalent both in Ukraine and Moldova, while nonmalignant hysterectomy and female reproductive organ surgeries are suspiciously widespread only in Ukraine. Likewise, glaucoma cases are remarkably prevalent in Moldova—accounting for 4.2 percent of all surgeries—as are lens corrections in Ukraine. The high shares of hysteroscopy and gastroscopy in Ukraine are a special case, as Ukraine introduced separate respective benefit packages to prevent oncological diseases. Ukraine could analyze the results of these measures to clarify the effectiveness of such a policy and decide if it needs adjustment. Figure 13. Surgical A-DRGs More Prevalent in Ukraine and/or Moldova (Group 2) Source: Authors’ calculations based on data provided by country teams. Notes: A-DRG = Adjacent diagnosis-related group. Korea is not presented in this figure as Korean country-level data could not be split by DRGs (such split is available only for 1,352,906 cases, grouped by seven A-DRGs, out of total 16,46,348 cases in 2021). Group 3 includes DRGs less widely provided in Ukraine or Moldova (Figure 14). Small and large bowel surgeries seem to be underprovided in both countries. Ukraine’s major gap is in one of the top-six MDCs, Musculoskeletal System and Connective Tissue. It includes hip replacement, as well as procedures on elbow, forearm, knee, back, femur, etc.; presumably the needed implants are not included in the benefit package in Ukraine. However, Moldova also shows insufficient provision of such surgeries if compared with Estonia and Croatia. Romania does too, but only in some of these A-DRGs (Other Back and Neck Procedures, Other Elbow and Forearm Procedures). Another top-six MDC group (Circulatory System) is underprovided 33 not only in Ukraine and Moldova but also in Romania. The Interventional Coronary Procedures (ICP) prevail in Croatia and Estonia but have low frequency in Romania. Ukraine underprovides in ICP Not Admitted for AMI, while Moldova data suggest underprovision in ICP, Admitted for AMI. Trans-vascular Percutaneous Cardiac Interventions play a major role in Croatia and Estonia but are almost absent in Ukraine, Romania, and Moldova. Figure 14. Surgical A-DRGs Less Widely Provided in Ukraine and/or Moldova (Group 3) Source: Authors’ calculations based on data provided by country teams. Notes: A-DRG = Adjacent diagnosis-related group; AMI = Acute myocardial infarction. Korea is not presented in this figure as Korean country-level data could not be split by DRGs (such split is available only for 1,352,906 cases, grouped by seven A-DRGs, out of total 16,46,348 cases in 2021). Group 4 includes outlier A-DRGs noticed in select countries. Such cases include Major Procedures for Breast Disorders for Croatia; Vascular Procedures, without Major Reconstruction, without cardiopulmonary bypass (CPB) Pump for Romania; and Other Ear, Nose, Mouth, and Throat Procedures for Estonia (Figure 15). Korea was not included in the DRG comparison as it provided DRG split for only seven disease groups (8 percent of all cases). However, lens procedures, which is one of these seven groups, showed remarkably high frequency in Korea, accounting for 19 percent of all surgical cases. 34 Figure 15. Surgical A-DRGs Widely Provided in Countries Other Than Ukraine and/or Moldova (Group 4) Source: Authors’ calculations based on data provided by country teams. Notes: A-DRGs = Adjacent diagnosis-related groups; w/o = Without; CPB = Cardiopulmonary bypass. Korea is not presented in this figure as Korean country-level data could not be split by DRGs (such split is available only for 1,352,906 cases, grouped by seven A-DRGs, out of total 16,46,348 cases in 2021). 3.3.6 Hospital length of stay for top A-DRGs Countries demonstrated major variation for the average length of stay (ALOS) for the most frequent A-DRGs. Figure 16 and Figure 17 present the results. The ALOS for diseases in Ukraine is generally much longer than for similar diseases in other countries, especially for Coronary Atherosclerosis, Antenatal and Other Obstetric Admissions, Otitis Media and Respiratory Infections, Cranial and Peripheral Nerve Disorders, Bone Diseases and Arthropathies, Nonsurgical Spinal Disorders, and Diabetes. For some of the A-DRGs, the ALOS in other countries is similar. This is true, for example, of Respiratory Infections, for which Estonia recorded the longest ALOS; 6 Stroke, for which Estonia and Ukraine have a similar ALOS; Disorders of Pancreas, for which Ukraine and Croatia have a similar ALOS; and for Deliveries, for which Ukraine and Romania have a similar ALOS. 6 The ALOS for Respiratory Infections is exceptionally long in Estonia, which is likely explained by the inclusion of COVID-19 cases in this A-DRG in Estonia. 35 Figure 16. Average Length of Stay for Top Medical A-DRGs Source: Authors’ calculations based on data provided by country teams. Notes: A-DRGs = Adjacent diagnosis-related groups; ALOS = Average length of stay. Korea is not presented in this figure as Korean country-level data could not be split by DRGs (such split is available only for 1,352,906 cases, grouped by seven A-DRGs, out of total 16,46,348 cases in 2021). For most A-DRGs, the second-highest ALOS was in Moldova. Moldova’s extreme case was the duration of hospital treatment for Alcoholic Intoxication (18 days), three times longer than in Ukraine or Romania (about six days). The disparity in the treatment of schizophrenia in Ukraine, Moldova, and other countries is striking. In Ukraine and Moldova, the ALOS for patients with schizophrenia is more than twice what it is in Estonia or Romania, where it is approximately 15 days, and triple the ALOS in Croatia, where it is only 11 days. This irregularity highlights the need for improvements in mental health care in Ukraine and Moldova, as prolonged hospitalization can have negative effects on patients and increase health care costs. 36 Figure 17. Average Length of Stay for Top Surgical A-DRGs Source: Authors’ calculations based on data provided by country teams. Notes: ALOS = Average length of stay; A-DRGs = Adjacent diagnosis-related group; AMI = Acute myocardial infarction. Korea’s data are included in this figure. A-DRGs should be split into DRGs based on complexity and differential resource use. This split should be reflected in higher ALOS for more complex cases. Results (Figure 18 and more examples in Annex 6) indicate that in some countries—specifically in Moldova and Romania for some A-DRGs—the DRG splitting may need adjustments, or upcoding may be occurring. For example, in Croatia, ALOS for hip replacements differed significantly depending on the severity of the case: it was more than 25 days if there were severe complications and comorbidities, and almost 13 days if no complications or comorbidities were present. The ALOS for hip replacement varied to a lesser extent in Estonia, where it was 7.8 days versus 5.9 days, and in Romania, where it was 8.8 days versus 8.6 days. In contrast, ALOS for hip replacement did not depend on case severity in Moldova (8.6 days irrespective of severity). 37 Figure 18. Difference in ALOS across Complexity Levels within Selected A-DRGs Source: Authors’ calculations based on data provided by country teams. Notes: ALOS = Average length of stay; A-DRGs = Adjacent diagnosis-related group; wo = Without. Ukraine is not presented in this figure as its DRG system is introduced only at A-DRG level. Korea is not presented because the structure of the Korean DRG system differs from the rest of the four countries and cannot be comparable in this figure. For instance, there are two level of complexity for Appendectomy in Moldova, Romania, Croatia, and Estonia, while in Korea there are four DRGs related to Appendectomy (“Appendectomy with Complicated Principal Diagnosis,” “Appendectomy without Complicated Principal Diagnosis,” “Laparoscopic Appendectomy with Complicated Principal Diagnosis,” “Laparoscopic Appendectomy without Complicated Principal Diagnosis”), which are further split by three levels of complexity. 3.5 Analysis by main diagnosis Another layer of analysis was conducted to match hospital cases by main diagnosis; this effort found viral pneumonia to be the most common main diagnosis across hospitals. As all of the study countries use International Classification of Diseases—10th Revision (ICD- 10), with some national modifications, the comparison and analysis of diagnosis information were relatively straightforward. In this analysis, the three-character-level ICD-10 codes were used, focusing on the top-10 most frequently used main diagnosis codes of each country within the top-six MDCs. As a first step, the three most commonly used diagnosis codes (as submitted in the original data set by the countries) were identified in each country. The results (Table 7) show that the diagnosis code J12.8 (Other Viral Pneumonia) is the most common in all countries with the exception of Estonia. Such a high proportion of cases with a main diagnosis of J12.8 is likely due to the impact of COVID-19, which is often the underlying cause of viral pneumonia. The proportion of cases with J12.8 as the main diagnosis varies among countries, ranging from 3.1 percent to 10.0 percent of all inpatient cases in 2021. The proportion of J12.8 is highest in Moldova (10.0 percent of all cases) and Ukraine (8.4 percent of all cases). In addition, Ukraine’s specific diagnosis code U07.1 for COVID-19 was among the most frequently used (2.7 percent of all hospital cases admitted had this main diagnosis). The 38 combined total of J12.8 and U07.1 is 11.1 percent of hospital cases in Ukraine, which likely reflects the high burden of the COVID-19 pandemic on hospital performance and capacity. The remaining diagnosis codes in all countries were related to spontaneous delivery and, surprisingly, to chemotherapy for neoplasms in Estonia. The presence of Z51.1 in Estonia indicates that despite the pandemic, the treatment of cancer patients in need of chemotherapy was likely prioritized. Table 7. Top-Three Main Diagnosis Codes in Study Countries Country Diagnosis Diagnosis code text Number of cases Cases as % of code total Ukraine J12.8 Other viral pneumonia 443,079 8.4 O80 Single spontaneous delivery 176,496 3.3 U07.1 COVID-19, virus identified 141,123 2.7 Moldova J128 Other viral pneumonia 54,824 10.0 O800 Spontaneous vertex delivery 13,771 2.5 Z380 Singleton, born in hospital 13,738 2.5 Romania J12.8 Other viral pneumonia 79,533 3.4 Z38.0 Singleton, born in hospital 45,831 2.0 J18.9 Pneumonia, unspecified 36,023 1.6 Croatia J12.8 Other viral pneumonia 14,487 3.1 O80.0 Spontaneous vertex delivery 14,412 3.1 J18.9 Pneumonia, unspecified 9,710 2.1 Estonia O80.0 Spontaneous vertex delivery 8,514 5.0 J12.8 Other viral pneumonia 8,097 4.7 Z51.1 Chemotherapy session for neoplasm 5,781 3.4 Source: Authors’ calculations based on data provided by country teams. Note: Korea is not presented in this figure as Korean country-level data could not be split by diagnosis-related groups (DRGs) (such split is available only for 1,352,906 cases, grouped by seven adjacent diagnosis-related groups [A-DRGs], out of total 16,46,348 cases in 2021). The study team observed major differences in the prevalence of various medical conditions treated in hospitals of the studied countries. Table 8 shows the percentages of top-10 diagnoses for included countries. On average, these diagnoses were responsible for 14– 17 percent of all hospital cases in Croatia, Estonia, and Romania, and 25–28 percent of hospital cases in Moldova and Ukraine. Among the top-10 diagnoses, MDC Diseases and Disorders of the Respiratory System is represented by three codes (J12, J18, and J20). MDC Pregnancy, Childbirth, and Puerperium comprises two of the top-10 codes (O80 and O82). MDC Diseases and Disorders of the Nervous System is likewise represented by two codes (I63 and I67), as is MDC Diseases and Disorders of the Circulatory System (I20 and I25). Interestingly, none of the top-10 diagnosis codes were related to MDC Diseases and Disorders of the Musculoskeletal System and Connective Tissue or MDC Diseases and Disorders of the Digestive System, which were among the top-six MDCs. This difference can be attributed to the different levels of aggregation for MDC and diagnosis information. Table 8. Top-Ten Main Diagnosis Codes and Share of Cases in Study Countries (Percent) Diagno- sis code Diagnosis text Ukraine Moldova Romania Croatia Estonia Total J12 Viral pneumonia, not elsewhere classified 8.7 10.1 4.8 4.2 4.8 7.4 O80 Single spontaneous delivery 3.3 2.5 0.9 3.3 5.3 2.7 J18 Pneumonia, organism unspecified 2.3 3.8 3.4 2.3 0.7 2.6 I63 Cerebral infarction 2.0 1.3 1.5 1.7 2.7 1.8 I20 Angina pectoris 2.3 1.7 0.6 1.2 0.9 1.7 U07 COVID-19 2.7 0.0 0.0 0.0 0.1 1.6 I25 Chronic ischemic heart disease 2.2 1.1 0.3 1.0 0.9 1.5 39 O82 Single delivery by cesarean section 1.3 1.1 1.4 1.8 0.6 1.3 I67 Other cerebrovascular diseases 1.7 2.0 0.5 0.2 0.1 1.3 J20 Acute bronchitis 1.8 1.8 0.1 0.2 0.6 1.2 % of top-10 28 25 14 16 17 23 Source: Authors’ calculations based on data provided by country teams. Note: Color-coding corresponds to top-six MDCs coloring (see Figure 4). Korea is not presented in this table as we have got only limited information on diagnoses codes. 3.6 Analysis by procedure codes Analysis by procedure codes was most challenging. Due to the differences in procedure classifications across the countries, there was no feasible and consistent way to aggregate single national procedure codes; this was in contrast to the case of diagnosis codes, which could be aggregated by using the three-character level codes. The focus was on codes indicating significant surgical activity. Procedure codes that were aggregated during the data processing were related to the repair of hernia and C-sections. Table 9 shows an example of different C-section procedure codes in the different countries. In all countries that have adopted the AR-DRG system and Australian Classification of Health Interventions (ACHI), the procedure codes for C-section are divided into elective and emergency procedures. But there is no such division in procedure classification in Estonia. And although the C-section procedure coding all countries considers the area of the uterus where the incision is made (lower or upper segment), there are major differences between the countries that are explained by the primary classifications they use. Table 9. Example of C-Section Procedure Codes in Study Countries Country Code (original) Code text Ukraine 16520-00 Elective classical cesarean section 16520-01 Emergency classical cesarean section 16520-02 Elective lower-segment cesarean section 16520-03 Emergency lower-segment cesarean section 16520-04 Elective cesarean section, not elsewhere classified 16520-05 Emergency cesarean section, not elsewhere classified Moldova 1652000 Elective classical cesarean section 1652001 Emergency classic C-section 1652002 Elective cesarean section of the lower segment 1652003 Emergency cesarean section of the lower segment Romania N01101 Elective classical cesarean section N01102 Emergency classical cesarean section N01103 Elective lower-segment cesarean section N01104 Emergency lower-segment cesarean section Croatia 16520-00 Elective classical cesarean section 16520-01 Emergency classic C-section 16520-02 Elective cesarean section of the lower segment 16520-03 Emergency cesarean section of the lower segment Estonia MCA00 Upper uterine segment cesarean section MCA10 Lower uterine segment cesarean section MCA96 Other cesarean section Source: Authors’ calculations based on data provided by country teams. Note: Korea’s data are not included as only limited data for cases assigned to seven adjacent diagnosis-related groups (A- DRGs) were available, which did not allow comparison with other countries. There are similarities across countries on the use of the three most common surgical procedure codes. As was the case for the diagnosis codes, there was one procedure code common for all countries: C-sections. Another procedure, Repair of hernia, was present in all countries except for Estonia. The most frequent procedures were common for Croatia, Romania, Moldova, and Ukraine. However, Laparoscopic cholecystectomy was among the 40 top-three procedure codes only in Croatia, and percutaneous transluminal coronary angioplasty (PTCA) with insertion of the stent was among the top-three procedures only in Estonia. Results are shown in Table 10. Table 10. Top-Three Surgical Procedure Codes in Study Countries Country Code text Number of Cases as % cases of total Ukraine C-section 68 578 1.3 Dilation & curettage of uterus 59 740 1.1 Repair of hernia 53 732 1.0 Moldova C-section 6 421 1.2 Dilation & curettage of uterus 6 082 1.1 Repair of hernia 5 147 0.9 Romania C-section 81 522 3.5 Repair of hernia 30 822 1.3 Laparoscopic cholecystectomy 24 400 1.1 Croatia C-section 10 653 2.3 Repair of hernia 7 161 1.5 Laparoscopic cholecystectomy 4 643 1.0 Estonia Repair of partial rupture of perineum 4 706 2.7 Percutaneous transluminal coronary angioplasty with insertion of stent 2 708 1.6 C-section 2 628 1.5 Source: Authors’ calculations based on data provided by country teams. Note: Korea’s data are not included as only limited data for cases assigned to seven adjacent diagnosis-related groups (A- DRGs) were available, which did not allow comparison with other countries. Four procedure codes (two of them on an aggregated level) are among the top 10 of all that are present in each country. These codes are related to MDC Pregnancy, Childbirth, and Puerperium and MDC Diseases and Disorders of the Digestive System (C-section, repair of hernia, laparoscopic cholecystectomy, and appendectomy). Table 11 presents the frequency of utilizing the most common procedures. Romania has the highest percentage of patients undergoing C-section, while Moldova has the highest percentage of patients undergoing dilation and curettage of uterus. Ukraine has the highest percentage of patients undergoing laparoscopic cholecystectomy and insertion of intraocular lens. The share of top-10 surgical procedures in all surgical procedures is highest in Ukraine and Romania, and lowest in Estonia. Such differences do not necessarily mean that the procedure codes do not exist in the national versions or that the procedures are not performed. It is possible that the procedures are not performed in inpatient settings or that different countries have different coding guidelines. Table 11. Top-Ten Surgical Procedure Codes as Share of Cases in Study Countries (Percent) Code text Ukraine Moldova Romania Croatia Estonia Total C-section 1.3 1.2 3.5 2.3 1.5 1.9 Repair of hernia 1.0 0.9 1.3 1.5 1.0 1.1 Dilation & curettage of uterus 1.1 1.1 0.4 0.4 0.9 Laparoscopic cholecystectomy 0.6 0.6 1.1 1.0 1.1 0.7 Appendectomy 0.6 0.6 0.4 0.4 0.2 0.5 Phacoemulsification of crystalline lens 0.7 0.4 Insertion of intraocular lens 0.6 0.003 0.3 Polypectomy of uterus via hysteroscopy 0.4 0.041 0.3 Suction curettage of uterus 0.4 0.3 Incision and drainage of abscess of soft tissue 0.4 0.1 0.3 Percentage of top-10 7.1 4.4 6.9 5.6 3.8 6.8 Source: Authors’ calculations based on data provided by country teams. Note: Korea’s data are not included as only limited data for cases assigned to seven adjacent diagnosis-related groups (A- DRGs) were available, which did not allow comparison with other countries. 41 3.7 Use of country-level findings to define the scope of the hospital-level analysis The study used country-level data analysis to define the scope of the hospital-level data analysis. The higher-level analysis helped identify the case types that shaped further hospital- level data collection and analysis. The aim was to include cases minimally impacted by the COVID-19 pandemic and of an acute nature. In the hospital-level analysis, four case types were determined to cover both medical and surgical activities and included in the study. Cases were defined using country-level analysis and comparison of MDCs, DRGs/A-DRGs, major diagnosis and procedure codes, discussion with the experts from participating countries, and EURODRG Project vignette definitions. The study identified two case types in the medical category and two in the surgical category for further in-depth analysis, as described below. The detailed definitions of each case type can be found in Annex 7. 1. Vaginal delivery (medical). Various DRGs related to vaginal deliveries were among the top-six MDCs (MDC 14 Pregnancy, Childbirth, and Puerperium) in all countries except Romania. However, the total number of cases in three Romanian DRGs related to vaginal deliveries 7 is 26,296 cases, making aggregated cases of vaginal delivery one of the top medical DRGs in Romania as well. The share of DRGs related to vaginal delivery in total cases is 6 percent in Estonia, 5 percent in Croatia, and 3 percent in Moldova and Ukraine. Single spontaneous delivery was also one of the most commonly used diagnosis codes in all countries, a finding that supports inclusion of this case type in hospital-level analysis. 2. AMI (medical). AMI was chosen as a case type for hospital-level analysis, even though no separate DRG for AMI was among the top-20 DRGs in any country or among the top- 10 diagnosis codes. However, when the COVID-19-related diagnosis codes were excluded, the AMI diagnosis turned out to be among the most frequent in each country. AMI was included as a case type to represent MDC 5 Diseases and Disorders of the Circulatory System, as this MDC is one of the most prevalent among the countries, and AMI cases could be relatively easily defined based on diagnosis information. Depending on the country, AMI cases without surgical intervention are often grouped into DRGs related to other circulatory disorders with or without AMI, which were present among the top-20 DRGs, for example in Romania. Because data were not available, AMI cases were not analyzed for Korea. 3. C-section (surgical). C-section is among the top surgical DRGs in all countries. The same is true for different procedure codes referring to C-section, which were also widely used in all countries. In the analysis of procedure codes, different C-section codes were aggregated. The selection of C-section was also supported by the use of the diagnosis code Single delivery by cesarean section, which appeared among the top-10 diagnosis codes. 4. Appendectomy (surgical). Appendectomies are among the top surgical DRGs in most countries. Appendectomy was also among the top-10 most commonly used procedure codes. Even though there is a separate procedure code for laparoscopic appendectomy, which was not among the top-10 procedure codes, appendectomies are still one of the most 7 These DRGs include (1) vaginal delivery with multiple complicating diagnoses, at least one of which is severe, (2) vaginal delivery with severe complicating diagnosis, and (3) vaginal delivery with moderate complicating diagnosis. 42 frequently performed procedures in the participating countries. Additionally, appendectomy is a relatively unified intervention with a straightforward definition, making it a good candidate for inclusion in the hospital-level data analysis. 43 4 Findings at the Hospital Level The analysis of findings at the hospital level specifically reviewed the following areas: 1. Coding activity by different hospitals, such as number of secondary diagnoses and procedures reported per case 2. Performance indicators for the hospitals, such as relation between ALOS and coding activity 3. Additional cases of interest that highlight differences between the hospitals 4.1 Coding activity The study reviewed coding activity across countries and case types. It looked at whether coding activity is similar in select countries across comparable facilities and across types of cases. It also compared countries to understand whether coding activity depends on the DRG system or duration of use of DRG grouping. To compare the coding activity across hospitals, the study looked at the average number of secondary diagnosis and procedure codes reported per selected cases. A general overview of coding frequencies by countries per case types is presented in Figure 19. There is more coding of secondary diagnoses in Croatia and Romania and relatively less in Estonia, Korea, and Ukraine. In Romania and Croatia, the high level of coding activity is largely driven by the case-mix system used in the country, which could explain the difference between these two countries and the others. For procedure codes, the highest level of coding activity is observed for Croatia (except in the case of medical AMI), while in Ukraine the most frequent use of procedure codes was for AMI cases. Figure 19. Average Number of Secondary Diagnosis Codes and Procedure Codes per Case Source: Authors’ calculations based on data provided by country teams. Note: AMI = Acute myocardial infarction. Consultations with country teams suggested the following possible explanations for the variations in the coding activity. In Romania, only payment-related procedures were included in the data due to digitalization and database size limits. In reality, the level of procedure coding in hospitals in Romania is higher than what was shown in the current study. In Estonia and Ukraine, hospital payments do not depend on the number of procedure codes. In Estonia, only those procedure codes that influence the grouping into DRGs are included in the database. 44 The high level of coding activity in Croatia may be explained by a set of specific reasons. First, payments per hospital cases relate to the reported severity of cases, which depends, among other things, on the coding of secondary diagnoses and additional procedures. Second, Croatia also uses coding as a tool for recording all the work performed by doctors, interns, and nurses. Third, the registration of complex cases in the hospital information system is often done using an automatic bundle of DRGs, procedures, and used materials that requires manual removal of unprovided components. If unprovided components are unintentionally left in place, a higher level of coding may result. Coding of secondary diagnosis for appendectomy cases is presented in Figure 20 (left side). The data suggest that the coding of secondary diagnoses in Croatia’s hospitals and a university hospital in Romania follow similar patterns. Studied hospitals in Estonia and Moldova provide less intense data on secondary diagnoses. In Ukraine, the coding of both secondary diagnoses and procedures seems to be nominal, with an average of one entry per category for diagnoses and procedures. Coding of secondary diagnoses for appendectomy cases is systematically least frequent in Estonia’s hospitals. Figure 20. Appendectomy Cases: Coding Activity Source: Authors’ calculations based on data provided by country teams. The right side of Figure 20 shows coding of procedures for appendectomy cases. University hospitals in Croatia and Romania report higher frequencies of procedures administered to patients. There is a twofold difference between the university and multiprofile hospitals in coding of procedures. This might be explained by the fact that more complicated cases are treated in university hospitals, though the Korean university hospital shows lower levels of coding activity than its nonuniversity counterpart. The data also do not support the hypothesis that hospitals of the same type show similar coding activity: one of the multiprofile hospitals in Moldova shows three times higher activity in coding procedures compared to its peers within the country, but does not differ from its peers in its coding of secondary diagnoses. Coding activity for C-sections shows variation, as presented in Figure 21. Croatian hospitals, regardless of size, are among the leaders for coding of both secondary diagnoses and procedures. However, the university hospital in Croatia shows even more intensive coding, with more than nine secondary diagnoses and 16 procedures per case on average. The number of secondary diagnosis and procedure codes in Estonia is moderate, though the university hospital has fewer procedures than the multiprofile hospital. Similarly, in Korea, the overall 45 coding activity is moderate, and the university hospital is more modest than the nonuniversity hospital in the number of coded diagnoses. In Moldova, among the three hospitals of the same type, coding activity differs, which might indicate some issues with consistent case reporting. Romanian hospitals, regardless of type, show relatively high activity in the coding of secondary diagnoses and relatively low activity in procedure coding. In Ukraine, the coding activity is moderate but varies from hospital to hospital; the number of procedure codes ranges from almost 13 to 2 codes per case. This variation may be explained by the different coding routines and training across hospitals. Figure 21. C-Section Cases: Coding Activity Source: Authors’ calculations based on data provided by country teams. Coding activity for vaginal delivery is shown in Figure 22. As was true for the previous case types, Croatia and Romania show the most intense coding activity—the highest level is that of the Croatian university hospital—and Estonia shows the lowest level of coding activity. Estonian and Korean coding activity does not support the hypothesis of higher levels of coding activity in university hospitals. In Moldova and Ukraine, the activity is moderate, but the difference across hospitals of the same type within each country varies significantly. Such variability of coding is surprising for Ukraine, which applies a flat rate for all deliveries so that there are no financial incentives to increase or decrease coding. Figure 22. Vaginal Delivery: Coding Activity Source: Authors’ calculations based on data provided by country teams. 46 Coding activity for AMI is shown in Figure 23. In three different countries, the largest number of secondary diagnoses is coded by university hospitals, which likely indicates that more complicated cases are treated in university hospitals. However, high levels of secondary diagnosis coding are also recorded in a multiprofile hospital in Croatia and in a multiprofile hospital in Romania. In Ukraine, two multiprofile hospitals have both the lowest number of secondary diagnoses and the highest number of procedures, which may be due to gaps in coding capacity or the absence of unified guidelines for AMI treatment. Figure 23. AMI: Coding Activity Source: Authors’ calculations based on data provided by country teams. Note: Korea is not presented as its acute myocardial infarction (AMI) data are not available (see explanation in Section 2.4). The analysis shows that coding activity differs across countries and disease types and is influenced by various factors. University hospitals in general have somewhat higher levels of coding activity than other hospitals, particularly in Croatia, Estonia, and Romania. However, in Korea, the pattern is the opposite. For other countries, the hospital type does not explain the variation in coding activity: hospitals of the same type in Ukraine and Moldova show different coding activities for procedures and secondary diagnoses. The coding activity may be influenced by the DRG system and its complexity and by use of DRGs for case-based payments. Estonia, which uses the Nordic DRG system, has the lowest level of coding activity. Ukraine does not yet use the full DRG classification, which likely means there are no incentives for coding cases, since case severity does not influence tariffs. In Croatia, hospital remuneration depends on the severity of cases, which explains the higher numbers of reported secondary diagnoses and procedures. The coding intensity may also depend on the duration of the use of DRG-grouped data. In Ukraine, where the DRG system was introduced in recent years, similar types of hospitals, such as maternity hospitals, exhibit significantly different coding activity. This is likely related to differences in training, lack of feedback provided on coding, and overall immaturity of the system. 4.2 Performance indicators This section explores the link between the reported complexity of cases and duration of hospital stay. Specifically, it examines whether coding activity correlates with the duration of hospital treatment and whether ALOS is consistently higher in more specialized or university 47 hospitals. Figure 24, Figure 25, Figure 26, and Figure 27 present the relationship between the ALOS and the number of secondary diagnosis codes per case for each case type. The bigger the gap between the data points, the less these two indicators are related to each other. Figure 24. Appendectomy Performance Indicators Source: Authors’ calculations based on data provided by country teams. Note: ALOS = Average length of stay. For appendectomy cases, a correlation between the ALOS and coding activity can be observed in Croatia, Estonia, and Romania (Figure 24). The pattern in Korea is the opposite, with relatively low levels of coding in both hospitals and a relatively long length of stay. In Moldova and Ukraine, the situation is ambiguous: select hospitals in both countries show the same consistent patterns as hospitals in Croatia, Estonia, and Romania. Yet the majority of the Ukrainian and Moldovan hospitals present relatively low numbers of secondary diagnoses, which do not correspond with the hospitals’ high ALOS. This discrepancy may be explained by the fact that the ALOS in Moldova and Ukraine has historically been higher than in other countries. 48 Figure 25. C-Section Performance Indicators Source: Authors’ calculations based on data provided by country teams. Note: ALOS = Average length of stay. For C-section hospital cases, the coding activity corresponds to the ALOS in almost all hospitals; the exception is select Ukrainian hospitals (Figure 25). As expected, university hospitals show a higher number of secondary diagnoses and longer ALOS. However, hospitals of the same type within a country do not necessarily show the same level of hospital activity or ALOS. This specifically applies to Moldova and Ukraine. Figure 26. Vaginal Delivery Performance Indicators Source: Authors’ calculations based on data provided by country teams. Note: ALOS = Average length of stay. 49 For vaginal delivery, higher levels of coding activity in hospitals of Romania and Croatia did not result in higher ALOS (Figure 26). Hospitals in Estonia and some hospitals in Moldova had both the lowest ALOS and the lowest number of secondary diagnoses. University hospitals in Croatia and Romania tended to have slightly higher numbers of secondary diagnoses but the same ALOS as other hospitals in the same country. Ukrainian hospitals presented diverse results, with the ALOS not necessarily corresponding to the number of secondary diagnoses. Maternity hospital 5 in Ukraine is an outlier, with almost the lowest level of coding activity (an average of 1.3 secondary diagnoses) and the longest ALOS (seven days). Figure 27. AMI Performance Indicators Source: Authors’ calculations based on data provided by country teams. Notes: AMI = Acute myocardial infarction; ALOS = Average length of stay. Korea is not presented as its AMI data are not available (see explanation in Section 2.4). The results of the data analysis for medical AMI are inconsistent. Data on AMI cases are presented in Figure 27. On average, Croatian and Romanian hospitals had the highest level of coding activity, but this did not translate into increased ALOS. Hospitals in Moldova presented shorter ALOS coupled with less intense coding activity compared to Croatian and Romanian hospitals. Ukrainian hospitals generally had longer ALOS and lower numbers of secondary diagnoses compared to hospitals in other countries. The case of Estonia was particularly surprising: ALOS in its multiprofile hospital was one of the longest across all hospitals in the sample (11.8 days) while also reporting almost the lowest complexity of cases. One notable finding in the analysis was the lack of a consistent relation between the ALOS and coding activity. This was particularly evident in Ukrainian hospitals for all types of cases; in Moldovan hospitals for appendectomy; and in Romanian hospitals for C-section, vaginal delivery, and AMI. This result may be due to issues with coding quality, but there could be other contributing factors. 50 A range of factors may have influenced the variation in ALOS among hospitals. For example, in Estonia, clinical protocols for treating AMI patients are used nationwide to reduce clinical variations, including ALOS. The maturity of the case-mix system in a country may also affect ALOS, which could explain why Romania and Estonia show a different trend than Ukraine and Moldova. It is known that the implementation of DRGs can initially have a significant impact on ALOS. However, this impact may decrease over time as the potential for further reduction diminishes. In Croatia, contracting arrangements may also play a role in ALOS. These and other country-specific factors may have contributed to differences in ALOS and cross-hospital variations. 4.3 Specific cases of interest Within the data analysis, several types of cases presented particular interest and were analyzed in greater detail. Two are described below: ALOS in Ukrainian maternity hospitals and use of different DRGs for the aggregated case types in different countries. 4.3.1 ALOS for vaginal delivery in Ukrainian maternity hospitals The study observed homogeneity in coding and ALOS in five out of six maternity hospitals in Ukraine. One of the hospitals had a significantly higher ALOS of seven days, but at the same time had one of the lowest levels of coding activity, with an average of 1.5 secondary diagnosis codes per case (Figure 28). To identify potential explanations for the exceptional ALOS at maternity hospital 5, the data for this hospital were analyzed in greater detail. Figure 28. Coding Activity and ALOS for Vaginal Delivery in Ukrainian Maternity Hospitals Source: Authors’ calculations based on data provided by country teams. Note: ALOS = Average length of stay. For a more in-depth review, a histogram was created based on the length of hospital stay for all cases at this particular hospital (see Figure 29). The data show that 67 percent of 2,229 cases had an ALOS below 4.2 days, which was similar to the ALOS at other hospitals. The remaining cases with higher ALOS may have different reasons. This analysis could help to identify such extremes and lead to more detailed examination to determine the causes of these differences. 51 Figure 29. Histogram of ALOS in Ukraine Maternity Hospital 5 (Total of 2,229 Cases) Source: Authors’ calculations based on data provided by country teams. Notes: ALOS = Average length of stay; LOS = Length of stay. . 4.3.2 Variety of different DRGs within case type Although this study grouped cases under more aggregated case types, some cases were analyzed to compare the use of more specific DRGs across hospitals. The greatest variability was found among appendectomy cases. In this group, Ukraine’s aggregation of cases was the least granular, with some hospitals grouping all appendectomy cases into a single DRG. By contrast, in Croatia appendectomy cases are grouped with a much more precise calibration; appendectomy cases may fall into eight DRGs in multiprofile general hospitals and 12 DRGs in a multiprofile university hospital. There were also notable differences among hospitals in Romania (six to eight different DRGs) and in Estonia (four different DRGs), as presented in Figure 30. Figure 30. Number of Different DRGs Used to Group Appendectomy Cases, by Hospital Source: Authors’ calculations based on data provided by country teams. Notes: DRGs = Diagnosis-related groups. Korea is not presented in this figure as we do not have raw Korean hospital-level data to calculate the number of DRGs. The observed differences in assignments to DRGs may be related to the various case-mix systems used by the participating countries. Other factors, such as incomplete data or variations in coding practices among countries, may also explain the observed differences. 52 Table 12 provides a comparison of different DRGs used to assign appendectomy cases in the Croatian multiprofile university hospital and Ukrainian multiprofile hospital 5. Understanding the reasons for the assignment of appendectomy cases to a relatively large number of DRGs may help organize them into more clinically homogenous groups. Table 12. Comparison of DRGs Used for Grouping Appendectomy Cases in Croatia and Ukraine Hospital type DRG DRG name Number Number Number of of cases of procedure secondary codes per diagnosis case codes per case Multiprofile 5 G07 Appendectomy 474 1.0 1.0 (Ukraine) Multiprofile G07B Appendectomy without very severe or severe CC 144 3.5 10.1 university G04C Peritoneum adhesion, age < 50 without CC 42 3.2 11.2 (Croatia) G70B Other disorders of the digestive system without CC 14 4.1 2.5 G07A Appendectomy with very severe or severe CC 11 5.8 11.1 G04B Peritoneum adhesion, age > 49, or CC 8 4.6 13.0 G02B Great procedures on the small and large intestine without 4 3.8 15.0 very heavy CC G12B Other operating procedures on the digestive system 1 3.0 4.0 without very severe or severe CC A06Z Tracheostomy or ventilation > 95 hours 1 20.0 18.0 G70A Other disorders of the digestive system with CC 1 6.0 2.0 G44B Other colonoscopy procedures without very difficult or 1 5.0 3.0 severe CC G04A Peritoneum adhesion, age > 49 with CC 1 11.0 17.0 A41B Intubation, age < 16 without CC 1 4.0 11.0 Source: Authors’ calculations based on data provided by country teams. Notes: DRGs = Diagnosis-related groups; CC = Complications and comorbidities. 53 5 Discussion of Study Limitations The reliability and quality of data are crucial for cross-country comparisons, but collecting data from multiple countries can present various challenges. In this study, the main challenges were related to data protection issues, language barriers, the time required for data extraction, and the availability of data based on the template in each country. For example, some countries had strict regulations on the collection and sharing of patient-level information, which required additional clearance procedures. Language barriers also posed a challenge in some cases, as some countries’ data were available in the original language and required additional translation efforts. Included countries use different scopes and classifications of hospital cases. The DRGs used for classifying patients in Korea and Estonia are different from those used by the other four countries (all of which use some variant of AR-DRG). Moldova includes rehabilitation in results. The AR-DRG includes some A-DRGs named “same-day” cases—inpatient cases with the admission and discharge on the same date. These A-DRGs exist in Ukraine, Romania, Croatia, and Moldova and include mental health treatment, lens procedures, endoscopy, and others. Estonia also includes day procedure cases in the results. Ukraine groups patients up to only A-DRG level, which does not distinguish levels of complexity among cases of the same diagnosis group. The study team reports countries’ willingness to join the study and their adherence to the proposed data collection procedures. Most of the participating countries were able to provide the necessary data within the given time frame, and the country representatives were responsive to any questions that arose during the analysis. The use of templates streamlined the collection and processing of country- and hospital-level data. However, despite these efforts, some limitations remain. One limitation of the study was that it did not allow for the evaluation of clinical coding quality, as access to the original data sources (i.e., medical records) was not within the scope of the study. For this reason, the accuracy of the data used in the analysis could not be fully assessed. This inability to consult original data sources could have affected the reliability of the study’s findings, given that differences in coding systems can impact the accuracy of DRG assignment and potentially lead to differences in resource use or outcomes. Another limitation of the study was related to differences in primary classifications, particularly for surgical procedures. Even countries that initially adopted the same DRG system (AR-DRG) and were supposed to use the same primary classifications later introduced national modifications or changed the primary classifications, making it difficult to map them. This limitation was also one of the reasons why the original plan to include all AMI cases without bypasses in the analysis was modified. To address this limitation in future studies, it may be necessary to prioritize the development of a standardized mapping between primary classifications in different countries. Finally, the study did not allow for the identification of exhaustive explanations or reasons for differences between hospitals. However, the findings did highlight potential issues at the individual-hospital level that could be of interest to participating countries for their communications with hospitals or purchaser organizations. Future studies could consider advocating for greater access to original data sources to improve the accuracy of clinical coding and better understand the reasons for differences between hospitals. This could help to identify opportunities to improve documentation or coding quality, adjust patient management, or increase efficiency in different processes. 54 6 Summary and Conclusions The cross-country and cross-hospital benchmarking conducted in this study helped to identify positive and negative performance deviations for hospital cases organized by diagnoses and procedures that account for a high share of total hospital outputs. This analysis provided valuable insights for identifying priorities for further investigation and review of reported hospital cases. By pinpointing areas of concern, health care systems can focus their efforts on addressing these issues and improving the overall quality of care. In addition to identifying performance deviations, the study also utilized DRGs to describe the structure of case-mix in each country. This information is particularly useful for health care systems in planning their service delivery capacity. In health care systems that recognize which specialties and case types are underserved or overprovided in inpatient care, it is possible to make informed decisions about resource allocation and service provision. This information can help to ensure that patients receive appropriate care and that health care resources are used efficiently and effectively. Overall, the study’s findings offer valuable information for improving health care systems and ultimately enhancing patient outcomes. An international comparison allows for the benchmarking and comparison of countries or individual hospitals, as well as the investigation of factors contributing to variations in performance. In this study, a comparative analysis was conducted across selected countries and hospitals. The results indicated that less aggregated and more detailed data were analyzed, the more differences within and between countries emerged. The admission rates within top MDCs show significant variations across countries, with some notable patterns. While circulatory diseases have consistent admission rates, other MDCs exhibit inconsistencies. Korea has higher admission rates for digestive, musculoskeletal, and nervous system disorders, while Moldova has a higher admission rate for respiratory diseases, possibly influenced by COVID cases. Surprisingly, Croatia and Korea, both having a higher proportion of older population, show lower admission rates for nervous system disorders. Estonia demonstrates a higher admission rate for musculoskeletal disorders, possibly because of better access to orthopedic surgeries. The study found that the rate of surgical admissions was relatively uniform across four countries, but varied significantly in Korea and Ukraine. Notably, the rate of surgical admissions was found to be very high in Korea and low in Ukraine. On the other hand, the rate of medical admissions was found to be extraordinarily high in both Korea and Moldova. These findings suggest that there may be differences in the patterns of health care utilization across countries, which may reflect differences in the prevalence of specific conditions, variations in clinical practices, or differences in patient preferences. Further investigation and analysis may be needed to better understand the factors contributing to these differences and to develop strategies to address them. The share of inpatient admissions for surgery varies across countries, with Croatia having the highest share at nearly 40 percent, while Ukraine, Moldova, and Korea have relatively lower shares at just over 20 percent. Surgeries tend to be concentrated in a limited number of types within each country. In Ukraine, for instance, each of top-20 surgery A-DRGs is responsible for at least 1 percent of all surgery episodes, and top-32 surgery A-DRGs collectively account for nearly 60 percent of all surgery episodes. Similarly, in Moldova, each of top-21 surgery A-DRGs represents more than 1 percent of all surgery episodes, and the top- 55 31 surgery A-DRGs encompass almost 62 percent of all surgery episodes. The most prevalent surgery A-DRGs in Ukraine include C-section, diagnostic curettage or hysteroscopy, and hernia procedures. In Moldova, the most prevalent surgery A-DRGs are vaginal delivery with operating room procedure, C-section, and eye procedures such as glaucoma and complex cataract procedures or penetrating eye injury. The analysis of discrepancies in average length of stay reveals notable instances where certain A-DRGs have excessively long durations, particularly in Ukraine and Moldova. Specifically, for medical A-DRGs, coronary atherosclerosis exhibits a prolonged average length of stay (Ukraine: over 10 days, Moldova: 8 days), as does hypertension (Ukraine: 10 days, Moldova: 8 days). Similarly, for surgical A-DRGs, major procedures for breast disorders have significantly longer average length of stay in Ukraine (17 days) and Moldova (10 days) compared to other countries. Benchmarking reveals that there is room for improvement by reducing LOS for these case types to levels of the other comparator countries in the study. The data analysis for different case types revealed significant variations in coding activity as well as in ALOS. While some of these differences were clarified with the assistance of country representatives, in many cases the reasons for the variations were not apparent. Wide variation in ALOS for medical and surgical cases suggests substantial room for efficiency improvement by reducing inefficiencies in hospitals within analyzed A-DRGs. Analysis of hospital cases should always relate to the organization of service delivery in countries. It is important to consider the specific context of each country in interpreting the findings, including factors such as contracting arrangements, the use of coding guidelines, the overall health system context, and the implementation of the DRG system. The study's results indicate that the performance of DRG systems is not impacted by how long DRGs have been in use or what type of system is used (Nordic or Australian). However, financial incentives were identified as a factor that influenced the relative frequency of each DRG. In Moldova and Ukraine, similar patterns were observed, indicating a focus on inpatient care and potential underutilization of primary and outpatient care. The study also found that there are only small differences in the average length of stay for A-DRGs of different complexity levels in Moldova and Romania, which suggests that DRG split should be further refined to respect the severity of cases. The study can motivate health care professionals to investigate the reasons behind the identified variation in hospital case-mix. These reasons may vary from demographics and resource differences to clinical practice variations and potential issues like waste or fraud. Armed with these findings, countries may take targeted steps for improvement based on their unique context, ultimately enhancing health care delivery and outcomes. 56 7 Recommendations The study team recommends conducting cross-country comparisons of hospital cases as a means of obtaining important insights into health care delivery strengthening. Using the report and templates for data collection provided in its annexes, other countries can customize and conduct similar comparisons based on their specific needs and circumstances. As the study shows, such comparisons may help identify potential over- or under-provision of inpatient care by specialty, surgical and medical case types, and specific common diagnoses or case types. To ensure the quality and relevance of such analysis, it is crucial to compare case-mix structure and admission rates with variations in demographics, epidemiology, health service capacity, and health financing across countries. Doing so will help assess the reasonableness of the observed variations and support the development of policy actions to reduce undercoverage or inefficient overprovision of inpatient care, both generally and in specific specialties or case types. It will be important for countries to monitor the variation in ALOS across hospitals by individual DRG in the future, aiming to identify outliers and investigate the underlying reasons for their occurrence. This recommendation is supported by the findings of the study, which indicate significant differentials across facilities for the same disease type, despite the limitations of a small and skewed hospital sample. In addition to comparing case-mix structure and admission rates, it is also important to compare coding activity and prevalent ICD-10 diagnosis codes and procedure codes. Such comparisons can help identify potential anomalies in coding that could indicate upcoding in some countries or national variation in coding practices that should be further explored. By systematically comparing and analyzing health care utilization patterns and outcomes across countries, policy makers and health care professionals can develop targeted interventions to optimize health care delivery, enhance patient outcomes, and promote efficient use of health care resources. Collaboration in comparing hospital activity across countries will require a dedicated effort. It will be important to establish clear rules and agreements between interested countries. This includes setting rules for routine information reporting, assigning responsibility for analysis, and ensuring uniformity in coding and grouping rules. To enable cross-country comparisons using DRGs, it is crucial to assess which specific or aggregated DRGs are comparable across countries. This should not be difficult for countries using a similar version of DRG but may require additional efforts for those using different versions or classification systems. It is important to ensure that the data used in cross-country comparisons are accurate, reliable, and standardized to minimize the risk of misinterpretation or misrepresentation of results. Overall, collaboration in comparing hospital activity can help participating countries achieve the same benefits mentioned above—improved health care delivery, enhanced patient outcomes, and more efficient use of health care resources. The exploratory nature of the study generated new ideas for potential analyses that can be done and would be useful with this kind of data. Possible topics include (1) examining the split between elective and unplanned cases, especially for surgical procedures, as a high proportion of unplanned surgeries could indicate a weak referral system; (2) exploring multimodal distributions of length of stay as it could help refine the DRG classification system; (3) estimating the prevalence of cases that could have been avoided or that could have been moved to day surgery. Such types of analyses could be done in the framework of a regional 57 case-mix network. More importantly, similar analyses could be linked to financial indicators, allowing for the calculation of potential cost savings if low-performing hospitals/countries were to achieve the mean, or if all hospitals reached the level of the best/frontier hospital. To enable routine cross-country case-mix comparison at the global or regional level, the study team suggests establishing a regional case-mix network. The network should not be a formal institution but rather an informal platform that links together country technical teams, enabling them to communicate with each other and collaborate on further analysis of the same type. It can lead to agreements about common set of rules for data collection and analysis, as well as a shared understanding of the limitations of the data. In addition, the team recommends creating a code map between different versions of ACHI and NOMESCO Classification of Surgical Procedures, similar to the existing map in Australia, to enable comparisons with NordDRG countries. These efforts will help improve the accuracy and consistency of case- mix data across regions and facilitate the identification of areas for improvement in health care provision and resource allocation. Routine use of hospital case-mix data has the potential to significantly improve health system management. To leverage this potential, it is important to identify key areas for data analysis. One approach is to compare hospitals to identify potential gaps in service provision capacity for the top-20 medical and surgical DRGs and develop plans to increase capacity or facilitate referrals for these case types. Additionally, identifying hospitals with unusually long ALOS for the most prevalent DRGs can help to determine if these hospitals have more complex cases or inefficient service provision, and allow for corrective action to be taken. Hospitals with short ALOS or same-day admission and discharge for surgical cases can be encouraged to use ambulatory surgery more frequently. DRG-related key performance indicators—such as the case-mix index, weighted activity units, and percentages of cases in surgical DRGs and those with high levels of complications—can be calculated to refine the designation of hospitals as tertiary referral versus secondary hospitals. Implementation of studies of ambulatory care–sensitive conditions can help assess which diseases need greater attention in primary care and which geographic areas are underperforming in primary care. Improving coding accuracy and uniformity in coding rules is also essential to ensure that DRGs accurately reflect hospital case-mix and to detect upcoding during claims review. For the participating countries, the following specific recommendations are made with the aim of improving hospital case-mix data analysis. First, reasonably generous limits for reporting secondary diagnosis codes should be set to ensure that cases are accurately assigned to the correct diagnostic category and level of complexity for DRG assignment. Second, the accuracy and standardization of clinical coding within each country could be improved by sharing national coding standards and comparing the most common ICD-10 diagnosis and procedure codes. Finally, Ukraine should consider adoption of a full DRG classification system. 58 References Bredenkamp, Caryn, Elina Dale, Olena Doroshenko, Yuriy Dzhygyr, Jarno Habicht, Loraine Hawkins, Alexandr Katsaga, et al. 2022. “Health Financing Reform in Ukraine: Progress and Future Directions.” Overview Booklet, International Development in Focus. Washington, DC: World Bank. https://openknowledge.worldbank.org/bitstream/handle/10986/37585/9781464818400- ov.pdf?sequence=5&isAllowed=y. Busse, Reinhard. 2012. “Do Diagnosis-Related Groups Explain Variations in Hospital Costs and Length of Stay?–Analyses from the EURODRG Project for 10 Episodes of Care across 10 European Countries.” Health Economics 21 (Supplement 2): 1–5. Busse, Reinhard, Alexander Geissler, Wilm Quentin, and Miriam Wiley. 2011. “Diagnosis-Related Groups in Europe: Moving Towards Transparency, Efficiency and Quality in Hospitals.” Open University Press. ISBN: 978-0-33-524557-4. https://www.bmj.com/content/346/bmj.f3197. Chun, C.-B., S.-Y. Kim, J.-Y. Lee, and S.-Y. Lee. 2009. “Republic of Korea: Health System Review.” Health Systems in Transition 11 (7): 1–184. https://apps.who.int/iris/handle/10665/330337. CNAM (Compania Nationala de Asigurari in Medicina). 2022. “Mandatory Health Insurance System in the Republic of Moldova.” Presentation at Korean Moldovan webinar organized by the World Bank, August 23–25. Džakula, A., A. Sagan, N. Pavić, K. Lončarek, and K. Sekelj-Kauzlarić. 2014. “Croatia: Health System Review.” Health Systems in Transition 16 (3): 1–162. Džakula, A., D. Vočanec, M. Banadinović, M. Vajagić, K. Lončarek, I. Lukačević Lovrenčić, D. Radin et al. 2021. “Croatia: Health System Review.” Health Systems in Transition 23 (2): 1– 146. https://eurohealthobservatory.who.int/publications/i/croatia-health-system-review-2021. Edwards, N. 2011. “Improving the Hospital System in the Republic of Moldova.” Health Policy Paper Series 1. Copenhagen: World Health Organization Regional Office for Europe. Garam, I., M. Zadnipru, V. Doronin, A. Matei, and I. Mosca. 2020. Can People Afford to Pay for Health Care? New Evidence on Financial Protection in the Republic of Moldova. Copenhagen: World Health Organization Regional Office for Europe. https://apps.who.int/iris/bitstream/handle/10665/331667/9789289054959-eng.pdf. Habicht, T., M. Reinap, K. Kasekamp, R. Sikkut, L. Aaben, and E. van Ginneken. 2018. “Estonia: Health System Review.” Health Systems in Transition 20 (1): 1–193. Kim, J.-Ho, J. Ah-Reum An, S.-J. J. Oh, J. Oh, and J.-K. Lee. 2021. “Emerging COVID-19 Success Story: South Korea Learned the Lessons of MERS,” March 5. https://ourworldindata.org/covid-exemplar-south-korea. Korea, NHIS (National Health Insurance Service). 2022. “National Health Insurance and Long-Term Care Insurance System in the Republic of Korea.” https://www.nhis.or.kr/english/wbheaa03500m01.do?mode=download&articleNo=10814171 &attachNo=323871. 59 Medin, Emma, Unto Häkkinen, Miika Linna, Kjartan S. Anthun, Sverre A. C. Kittelsen, and Clas Rehnberg (on behalf of the EuroHOPE Study Group). 2013. “International Hospital Productivity Comparison: Experiences from the Nordic Countries.” Health Policy 112: 80– 87. https://www.sciencedirect.com/science/article/pii/S0168851013000444#sec0065. Moldova, National Agency for Public Health. 2021. Statistical Yearbook of the Health System 2020. Cited in World Bank, “Benchmarking of Human Development Outcomes: Moldova” (draft), November 18. OECD (Organisation for Economic Co-operation and Development). 2016. “OECD Health Policy Overview: Health Policy in Korea.” April. https://www.oecd.org/korea/Health-Policy-in- Korea-April-2016.pdf. OECD. 2020. “OECD Reviews of Public Health: Korea: A Healthier Tomorrow.” Paris: OECD Publishing. https://doi.org/10.1787/be2b7063-en. OECD (Organisation for Economic Co-operation and Development) and European Observatory on Health Systems and Policies. 2021a. “Croatia: Country Health Profile 2021.” State of Health in the EU. Paris: OECD Publishing. https://doi.org/10.1787/717e5510-en. OECD and European Observatory on Health Systems and Policies. 2021b. “Estonia: Country Health Profile 2021.” State of Health in the EU. Paris: OECD Publishing, and Brussels: European Observatory on Health Systems and Policies. OECD and European Observatory on Health Systems and Policies. 2021c. “Health Systems in Action: Ukraine.” Paris: OECD Publishing. https://eurohealthobservatory.who.int/publications/i/health-systems-in-action- ukraine#:~:text=Overview,care%20or%20medicines%20or%20not. OECD and European Observatory on Health Systems and Policies. 2021d. “Romania: Country Health Profile 2021.” State of Health in the EU. Paris: OECD Publishing. https://doi.org/10.1787/74ad9999-en. OECDStat. n.d. “Health Care Resources.” https://stats.oecd.org/Index.aspx?DataSetCode=HEALTH_REAC#. Our World in Data. n.d. “COVID-19 Data.” https://github.com/owid/covid-19- data/tree/master/public/data Scîntee, S. G., and C. Vlãdescu. 2022. “Overview of the Main Incremental Health Care Reforms Introduced between 2014 and 2020 in Romania.” South Eastern European Journal of Public Health 1. DOI: 10.11576/seejph-5477. Street, Andrew, Conrad Kobel, Thomas Renaud, and Josselin Thuiliez (on behalf of the Euro DRG Group). 2012. “How Well Do Diagnosis-Related Groups Explain Variations in Costs or Length of Stay among Patients and across Hospitals? Methods for Analysing Routine Patient Data.” Health Economics 21 (Supplement 2): 6–18. https://onlinelibrary.wiley.com/doi/full/10.1002/hec.2837. Turcanu, G., S. Domente, M. Buga, and E. Richardson. 2012. “Republic of Moldova Health System Review.” Health Systems in Transition 14 (7): 1–151. 60 https://www.researchgate.net/publication/233848539_Republic_of_Moldova_Health_System _Review#pf78. Vlãdescu, C., S. G. Scîntee, V. Olsavszky, C. Hernández-Quevedo, and A. Sagan. 2016. “Romania: Health System Review.” Health Systems in Transition 18 (4): 1–170. WHO (World Health Organization). n.d. (a). Coronavirus Disease (COVID-19) Database. https://github.com/owid/covid-19-data/tree/master/public/data. WHO. n.d. (b). Global Health Expenditure Database. https://apps.who.int/nha/database/ViewData/Indicators/en. WHO. n.d. (c). Global Health Observatory Database. https://www.who.int/data/gho/data/indicators. WHO (World Health Organization) European Region and European Health Information Gateway. 2022. “European Health for All Database” (HFA-DB). https://gateway.euro.who.int/en/datasets/european-health-for-all-database/#health-care- utilization-and-expenditure. Accessed September 19, 2022. WHO European Region and European Health Information Gateway. n.d. (a). “Acute Care Hospital Discharges per 100.” https://gateway.euro.who.int/en/indicators/hfa_536-6020-acute-care- hospital-discharges-per-100/visualizations/#id=19631&tab=table. WHO European Region and European Health Information Gateway. N.d. (b). “Indicators Explorer.” https://gateway.euro.who.int/en/hfa-explorer/. World Bank. 2022. Country Note on Resilient Health System in Moldova (unpublished). Washington, DC. 61 Annexes Annex 1. Hospital-Level Data Processing and Cleaning At the hospital level, data processing involves the amalgamation of data from various countries into a single Microsoft Excel spreadsheet and the cleaning of these data, as described below. Merging of data Even though countries were supposed to submit the data according to the template, minor adjustments were needed before merging all data into one spreadsheet. For example, when the secondary diagnosis codes and procedure codes of the case were submitted in separate fields (and not in one field separated by a comma as suggested in the template), the codes were merged into one field by using the Excel CONCATENATE function. After merging the single-country data into one spreadsheet, additional information was added to facilitate further analysis. For example: • Columns with the English text for diagnosis, procedure, and DRG codes were added. • End-of-treatment information was adjusted according to the template, where there were three options: (1) discharged home, (2) referred to another hospital, and (3) death. It appeared that an additional option—“other”—was needed for cases when the discharge information was unclear, the patient was referred internally within the same institution, the patient left against medical advice, or the original file indicated “other.” Hospital-level and profile information was harmonized according to the typology developed specifically for this study. The typology was discussed and agreed upon with participating countries. According to the typology, hospitals were divided into mono-profile and multiprofile hospitals, as follows: Monoprofile included cardiology hospitals and maternity hospitals. Multiprofile included multiprofile nonuniversity hospitals and multiprofile university hospitals. Length of stay (LOS) of each case was checked and recalculated (if needed) according to the following formula: LOS (days) = date of discharge minus date of admission. Gender information was harmonized, as follows: 1 = male, 2 = female, and 3 = unknown. Additional columns were added with information needed for the further data analyses, for example, number of secondary diagnoses, number of procedure codes, etc. Cleaning of data All rows were checked against the case definitions. • Nonrelevant rows were removed from further analysis. • In some instances, the case type was changed manually. For example, some cases with main diagnosis O82 Single delivery by cesarean section (and subcodes) were defined by the original- country database as vaginal delivery. As O82 is for coding C-section, those cases were changed 62 from Vaginal delivery to C-section. Cases with the following conditions were excluded from the analysis: • Discharge date other than year 2021. • Negative LOS without DRG information • Grouped to error DRGs • AMI diagnosis I21 (Acute myocardial infarction) and I22 (Subsequent myocardial infarction) and subcodes as secondary diagnosis codes because they did not meet the case definition criteria • Cases grouped to MDC 15 Newborns and Other Neonates but submitted under case definition Vaginal delivery or C-section Table 1A-1. Reasons for Excluding the Data from Analysis Country Reason for exclusion Number of excluded Percentage of cases excluded cases Ukraine AMI in surgical DRG 1,845 5.8 Error DRG 165 0.5 Duplication 38 0.1 Negative LOS 21 0.1 Case not meeting case definition criteria 1 0.003 Ukraine total 2,070 6.5 Moldova AMI in surgical DRG 914 6.0 Case not meeting case definition criteria 7 0.0 Moldova total 921 6.0 Romania Year of discharge other than 2021 8,313 48.1 Case not meeting case definition criteria 1,029 5.9 AMI in surgical DRG 175 1.0 Identical case after creation of new case ID 1 0.01 Romania total 9,518 55.0 Croatia AMI in surgical DRG 758 12.0 Case not meeting case definition criteria 516 8.2 No DRG information 31 0.5 DRG and main diagnosis of newborn 30 0.5 Croatia total 1335 21.2 Estonia AMI in surgical DRG 1,319 15.1 Identical case after the creation of new case 227 2.6 ID Year of discharge other than 2021 79 0.9 Case not meeting case definition criteria 4 0.05 Estonia total 1,629 18.6 Grand total 15,473 19.5 Source: Authors’ calculations based on data provided by country teams. Notes: AMI = Acute myocardial infarction; DRG = Diagnosis -related group; LOS = Length of stay. All plus (+) signs were removed from diagnosis codes. Duplications were addressed. All countries but Moldova had numerous case IDs that were not unique. Based on discussion with countries, corrections were made in the database as follows: • For Romania, an artificial case ID was created for all cases by merging (using MS Excel CONCATENATE function) old case ID, provider ID, age, DRG, and year of discharge. • For Croatia, duplicates were removed by replacing originally submitted data with a new data set. • For Ukraine, procedure codes of cases with identical case IDs were merged, and the redundant 63 cases were excluded. • For Estonia, among cases with identical case IDs, those grouped to C-section DRG but indicated as vaginal delivery case type were excluded. Missing MDC information was added if DRG information was available. Punctuation marks were removed from diagnosis codes, for example, Z37.0 became Z370. Among AMI cases, only those grouped to conservative DRGs were included in further analysis. Initially the idea was to exclude the cases with bypass procedures, but due to the difficulty in excluding only bypass procedures in the final database, all cases with surgical interventions were excluded from the analysis. 64 Annex 2. Templates for Collection of Country- and Hospital-Level Data Guidance for data collection Information in all tables concerns only inpatient cases. Information in MDC and DRG tables concerns the cases that have been assigned into DRGs regardless of the payment method. Table 2A-1. Country-Level Data 1. Distribution of cases between MDCs 2019 2020 2021 MDC code MDC text # of cases # of cases # of cases 2. Distribution of cases between DRGs 2019 2020 2021 Surgical (S) / # of ALOS # of ALOS # of ALOS DRG code DRG text Medical (M) DRG cases (days) cases (days) cases (days) 3. Number of diagnoses (only main diagnosis) 2019 2020 2021 Diagnosis code Diagnosis text # of cases # of cases # of cases 4. Number of interventions 2019 2020 2021 Intervention code Intervention text # of cases # of cases # of cases 5. Distribution of cases between hospitals 2019 2020 2021 Hospital Hospital name # of cases # of cases # of cases profile/level Source: Developed by authors Notes: MDC = Major diagnostic category; DRG = Diagnostic-related group. Table 2A-2. Hospital-Level Data Variable Example Comments 123456A Case ID BC Unique value for each case in the data sample Provider's ID or name Main diagnosis K80.2 ICD-10-AM, ICD-10 original, other Secondary O32.1, As many as coded, comma-separated if more than one diagnosis code(s) Z37.0, code 96197- 19, As many as coded, comma-separated if more than one Procedure code(s) JAA00 code Gender 1 1 = male, 2 = female, 3 = unknown Age of patient 42 Patient age in days at admission End of treatment Discharged home, referred to other hospital, death Date of admission 5.2.2021 Date of discharge 5.8.2021 Length of stay LOS in days = Date of discharge minus date of (LOS) 6 admission In case the case has been assigned to DRG. If not, the MDC field remains empty 65 In case the case has been assigned to DRG. If not, the DRG field remains empty Country Name of the country Profile of hospital Multiprofile, mono-profile, children’s, maternity, etc. Level of hospital University, county, regional, secondary, tertiary, etc. Number of beds Source: Developed by authors Notes: MDC = Major diagnostic category; DRG = Diagnostic-related group; ICD-10 = International Classification of Diseases-10th Revision. 66 Annex 3. Questionnaire for Collecting the Country-Specific DRG System Data Question/topic Answer/ Additional description comments DRG system in use (e.g., NordDRG, AR-DRG, etc.) Since when the DRG system is in use Was the system adopted, self-developed, something else (please specify)? Frequency of update of: ● DRG grouping algorithm ● DRG tariffs/cost weights Location of the grouper software: ● Centrally at purchaser or any other organization ● At each hospital ● Other (please specify) Organization(s) responsible for maintenance and update of ● DRG grouping algorithm ● DRG tariffs/cost weights Are other stakeholders/parties involved in maintenance and update of DRG system? If yes, please specify Use of DRGs, for example, for ● Patient classification (since when) ● Performance monitoring (since when) ● Benchmarking (since when) ● Budget allocation (since when) ● Contracting (since when) ● Reporting ● Invoicing ● Other (please specify, since when) Number of MDCs Number of DRGs Are DRGs spilt based on severity? If yes, describe the severity levels Which primary classifications are used in DRG system: ● For diagnosis (name of the version, last update) ● For interventions (name of the version, last update) Organization responsible for maintenance and update of primary classifications Use of coding guidelines ● National ● International Organization responsible for maintenance and development of coding guidelines Are clinical coding audits in place? If yes, which is the organization responsible for conducting the coding audits? Scope/coverage of DRGs, for example: ● Inpatient care ● Day surgery ● Day care ● Outpatient care ● Specific clinical area or specialty ● Other (please specify) 67 Are DRGs used in non- or subacute care (e.g., rehabilitation, psychiatric care, follow-up care, other)? If yes, are the DRGs also used for payment? Are DRGs combined with other payment methods in different care settings (inpatient, outpatient, day care, day surgery, etc.)? For example: ● DRGs + FFS (fee-for-service) ● DRGs + per diem ● DRGs + global budget ● Other (please specify) Are outliers taken into account if DRGs are used for reimbursement? If yes, are they based on LOS, cost, something else? Are the DRG tariffs, cost weights, and base rate the same for all providers? If not, explain the differences. Source: Developed by authors Notes: NordDRG = Nordic DRG; AR-DRG = Australian Refined diagnosis-related group; MDCs = Major diagnostic categories; LOS = Length of stay. 68 Annex 4. Summary of DRG systems in countries Table 4A-1. Summary of Countries’ DRG System Indicator Ukraine Moldova Romania Croatia Estonia Korea DRG version AR-DRG (A-DRG AR-DRG RO-DRG (variation on AR-DRG NordDRG, Estonian K-DRG (V 4.4) level) AR-DRG) version Use of DRG other Patient classification-– Patient classification-– Patient classification-– Patient classification-– Patient classification-– Patient classification than for payment 2020 2011 2003 (for 23 hospitals, 2009 2001 and performance now for all acute cases) monitoring—1994 Performance Benchmarking: Performance Benchmarking-–2012 monitoring-–2021 analysis of hospital Performance monitoring-–2009 Contracting—2012 (for Performance activity if necessary monitoring-– 2003 clinics and hospitals); Budget allocation-– Benchmarking-–2009 monitoring-–2012 2013 (for tertiary and 2020 Budget allocation-– Benchmarking-–2003 Reporting-–2009 Budget prognosis only general hospitals) 2012 Contracting—2020 Budget allocation-– (without budget Invoicing—2009 (not Contracting-–2012 2003 allocation) for payment; only for Contracting-–2003 justifying the budget) Procedure ACHI Australian procedure Australian procedure ACHI NOMESCO Electronic Data classification codes before ACHI codes before ACHI Classification of Interchange (EDI) Surgical Procedures medical procedure code Number of DRGs 372 698 669 671 677 2,721 in the classification Number of 372 399 403 405 Not applicable 1,128 adjacent DRGs Number of MDCs 27 25 25 25 27 26 Scope of services Acute inpatient Acute inpatient; Acute inpatient; Acute inpatient Acute inpatient; Acute inpatient covered by DRG day care subacute care and day day surgery classification care for reporting, performance monitoring Source: Authors’ representation of information provided by country teams. Notes: AR-DRG = Australian Refined diagnosis-related group; NordDRG = Nordic DRG; ACHI = Australian Classification of Health Interventions; NOMESCO = Nordic Medico-Statistical Committee; MDCs = Major diagnostic categories Annex 5. Overview of Countries’ Health Care Systems This annex presents an overview of the health care systems of the countries that participated in the study. The purpose of this overview is to provide context for the case-mix analysis described in the body of report. It summarizes key indicators that reflect the performance of these systems, the ways in which they are financed, and the composition of the health care workforce and hospital network in each country. It also discusses the impact of the COVID- 19 pandemic on these systems and the recent changes that have occurred in them. Whenever possible, data from the same year have been used for cross-country comparison, even if this is not the most recent available year. Ukraine In 2019, Ukraine spent US$907 PPP per capita on health care, representing 7.10 percent of its GDP. Out-of-pocket contributions funded 51.1 percent of health spending in Ukraine, followed by government spending, which covered 45.5 percent. The remaining 3.4 percent was funded through voluntary health insurance (VHI) (WHO n.d. [b]). In Ukraine, the majority of government health spending is allocated to inpatient and outpatient care, while OOP payments primarily cover spending on pharmaceuticals. Financing system. In Ukraine, the National Health Service of Ukraine (NHSU) is the single purchaser of health care services. It is funded through general taxation from the state budget (OECD and European Observatory on Health Systems and Policies 2021c), with additional financing provided at the regional level. All citizens and residents of Ukraine are entitled to the services outlined in the state-guaranteed program. The NHSU's Program of Medical Guarantees (PMG) defines the benefits package available in Ukraine, which includes inpatient, outpatient, and primary health care services. This package is supplemented by the Affordable Medicines Program, which covers outpatient medicines for conditions such as type 2 diabetes and cardiovascular diseases. Health workforce. In 2019, the number of physicians per 1,000 population in Ukraine was 4.17, and the number of nurses and midwives per 1,000 population was 7.43. Both of these figures have decreased in recent years due to an aging workforce and low wages (OECD and European Observatory on Health Systems and Policies 2021c). Organization of the hospital system. The health care system in Ukraine has undergone several reforms in recent years, moving away from a hospital-centric model and promoting primary health care. Secondary- and tertiary-care hospitals in Ukraine include maternity and childcare facilities, city hospitals, and rayon-level hospitals. The hospital system is still undergoing reform, shifting from mono-profile hospitals to multiprofile hospitals. The number of hospital beds has significantly decreased in recent years, reaching 6.65 per 1,000 population in 2019 (down from 9.5 in 2000) (OECD and European Observatory on Health Systems and Policies 2021c). There is a high level of unmet health care needs in Ukraine due to low government spending on health and high OOP expenses (OECD and European Observatory on Health Systems and Policies 2021c). Reforms. The health care reform in Ukraine was launched in 2015 with the passage of the Law on Financial Guarantees for Health Care Services, which defined the benefits package and established the NHSU. The reform process in Ukraine is ongoing, and the benefits package is being expanded to provide more services to patients. As the health care financing system in Ukraine continues to evolve, case-based payments, including DRGs, are being used to pay for some types of care and are being adapted to the Ukrainian health care system. In addition to reforming health care financing, the government has developed a plan to reorganize the hospital system in Ukraine to meet patients' needs better. Impact of COVID-19. To support the health care system in providing care during the COVID- 19 pandemic, Ukraine implemented new health financing measures focused on treatment, diagnostics, and vaccination. The pandemic had a significant impact on the health care system and delayed the implementation of health care reforms (Bredenkamp et al. 2022). Moldova In Moldova, health care expenditure represented 6.38 percent of GDP, or US$860 per capita (in PPP terms) in 2019. Government spending and mandatory contributions account for 60 percent of total health care expenditures, while out-of-pocket payments reach 36 percent, which is one of the highest shares among European countries. VHI is not widely popular in Moldova and makes up slightly over 1 percent of total health care expenditure (WHO n.d. [c]). Financing system. The health care system in Moldova is based on a mandatory health insurance scheme and a centralized institution, the National Health Insurance Company (CNAM), which pools public funding for health and acts as the single purchaser of health care services. In 2016, CNAM's revenue consisted of mandatory contributions from employees (9 percent of payroll) (CNAM 2022), transfers from the state budget for unemployed and pensioners (a total of 13 categories of citizens) (CNAM 2022), and individual contributions from self-employed individuals. These contributions represented 55 percent, 44 percent, and 1.5 percent of CNAM's revenue, respectively (Garam et al. 2020). Both public and private health care providers can enter into contracts with CNAM and provide services to the population under CNAM arrangements. Currently, approximately 88 percent of the population of Moldova is insured (Garam et al. 2020). Basic services, such as emergency care, primary health care, and certain drugs and hospitalizations, are provided to all individuals regardless of insurance status. In Moldova, hospital treatment, including day-care surgery, is financed through the use of DRGs, up to a contracted volume set for each hospital. As advised by CNAM, DRG payments accounted for over 80 percent of its expenditures for acute inpatient care in 2017, with the remainder reimbursed through a per bed-day mechanism, global budgets, or retrospective payments. Expensive medications and consumables, dialysis, radiotherapy, and doctors' salaries are financed in addition to DRG payments through separate budgets, with the conditions varying across hospitals. Health workforce. In 2020, Moldova had approximately 3.1 doctors and 4.7 nurses (or midwives) per 1,000 inhabitants, a significantly lower share than the European Union (EU) average of 3.9 and 8.4, respectively. However, these indicators mask significant regional disparities: in 2016, the average number of physicians per 1,000 inhabitants was 3.7, but it was only 0.6 in rural areas, compared to 7.8 in urban areas (Garam et al. 2020). Organization of the hospital system. In Moldova, there are approximately 85 hospitals, with about half located in the capital city of Chisinau (World Bank 2022). These hospitals include 10 municipal, 35 district, 17 republican, 6 other than Ministry of Health subordinates, and 17 private hospitals (Moldova, National Agency for Public Health 2021). 71 Moldovan hospitals primarily offer acute care rather than palliative, long-term, or rehabilitation care. Capacity for the latter is not well developed, affecting the system's overall efficiency (Turcanu et al. 2012). The number of beds in Moldova has been decreasing, from 7.1 per 1,000 population in 2000 to 5.7 per 1,000 population in 2014 (WHO n.d. [c]), which is still above the European average. However, the uneven distribution of beds across the country creates issues with equal access to high-quality care for Moldovan citizens. Another challenge is that the average length of stay, at 8 days, remains above the European average of 7.5 days, though it has decreased over time (it was 14.2 days in 2000) (WHO European Region and European Health Information Gateway 2022). Reforms. Moldova has implemented various purchasing mechanisms to improve health care, including age-adjusted capitation in primary health care, case-mix funding for hospitals, and performance-based financial incentives. In the early 2000s, the country sought to optimize its hospital network by closing many rural hospitals. However, the next step, addressing larger hospitals in district centers and cities, proved more challenging. The National Hospital Masterplan 2009–2018, which has yet to be implemented, proposes the replacement of district hospitals with a network of nine regional hospitals and two specialized centers in Cahul and Balti (Edwards 2011). In addition, hospitals operated by the Ministry of Defense and Ministry of Interior (some of which are contracted by the National Health Insurance Company) continue to provide services for the military and police, and this issue must also be addressed. Impact of COVID-19. The rapid increase in COVID-19 cases strained human resources and bed capacity, reducing the availability of all types of care, including essential care. A comparison of services provided in the first nine months of 2019 and 2020 demonstrates that the pandemic significantly impacted the delivery of essential care. However, certain interventions implemented during the pandemic, such as disease surveillance, public-private partnerships, and telemedicine, could potentially improve the health care system in the long term (Garam et al. 2020). Romania In 2019, Romania spent US$1,907 PPP per capita on health care, equivalent to 5.7 percent of its GDP. The majority of health spending in Romania is financed through government and compulsory insurance schemes (over 80 percent), with approximately one-fifth funded through out-of-pocket contributions and less than 1 percent through VHI. Of the EU countries, Romania allocates the highest share of its health spending to inpatient care (44 percent), although the absolute amount is relatively low. Outpatient medicines and medical goods comprise 27 percent of health spending, followed by outpatient care (19 percent), long-term care (6 percent), and prevention (1 percent). OOP payments primarily cover pharmaceutical spending, while inpatient care is almost entirely funded by government and compulsory funding (99 percent) (OECD and European Observatory on Health Systems and Policies 2021d). Financing system. Romania’s compulsory social health insurance system requires employees to contribute to the National Health Insurance Fund. This system guarantees a benefits package to all insured individuals, including health care services (such as prevention, outpatient, specialist, and hospital care) and pharmaceuticals and medical devices (Vlãdescu et al. 2016). Approximately 11 percent of the population in Romania remains uninsured despite the compulsory social health insurance system and is eligible for only the minimum benefits 72 package, which covers life-threatening emergencies, infectious diseases (including COVID- 19), and care during pregnancy. Health workforce. In 2019, the number of physicians per 1,000 population in Romania was 3.2, and the number of nurses per 1,000 population was 7.5. These relatively small numbers are due to the migration of medical personnel, which has had a negative impact on access to care (OECD and European Observatory on Health Systems and Policies 2021d). Organization of the hospital system. The health care system in Romania is hospital-centric; primary care services are underutilized, and the number of hospital beds is relatively high (seven per 1,000 population). Patients often bypass primary care and turn to emergency departments in hospitals for nonurgent care. Despite this, ALOS in hospitals in Romania is close to the EU average, at 7.3 days. Unmet health care needs in Romania have continuously declined, falling by more than half between 2011 to 2019, from 12.2 percent to 4.9 percent (OECD and European Observatory on Health Systems and Policies 2021d). Reforms. As part of the 2014–2020 Health Strategy, Romania has reformed and restructured its regional hospital network, redistributing acute and long-term beds, improving the quality of services in hospitals, and updating its DRG system. The reform has focused on key areas, including improving governance, increasing resources for health, expanding coverage and access, and optimizing health care organization and quality of care (Scîntee and Vlãdescu 2022). In Romania, hospitals receive prospective payments, through a mix of payment methods, including the Romanian DRG (RO-DRG) system (Vlãdescu et al. 2016). In recent years, the DRG system initially introduced in Romania between 1999 and 2000 has been updated, and tools have been developed to improve the quality and performance of hospitals (Scîntee and Vlãdescu 2022). Impact of COVID-19. As of February 2022, there were 167 COVID-19 cases and 3.4 COVID- 19 deaths per 1,000 population in Romania. The number of cases was significantly lower than the European average, while the number of deaths was significantly higher, suggesting a substantial underreporting of COVID-19 cases in Romania (OECD and European Observatory on Health Systems and Policies 2021d). Despite facing challenges, such as lack of a sufficient health workforce, Romania managed to double the number of intensive care unit (ICU) beds with the aid of external financial support for building mobile ICUs (OECD and European Observatory on Health Systems and Policies 2021d). Services were also reorganized to support COVID-19 patients by designating specific hospitals, wards, and outpatient facilities for treatment of COVID-19. Croatia In 2019, Croatia allocated approximately 7 percent of its GDP, or US$2,168 PPP per capita, to health care spending. The majority of this spending was financed through government and compulsory insurance schemes (80 percent), with the remaining amount supported by out-of- pocket contributions (11.5 percent) and voluntary health insurance (6.6 percent). The majority of health care spending was allocated to outpatient care (38 percent), followed by inpatient care (30 percent), pharmaceuticals and medical devices (23 percent), long-term care (3 percent), and prevention (3 percent). Out-of-pocket payments primarily covered spending on pharmaceuticals and dental care, while inpatient care was predominantly funded through government and compulsory sources (89 percent) (OECD and European Observatory on Health Systems and Policies 2021a). 73 Financing system. The primary source of public funding for health care in Croatia is mandatory health insurance, which is administered and purchased by the Croatian Health Insurance Fund (CHIF) (Džakula et al. 2021). Contributions to the CHIF are made by employees, who also cover their dependents, and by the government, which covers the unemployed, pensioners, and vulnerable groups. In addition to its mandatory insurance offerings, the CHIF and private insurers also provide complementary insurance, which can be purchased voluntarily to cover copayments. The benefits package covered by the mandatory health insurance system includes most curative and preventive health care services. The provision of these services is subject to a contract with CHIF (OECD and European Observatory on Health Systems and Policies 2021a). Health workforce. In 2019, Croatia had 3.5 doctors and 6.8 nurses per 1,000 population. These ratios have been steadily increasing over the past two decades (OECD and European Observatory on Health Systems and Policies 2021a; Džakula et al. 2021). Organization of the hospital system. Croatia has a relatively high number of hospital beds— 5.7 beds per 1,000 population in 2019—with only a slight decrease over the past two decades. One of the main challenges facing the hospital system in Croatia is the focus on rehabilitation and long-term care in hospitals rather than in the community. The average length of stay in hospitals in Croatia was 8.1 days in 2019; while decreasing, this measure remains above the EU average and places a heavier burden on hospital outpatient departments. However, the number of hospital discharges and doctor consultations per 1,000 population is at the EU average, and the prevalence of unmet needs for medical care is relatively low, at 1.4 percent (OECD and European Observatory on Health Systems and Policies 2021a). In Croatia, hospitals that have contracted with the Croatian Health Insurance Fund are part of the National Health Care Network of health care centers and are primarily funded through global budgets (90 percent), with only 10 percent of funding tied to the services provided. These hospitals are divided into clinical, general, and specialized hospitals; clinical hospitals are established under the Ministry of Health and the others established under the counties (Džakula et al. 2021). In 2019, there were 63 hospital institutions and centers: five university hospital centers (two located in Zagreb and one each in Rijeka, Split, and Osijek); three university hospitals; five clinics (for infectious diseases, children, orthopedics, psychiatry, and cardiovascular diseases); 22 general hospitals; and 28 special hospitals, treatment centers, and hospices. Most of these institutions are publicly owned; only 11 hospitals are private. The hospital network in Croatia is characterized by well-developed infrastructure, although there are challenges related to geographical distribution, particularly in rural areas and on the country's islands (Džakula et al. 2021), where there are higher rates of unmet health care needs. In other parts of the country, there may be a surplus of hospitals relative to the need. Reforms. Croatia has implemented several reforms in the health care sector, including changes to financing and broader reforms of the hospital system. In 2009, Croatia implemented a modified version of the Australian Refined-DRG (AR-DRG) system for health care financing (Džakula et al. 2014), replacing fee-for-service payments. However, hospitals in Croatia are primarily funded through global budget schemes. As a result, the new financing system is largely used for budget justification and reporting purposes rather than determining payment for individual services. In 2015, CHIF was separated from the State Treasury in an effort to increase transparency, stabilize the health care budget, and support efficiency. At the same time, a new financing model for hospitals was partially implemented (Džakula et al. 2021). 74 In the years following the 2015 reforms, additional changes have been implemented to reorganize and restructure hospitals (through the National Hospital Development Plan), alter the hospital payment model, address hospital debts, reduce waiting times, decrease the average length of stay in hospitals, reduce the number of acute care beds, and improve the overall efficiency of the health care system (Džakula et al. 2021). Impact of COVID-19. The COVID-19 pandemic had a significant impact on the health of the population of Croatia, including a high number of cases during the second wave. As of February 2022, the country had a death rate much higher than the European average, at 4.1 cases per 1,000 population. In addition to the direct impact of the pandemic, there was also a negative impact on access to health care services, as the prevalence of unmet needs for medical care increased, and approximately 25 percent of the population had not received necessary treatment or examination within the first year of the pandemic. This increase in unmet needs was largely due to the postponement of nonessential services and patients’ reluctance to seek care lest they contract COVID-19 (OECD and European Observatory on Health Systems and Policies 2021a). To support the health care system during the pandemic, the surge capacity of health care workers was strengthened, and some hospitals and outpatient facilities were designated as COVID-19 facilities. There were no reports of shortages of equipment, supplies, or capacity (OECD and European Observatory on Health Systems and Policies 2021a). Estonia In 2019, Estonia allocated 6.7 percent of its GDP, or US$2,617 PPP per capita, to health care spending. The majority of this spending was financed through government and compulsory insurance schemes (74.4 percent), with the remaining quarter (24 percent) coming from out- of-pocket contributions and voluntary health insurance. The majority of health care spending was allocated to outpatient care (42 percent), followed by inpatient care (25 percent), pharmaceuticals and medical devices (19 percent), long-term care (9 percent), and prevention (4 percent). Out-of-pocket payments primarily covered spending on pharmaceuticals and dental care, while inpatient care was predominantly funded through government and compulsory sources (98 percent) (OECD and European Observatory on Health Systems and Policies 2021b). Financing system. Estonia has a centralized health care system with a single health insurance fund, the Estonian Health Insurance Fund (EHIF), which is largely funded through payroll taxes based on employer contributions of 13 percent for health benefits (OECD and European Observatory on Health Systems and Policies 2021b). The EHIF pools the majority of public funding for health care and manages the purchasing of health care services for 95 percent of the population (Habicht et al. 2018). It covers a comprehensive benefits package that includes most hospital and outpatient medical care and uses cost sharing to cover pharmaceuticals and dental care. Both public and private health care providers can hold contracts with the EHIF, although the EHIF is also allowed to selectively contract providers that are not included in the Hospital Network Development Plan (Habicht et al. 2018). Health workforce. In 2019, Estonia had an estimated approximately 3.5 physicians and 6.2 nurses per 1,000 inhabitants (OECD and European Observatory on Health Systems and Policies 2021b). Organization of the hospital system. In 2019, the number of hospital beds in Estonia was 4.5 per 1,000 population, with a trend toward a gradual decrease as services shift toward outpatient 75 care (OECD and European Observatory on Health Systems and Policies 2021b). The rate of acute care hospital discharges in Estonia was 15.4 per 100 inhabitants, while the average length of hospital stay was 7.6 days in 2014 (WHO European Region and European Health Information Gateway n.d. [a]). Despite the relatively low utilization of the health care system in Estonia, long waiting times have been linked to high levels of unmet health care needs, with approximately 15 percent of the population reporting unmet needs due to waiting times (Habicht et al. 2018). The bed occupancy rate in 2018 varied significantly across hospitals in Estonia, ranging from 45 percent to 86 percent. 8 In Estonia, about half of the total volume of specialized health care services is provided at the two largest hospitals, located in Tallinn and Tartu; the hospital network connects smaller hospitals with these two major centers. Hospitals in Estonia are divided into regional, central, general, local, specialized, rehabilitation care, and nursing care hospitals, which differ in the services they provide, their location, and their catchment area. Regional hospitals offer a full range of services, while central hospitals exclude certain services such as neurosurgery or cardiac surgery. General hospitals provide 24/7 emergency care, intensive care, and some surgical specialties, while local hospitals provide only emergency care. Ambulatory specialist care is provided by hospital outpatient departments and independent specialists, while specialized outpatient care providers can be private entrepreneurs or joint-stock companies (Habicht et al. 2018). Reforms. The Estonian Health Insurance Fund introduced the Nordic DRG–based payment system (NordDRG) for inpatient services in 2004. The DRG system was implemented to increase efficiency and is used in conjunction with other payment methods. Its use for payments has gradually increased, from 10 percent upon introduction to 70 percent in 2009 (Habicht et al. 2018). All inpatient care cases, as well as outpatient care cases involving surgical procedures, fall under the DRG system. In 2015, further reforms were implemented to optimize the number of acute care hospitals in Estonia. The Estonian Hospital Master Plan 2015 set a target to reduce the number of acute care hospitals by half and decrease the number of acute care beds by 2.5 times through reduction of the average length of stay. 9 Impact of COVID-19. As of February 2022, Estonia had a relatively high number of COVID- 19 cases, at 450 per 1,000 population; and a lower-than-average number of deaths, at two per 1,000 inhabitants (compared to European averages of 210 cases and 2.3 deaths per 1,000 population) (Our World in Data n.d.). The impact of the COVID-19 pandemic led to changes in financing and organization of hospitals, as well as in the health care workforce. To cover the costs associated with the pandemic, Estonia implemented a supplementary state budget, with state budget transfers to the EHIF covering the deficit (OECD and European Observatory on Health Systems and Policies 2021b). Elective inpatient and outpatient services were briefly suspended during the first wave of the pandemic in March–April 2020, and some elective services were postponed on an ad hoc basis during the second wave (OECD and European Observatory on Health Systems and Policies 2021b). To ensure sufficient physician availability for treating COVID-19 patients, some network hospitals established separate departments and required one doctor for every 10 COVID-19 patients. Various measures were also implemented to address workforce shortages in Estonia (OECD and European Observatory on Health Systems and Policies 2021b). Throughout the pandemic, there were no reports of shortages of hospital beds. 8 Person-Centred Integrated Hospital Master Plan in Estonia—SRSS/SC2019/140—Inception Report, prepared within DG REFORM project “Person-Centred Integrated Hospital Master Plan in Estonia” (EHMP2040), https://www.sm.ee/media/2243/download. 9 Person-Centred Integrated Hospital Master Plan in Estonia, https://www.sm.ee/media/2243/download. 76 Republic of Korea In 2019, health care expenditure in Korea accounted for 8.16 percent of GDP, or US$3,521 PPP per capita. Government spending and mandatory contributions comprised only 60 percent of total health care expenditure; 30 percent was funded through out-of-pocket payments and 10 percent through VHI. The latter is a substantial portion compared to other countries (WHO n.d. [c]). Financing system. Like many other countries, Korea has a mandatory health insurance system with a single insurer (NHIS) that pools contributor payments and purchases health care services at pre-agreed prices. The NHIS is funded through contributions (80 percent in 2020), government subsidies (14 percent), and tobacco surcharges (6 percent). Contributions come from employees (86 percent) and the self-employed (14 percent) and cover the health insurance of contributors and their dependents. The employee contribution rate as a percentage of payroll has increased over time, from 5.33 percent in 2010 to 6.86 percent in 2021. Tobacco surcharges designated for the NHIS are limited to a maximum of 65 percent of the total estimated tobacco surcharges (Korea, NHIS 2022). Currently, approximately 97 percent of Koreans are insured, and the remaining individuals are covered by a separate Medical Aid Program (Korea, NHIS 2022). Regardless of their status, all individuals are eligible for equal health care services. However, to prevent the overuse of medical services, especially at higher-level facilities, patients are required to make copayments. The share of copayments depends on the type of service (inpatient vs. outpatient), facility type (higher-level facilities require a larger copayment), and disease type (lower copayments are required for rare and/or severe diseases). There is an annual copayment ceiling that depends on the patient's income. Copayments are a significant source of revenue for the health care budget. In Korea, most health care providers are privately owned. These providers are required to offer services to the insured on previously agreed terms, subject to negotiations between the NHIS and the medical service provider. The Korean health care sector primarily relies on fee-for-service payment methods and has implemented the DRG system only to a limited extent (i.e., only for seven DRG groups such as cataract, tonsil/adenoidectomy, appendectomy, C-section, hernia surgeries, uterine surgeries, and anal surgeries). The sector also uses elements of performance-based compensation (for acute myocardial infarction, C-section, stroke in the acute phase, preventive antibiotics for surgery, hypertension, diabetes, and medicines) (Korea, NHIS 2022). Health workforce. In 2020, Korea had approximately 2.5 doctors per 1,000 inhabitants—lower than the EU average—and 8.4 nurses per 1,000 inhabitants (OECDStat n.d.). Organization of the hospital system. Despite the relatively low number of doctors, Korea has a large number of hospitals and beds, with 7.8 hospitals and 124 beds per 100,000 inhabitants in 2019. These numbers have increased over time (e.g., in 2010, there were only 5.7 hospitals with 87 beds per 100,000 inhabitants) (OECDStat n.d.). In Korea, there are several types of hospitals: clinics, hospitals, general hospitals (including dental and Oriental medicine hospitals), and tertiary hospitals (Chun et al. 2009) (managed by the Korean government) (OECD 2020). However, there are no well-defined boundaries between primary health care and inpatient care. Patients can receive their first consultation without a referral at any hospital except a tertiary hospital (in principle, it is possible to receive a consultation without a referral at a tertiary hospital, but the patient must pay the bill). This free access results in a high frequency of medical consultations: in Korea, the rate of doctor 77 consultations was 16.6 times per year, compared to the Organisation for Economic Co- operation and Development (OECD) average of 6.8 in 2017 (OECD 2020). Reforms. Over the last several decades, Korea has made significant progress in developing its health care system, resulting in universal health care coverage and one of the highest life expectancies in the world. However, there are ongoing health care reforms in the country. The NHIS recognizes the problems associated with fee-for-service payment methods and is attempting to introduce a hybrid payment method that combines DRGs, fee-for-service, and performance-based payment compensation in the inpatient sector. According to the OECD, Korea must also address the issue of high copayments (Korea’s are among the highest in the OECD) and the overuse of inpatient care (Korea has the most beds per person and the longest length of stay among OECD countries) (OECD 2016). Impact of COVID-19. As of February 2022, Korea had had a relatively successful response to the COVID-19 pandemic, with only 528 deaths per million people, compared to the European average of 2,326 deaths per million people (WHO n.d. [a]). Effective communication with the public, aggressive testing and contact tracing, and strict quarantine policies enabled the government to effectively manage outbreaks without causing unnecessary harm to the economy and health care sector (Kim et al. 2021). 78 Annex 6. Overview of ALOS differences This annex shows how large the difference is in average length of stay across DRGs that vary in complexity level. As Ukraine has only A-DRGs and, thus, does not distinguish complexity, we do not present it in this annex. We also do not have a sufficiently comparable data set to present the Koreans’ results. Table 6A-1. Difference in ALOS across Complexity Levels DRG DRG DRG DRG DRG text ALOS DRG text ALOS DRG text ALOS DRG text ALOS code code code code Croatia Estonia Moldova Romania SURGICAL G07A Appendectomy w very severe or severe 9.70 166N Appendectomy w complicated 4.90 G07 Appendectomy w very severe or 6.13 G1071 Appendicectomy w catastrophic or 4.84 CC principal diagnosis A severe CC severe CC G07B Appendectomy w/o very severe or 3.76 167 Appendectomy w/o complicated 2.00 G07B Appendectomy w/o very severe or 4.79 G1072 Appendicectomy w/o catastrophic or 4.12 severe CC principal diagnosis w/o CC severe CC severe CC I13A Procedures on humerus, tibia, fibula, 16.37 218 Lower extremity & humerus 7.60 I13A Procedures on humerus, tibia, fibula, 10.27 I1131 Humerus, tibia, fibula, and ankle 7.12 and ankle w very severe or severe CC procedures except hip, foot, and ankle w very severe or severe CC procedures w catastrophic/severe CC femur age > 17 w CC I13B Procedures on humerus, tibia, fibula, 6.70 219 Lower extremity & humerus 3.80 I13B Procedures on humerus, tibia, fibula, 8.59 I1132 Humerus, tibia, fibula, and ankle 7.31 and ankle, age > 59 w/o very heavy or procedures except hip, foot, and ankle, w/o CC procedures age > 59 w/o severe CC femur age > 17 w/o CC catastrophic/severe CC I13C Procedures on humerus, tibia, fibula, 5.20 220 1.9 I1133 Humerus, tibia, fibula, and ankle 6.08 and ankle, age < 60 w/o very heavy or Lower extrem & humer proc procedures age < 60 w/o catastrophic severe CC except hip, foot, femur age 0–17 /severe CC G11A Procedures on the anus and stoma w 11.55 G11Z Procedures on the anus and stoma 6.59 G1111 Anal and stomal procedures w 4.74 very heavy or severe CC catastrophic or severe CC G11B Procedures on the anus and stoma w/o 3.15 G1112 Anal and stomal procedures w/o 3.92 very heavy or severe CC catastrophic or severe CC F10Z Percutaneous coronary intervention w 4.42 112E PCI w myocardial infarction, w/o 4.50 F10A Percutaneous coronary intervention w 6.67 F1100 Percutaneous coronary angioplasty w 5.13 AMI CC AMI w CC AMI 112F PCI w myocardial infarction, w 7.20 F10B Percutaneous coronary intervention w 6.39 CC AMI w/o catastrophic CC F15Z Percutaneous coronary intervention, 3.11 112C PCI w/o myocardial infarction, 2.40 F15A Percutaneous coronary intervention, 2.77 F1150 Percutaneous coronary angioplasty w/o 3.07 w/o AMI, w stent insertion w/o CC w/o AMI, w stent insertion, w AMI w stent implantation catastrophic or severe CC F16Z Percutaneous coronary intervention, no 2.81 112D PCI w/o myocardial infarction, w 4.00 F16A Percutaneous coronary intervention, 2.72 F1160 Percutaneous coronary angioplasty w/o 3.37 AMI, no stent insertion CC w/o AMI, w/o stent insertion w CC AMI w/o stent implantation F15B Percutaneous coronary intervention, 4.22 w/o AMI, w stent insertion, w/o catastrophic or severe CC G02A Great procedures on the small and 24.97 148 Major small & large bowel 15.10 G02 Proceduri majore pe intestinul subtire 14.94 G1021 Major small and large bowel procedures 11.64 large intestines w very heavy CC procedures w CC A si gros cu CC catastrofale w catastrophic CC G02B Great procedures on the small and 10.21 149 Major small & large bowel 8.30 G02B Great procedures on the small and 12.14 G1022 Major small and large bowel procedures 9.18 large intestine w/o very heavy CC procedures w/o CC large intestine w/o very heavy CC w/o catastrophic CC 79 I03A Hip revision w very heavy or severe 25.37 209C Major joint secondary procedure 11.70 I03A Hip revision w very heavy or severe 8.63 I1031 Hip revision w catastrophic or severe 13.44 CC on hip CC CC I03B Hip replacement w very heavy or 12.74 209D Major joint primary procedure 7.80 I03B Hip replacement w very heavy or 8.64 I1032 Hip replacement w cat or severe CC or 8.80 severe CC or hip revision w/o very on hip w CC severe CC or hip revision w/o very hip revision w/o catastrophic/severe CC heavy or severe CC heavy or severe CC I03C Hip replacement w/o very heavy or 9.93 209E Major joint primary procedure 5.90 I1033 Hip replacement w/o catastrophic or 8.63 heavy CC on hip w/o CC severe CC H08A Laparoscopic cholecystectomy w 10.62 493 Endoscopic or laparoscopic 6.50 H08 Laparoscopic cholecystectomy w 5.97 H1061 Laparoscopic cholecystectomy w closed 4.90 closed ductus choledocus passability operation on gallbladder w CC A closed ductus choledocus passability CDE or w catastrophic or severe CC test or w very severe or severe CC test or w very severe or severe CC H08B Laparoscopic cholecystectomy, w/o a 3.19 494 Endoscopic or laparoscopic 2.50 H08B Laparoscopic cholecystectomy, w/o a 4.83 H1062 Laparoscopic cholecystectomy w/o 3.87 closed study of the passability of operation on gallbladder w/o CC closed study of the passability of closed CDE w/o catastrophic or severe ductus choledocus, w/o very severe or ductus choledocus, w/o very severe or CC severe CC severe CC O01A Cesarean delivery w very severe CC 25.35 370 Cesarean section w CC 5.00 O01 Cesarean delivery w very severe CC 6.11 O1011 Cesarean delivery w multiple 5.41 A complicating diagnoses, at least one severe O01B Cesarean delivery w severe CC 8.08 371 Cesarean section w/o CC 3.70 O01B Cesarean delivery w severe CC 4.09 O1012 Cesarean delivery w severe 4.80 complicating diagnosis O01C Cesarean delivery w/o very severe or 5.48 O1013 Cesarean delivery w moderate 4.53 severe CC complicating diagnosis O02A Vaginal delivery w an operational 11.28 O02 Vaginal delivery w an operational 3.87 O1021 Vaginal delivery w OR procedure w 4.39 procedure w very severe or severe CC A procedure w very severe or severe CC catastrophic or severe CC O02B Vaginal delivery w an operational 3.69 O02B Vaginal delivery w an operational 3.61 O1022 Vaginal delivery w OR procedure w 4.35 procedure w/o very severe or severe procedure w/o very severe or severe catastrophic or severe CC CC CC MEDICAL O60A Vaginal delivery w very severe or 11.53 372 Vaginal delivery w complicating 3.40 O60Z Vaginal delivery O60Z O3011 Vaginal delivery w multiple 4.41 severe CC diagnoses complicating diagnoses, at least one severe O60B Vaginal delivery w/o very severe or 3.87 373 Vaginal delivery w/o 2.80 O3012 Vaginal delivery w severe complicating 4.41 severe CC complicating diagnoses diagnosis O60C Single-prolific uncomplicated vaginal 3.47 374 Vaginal delivery w sterilization 10.00 O3013 Vaginal delivery w moderate 3.93 delivery w no other conditions &/or D&C complicating diagnosis 375 Vaginal delivery w OR 3.80 procedure except sterilization &/or D&C F66A Coronary blood vessel atherosclerosis 10.79 132 Atherosclerosis w CC 6.20 F66A Coronary blood vessel atherosclerosis 8.2 F3071 Coronary atherosclerosis w CC 5.44 w CC w CC F66B Coronary blood vessel atherosclerosis 3.67 133 Atherosclerosis w/o CC 2.60 F66B Coronary blood vessel atherosclerosis 8.1 F3072 Coronary atherosclerosis w/o CC 3.91 w/o CC w/o CC D63A Inflammation of the middle ear and 8.72 068 Otitis media & URI, age > 17 w 8.60 D63Z Otitis media and upper respiratory 5.02 D3041 Otitis media and URI w CC 4.15 upper respiratory tract infections w CC CC tract infections D63B Inflammation of the middle ear and 3.33 069 Otitis media & URI, age > 17 3.10 D3042 Otitis media and URI w/o CC 4.18 upper respiratory tract infections w/o w/o CC CC 80 070A Otitis media & URI, age 0–17, w 3.50 CC 070B Otitis media & URI, age 0–17, 2.60 w/o CC B70A Stroke w very heavy CC 18.21 014A Specific cerebrovascular 11.20 B70A Stroke w very heavy CC 9.11 B3111 Stroke w severe or complicating 11.80 disorders except TIA, w/o diagnosis/procedure thrombolysis, w CC B70B Stroke w heavy CC 8.33 014B Specific cerebrovascular 8.60 B70B Stroke w heavy CC 8.33 B3112 Stroke w other CC 8.57 disorders except TIA, w/o thrombolysis, w/o CC B70C Stroke w/o very heavy or heavy CC 8.23 014C Specific cerebrovascular 5.70 B70C Stroke w/o very heavy or heavy CC 7.87 B3113 Stroke w/o other CC 7.16 disorders except TIA, w thrombolysis, w/o CC B70D Stroke, fatality, or transfer to another 1.89 014D Specific cerebrovascular 6.90 B3114 Stroke, died or transferred < 5 days 1.94 acute care facility, stay < 5 days disorders except TIA, w thrombolysis, w CC E69A Bronchitis and asthma, age > 49 w CC 10.38 096 Bronchitis & asthma, age > 17 w 8.20 E69A Bronchitis and asthma, w CC 6.79 E3101 Bronchitis and asthma age > 49 w CC 6.74 CC E69B Bronchitis and asthma, age > 49, or w 4.51 097 Bronchitis & asthma, age > 17 5.60 E69B Bronchitis and asthma, or w CC 6.44 E3102 Bronchitis and asthma (age < 50 w CC) 4.60 CC w/o CC or (age > 49 w/o CC) E69C Bronchitis and asthma, age < 50 w/o 4.00 098A Bronchitis & asthma, age 0–17, 4.30 E3103 Bronchitis and asthma age < 50 w/o CC 3.85 CC w CC 098B Bronchitis & asthma, age 0–17, 2.90 w/o CC F67A Hypertension w CC 9.59 134 Hypertension 4.50 F67A Hypertension w CC 7.84 F3081 Hypertension w CC 5.39 F67B Hypertension w/o CC 4.13 F67B Hypertension w/o CC 8.01 F3082 Hypertension w/o CC 5.23 B71A Disruption of cranial and peripheral 12.83 018 Cranial & peripheral nerve 7.80 B71A Disruption of cranial and peripheral 8.10 B3121 Cranial and peripheral nerve disorders w 5.56 nerves w CC disorders w CC nerves w CC CC B71B Disorder of cranial and peripheral 4.85 019 Cranial & peripheral nerve 4.60 B71B Disorder of cranial and peripheral 7.79 B3122 Cranial and peripheral nerve disorders 4.76 nerves w/o CC disorders w/o CC nerves w/o CC w/o CC I69A Bone diseases and specific arthropathy, 14.78 244 Bone diseases & specific 9.70 I69A Bone diseases and specific 7.96 I3091 Bone diseases & specific arthropathies 6.25 age > 74 w very severe or severe CC arthropathies w CC arthropathy, w very severe or severe age > 74 w catastrophic/severe CC CC I69B Bone diseases and specific arthropathy, 4.85 245 Bone diseases & specific 6.90 I69B Bone diseases and specific 7.89 I3092 Bone diseases, specific arthropathies 4.74 age > 74, w very severe or severe CC arthropathies w/o CC arthropathy, w very severe or severe age > 74 w/o catastrophic/severe CC CC I69C Bone diseases and specific arthropathy, 3.60 I3093 Bone diseases and specific arthropathies 3.59 age < 75 w/o very severe or severe CC age < 75 F62A Heart failure and shock w very severe 13.86 127 Heart failure & shock 8.20 F62A Heart failure and shock w very severe 7.82 F3031 Heart failure and shock w catastrophic 7.07 CC CC CC F62B Heart failure and shock w/o very 6.07 F62B Heart failure and shock w/o very 8.01 F3032 Heart failure and shock w/o catastrophic 6.19 severe CC severe CC CC J62A Malignant breast disease, age > 69 w 16.39 366 Malignancy, female reproductive 7.70 J62A Malignant breast disease, w CC, or w 2.92 J3021 Malignant breast disorders age > 69 w 4.43 CC, or w very severe or severe CC system w CC very severe or severe CC CC J62B Malignant breast disease, age > 69 w/o 4.16 367 Malignancy, female reproductive 3.40 J62B Malignant breast disease, w/o CC, or 4.12 J3022 Malignant breast disorders (age < 70 w 3.44 CC, or w/o very severe or severe CC system w/o CC w/o very severe or severe CC CC) or (age > 69 w/o CC) 81 I68A Nonsurgical spinal disorders w CC 11.37 009 Spinal disorders & injuries 9.00 I68A Nonsurgical spinal disorders w CC 8.19 B3021 Spinal cord conditions w or w/o OR 7.35 procedures w catastrophic or severe CC I68B Nonsurgical spinal disorders w/o CC 4.03 I68B Nonsurgical spinal disorders w/o CC 8.23 B3022 Spinal cord conditions w or w/o OR 5.94 procedures w/o catastrophic or severe CC H62A Pancreatic disorders, in addition to 15.14 204 Disorders of pancreas except 6.60 H62 Pancreatic disorders, in addition to 7.94 H3031 Disorders of pancreas except for 6.87 malignant disease, w very severe or malignancy A malignant disease, w very severe or malignancy w catastrophic/severe CC severe CC severe CC H62B Pancreatic disorders, excluding 7.28 H62B Pancreatic disorders, excluding 6.64 H3032 Disorders of pancreas except for 5.59 malignant disease, w/o very severe or malignant disease, w/o very severe or malignancy w/o catastrophic/severe CC severe CC severe CC U61A Schizophrenic disorders—forced 20.52 430A Schizophrenia, age < 30 years 15.60 U61Z Schizophrenic disorders U3021 Schizophrenia disorders w mental health 16.44 treatment legal status U61B Schizophrenic disorders—self- 11.08 430B Schizophrenia, age 30–59 years 14.90 U3022 Schizophrenia disorders w/o mental 15.51 treatment health legal status 430C Schizophrenia, > 59 years 15.80 G60A Malignant disease of the digestive 10.82 172 Digestive malignancy w CC 7.50 G60 Malignant disease of the digestive 6.48 G3011 Digestive malignancy w catastrophic or 4.11 system w very severe or severe CC A system w very severe or severe CC severe CC G60B Malignant disease of the digestive 3.38 173 Digestive malignancy w/o CC 4.60 G60B Malignant disease of the digestive 4.54 G3012 Digestive malignancy w/o catastrophic 3.42 system w/o very severe or severe CC system w/o very severe or severe CC or severe CC K60A Diabetes w very severe or severe CC 9.73 294 Diabetes age > 35 8.40 K60 Diabetes w very severe or severe CC 8.87 K3011 Diabetes w catastrophic or severe CC 6.40 A K60B Diabetes w/o very severe or severe CC 5.64 295 Diabetes age 0–35 4.00 K60B Diabetes w/o very severe or severe 6.89 K3012 Diabetes w/o catastrophic or severe CC 5.14 CC Source: Authors, based on country data Notes: CC = Complications and comorbidities; CDE = Common duct exploration; D&C = Dilation and curettage; OR = Operating room; PCI = Percutaneous coronary intervention SD = Standard deviation; URI = Upper respiratory infection; DRG = Diagnosis-related group; ALOS = Average length of stay; w = With; w/o = Without; TIA = transient ischemic attack; AMI = Acute myocardial infarction. 82 Annex 7. Definition of the Case Types Surgical partition: ● C-section o Defined by main or secondary diagnosis OR procedure code o Main or secondary diagnosis O82 Single delivery by cesarean section (and subcodes) or O842 Multiple delivery, all by cesarean section OR o Procedure code of C-section (according to surgical classification in use) ● Appendectomy o Defined by main diagnosis AND procedure code o Main diagnosis K35–K38 (Diseases of appendix) and subcodes (when used) AND o Procedure codes of Appendectomy (according to surgical classification in use) Medical partition: ● Vaginal delivery o Defined by main or secondary diagnosis OR DRG o Main or secondary diagnosis O80, O81, O83, and O84 (exclude O842) and subcodes OR o DRG of Vaginal Delivery (according to DRG system in use) ● Acute myocardial infarction (only medical) o Defined by main diagnosis AND exclude bypass procedure o Main diagnosis I21 (Acute myocardial infarction) and I22 (Subsequent myocardial infarction) and subcodes AND o No bypass procedure (according to surgical classification in use) 83 This study presents a comparative analysis of inpatient care data from six countries (Ukraine, Moldova, Romania, Croatia, Estonia, and the Republic of Korea). It uses diagnosis-related groups (DRGs) as a framework for benchmarking. The primary aim of the study is to propose methods for comparing hospital outputs across countries that use different systems for coding hospital activities. Additionally, this study seeks to identify best practices in coding and hospital performance indicators, which could be of interest to countries and hospitals included in this analysis and beyond. It compares hospital outputs from 2021 at both country and hospital levels, using indicators such as the frequency of cases within different adjacent DRGs (A-DRGs) and major diagnostic categories, length of stay, coding activities, and case complexity. It also employs country-level analysis to determine the scope of hospital-level data collection and analysis. The study draws upon data from 30 hospitals of different types across all six countries. Its findings provide insights and identify areas for further investigation that can orient and guide hospital reforms. Such reforms may aim at ensuring better quality and cost-effectiveness of care, as well as patient access to hospital services. ABOUT THIS SERIES: This series is produced by the Health, Nutrition, and Population Global Practice of the World Bank. The papers in this series aim to provide a vehicle for publishing preliminary results on HNP topics to encourage discussion and debate. The findings, interpretations, and conclusions expressed in this paper are entirely those of the author(s) and should not be attributed in any manner to the World Bank, to its affiliated organizations or to members of its Board of Executive Directors or the countries they represent. Citation and the use of material presented in this series should take into account this provisional character. For free copies of papers in this series please contact the individual author/s whose name appears on the paper. Enquiries about the series and submissions should be made directly to the Editor Jung-Hwan Choi (jchoi@worldbank.org) or HNP Advisory Service (askhnp@worldbank.org). For more information, see also www.worldbank.org/hnppublications. 1818 H Street, NW Washington, DC USA 20433 Telephone: 202 473 1000 Facsimile: 202 477 6391 Internet: www.worldbank.org E-mail: feedback@worldbank.org