CHILDREN AND FAMILIES The RAND Corporation is a nonprofit institution that helps improve policy and EDUCATION AND THE ARTS decisionmaking through research and analysis. ENERGY AND ENVIRONMENT
HEALTH AND HEALTH CARE This electronic document was made available from www.rand.org as a public service
INFRASTRUCTURE AND of the RAND Corporation. TRANSPORTATION
INTERNATIONAL AFFAIRS LAW AND BUSINESS Skip all front matter: Jump to Page 16 NATIONAL SECURITY
POPULATION AND AGING
PUBLIC SAFETY Support RAND
SCIENCE AND TECHNOLOGY Browse Reports & Bookstore
TERRORISM AND Make a charitable contribution HOMELAND SECURITY
For More Information Visit RAND at www.rand.org Explore the Pardee RAND Graduate School View document details
Limited Electronic Distribution Rights This document and trademark(s) contained herein are protected by law as indicated in a notice appearing later in this work. This electronic representation of RAND intellectual property is provided for non- commercial use only. Unauthorized posting of RAND electronic documents to a non-RAND website is prohibited. RAND electronic documents are protected under copyright law. Permission is required from RAND to reproduce, or reuse in another form, any of our research documents for commercial use. For information on reprint and linking permissions, please see RAND Permissions. This product is part of the Pardee RAND Graduate School (PRGS) dissertation series. PRGS dissertations are produced by graduate fellows of the Pardee RAND Graduate School, the world’s leading producer of Ph.D.’s in policy analysis. The dissertation has been supervised, reviewed, and approved by the graduate fellow’s faculty committee. Dissertation
China’s Health Insurance Reform and Disparities in Healthcare Utilization and Costs A Longitudinal Analysis
Henu Zhao
C O R P O R A T I O N Dissertation
China’s Health Insurance Reform and Disparities in Healthcare Utilization and Costs A Longitudinal Analysis
Henu Zhao
This document was submitted as a dissertation in October 2014 in partial fulfillment of the requirements of the doctoral degree in public policy analysis at the Pardee RAND Graduate School. The faculty committee that supervised and approved the dissertation consisted of Hao Yu (Chair), Emmett Keeler, and Gema Zamarro.
PARDEE RAND GRADUATE SCHOOL The Pardee RAND Graduate School dissertation series reproduces dissertations that have been approved by the student’s dissertation committee.
The RAND Corporation is a nonprofit institution that helps improve policy and decisionmaking through research and analysis. RAND’s publications do not necessarily reflect the opinions of its research clients and sponsors.
R® is a registered trademark.
Permission is given to duplicate this document for personal use only, as long as it is unaltered and complete. Copies may not be duplicated for commercial purposes. Unauthorized posting of RAND documents to a non-RAND website is prohibited. RAND documents are protected under copyright law. For information on reprint and linking permissions, please visit the RAND permissions page (http://www.rand.org/publications/permissions.html).
Published 2015 by the RAND Corporation 1776 Main Street, P.O. Box 2138, Santa Monica, CA 90407-2138 1200 South Hayes Street, Arlington, VA 22202-5050 4570 Fifth Avenue, Suite 600, Pittsburgh, PA 15213-2665 RAND URL: http://www.rand.org/ To order RAND documents or to obtain additional information, contact Distribution Services: Telephone: (310) 451-7002; Fax: (310) 451-6915; Email: [email protected] Table of Contents
Tables ...... v Figures ...... ix Abstract ...... xi Acknowledgements ...... xiii Chapter 1 Introduction ...... 1 Chapter 2 Background ...... 3 2.1 Health insurance reform in China ...... 3 2.1.1 Collapse of health insurance schemes in the 1970s and 1980s ...... 4 2.1.2 Early efforts in the 1980s and early 1990s ...... 5 2.1.3 Health insurance reform since the late 1990s ...... 6 2.1.4 Healthcare reform after 2009 ...... 9 2.2 Three Major Health Insurance Schemes ...... 10 2.2.1 The Basic Medical Insurance for Urban Employees ...... 10 2.2.2 The Basic Medical Insurance for Urban Residents...... 11 2.2.3 The New Rural Cooperative Medical Insurance ...... 13 2.3 Trends and Current Status of Healthcare Disparities ...... 13 Chapter 3 Literature Review and Study Objectives ...... 19 3.1 Existing Research ...... 19 3.1.1 Literature on Rural–Urban Disparities in Healthcare Utilization ...... 19 3.1.2 Literature on Disparities in Out‐of‐Pocket Expenditure and Healthcare Costs ...... 21 3.1.3 Literature on Disparities in Health Insurance Coverage ...... 22 3.1.4 Methodological Issues ...... 22 3.2 Gap in the Existing Literature ...... 26 3.3 Objectives and Research Questions ...... 27 Chapter 4 Study Design ...... 28 4.1 Data ...... 28 4.2 Study Periods ...... 30 4.3 Conceptual Model and Variable Selection ...... 30 4.3.1 Dependent Variables ...... 31 4.3.2 Independent Variables ...... 33 4.4 Analytic Approach ...... 38 4.4.1 Difference‐in‐Differences Analysis with Multiple Groups and Multiple Time Periods ...... 38 4.4.2 Multivariate Regression for the Variables that do not meet the Assumption of Parallel Trends ...... 44 4.5 Sensitivity analysis ...... 46 4.5.1 Controlling for Insurance Status ...... 46 4.5.2 Dropping the Richest Province or the Poorest Province ...... 4 6
iii 4.5.3 Including Interaction Terms with Household Income ...... 4 7 4.5.4 DID Analysis Results for Variables in Which Parallel Trends did not Hold ...... 47 Chapter 5 Results: Disparities in Healthcare Utilization ...... 48 5.1 Descriptive Analysis ...... 48 5.2 DID Analysis for Formal Care Utilization and Outpatient Utilization ...... 51 5.3 Multivariate Analysis Controlling for Existing Trends for Inpatient Utilization ...... 57 5.4 Sensitivity Analysis ...... 64 5.4.1 Controlling for Insurance Status ...... 64 5.4.2 Dropping the Richest Province or the Poorest Province ...... 7 1 5.4.3 Including Interaction Terms with Household Income ...... 8 0 5.4.4 DID Analysis for Inpatient Care ...... 84 5.5 Summary of Findings ...... 85 Chapter 6 Results: Disparities in healthcare costs ...... 88 6.1 Descriptive Analysis ...... 88 6.2 Multivariate Analysis Controlling for Existing Trends ...... 91 6.3 Sensitivity Analysis ...... 103 6.3.1 controlling for health insurance status ...... 103 6.3.2 dropping the richest province or the poorest province...... 107 6.3.3. Including interaction terms with household income ...... 116 6.3.4 DID analysis results for cost variables ...... 131 6.4 Summary of Findings ...... 133 Chapter 7 Conclusion, Discussion, and Policy Implications ...... 135 7.1 Conclusion ...... 135 7.2 Discussion ...... 137 7.2.1 Comparing With the Published Research ...... 137 7.2.2 Strengths ...... 138 7.2.3 Limitations ...... 139 7.2.4 Future Directions ...... 140 7.3 Policy Implications ...... 140 Appendix ...... 143 Reference ...... 145
iv Tables
Table 4.1 Sample Size by Rural and Urban Residences and Registrations ...... 29
Table 4.2 Descriptive Statistics of Independent Variables by Rural and Urban Residences and Registrations ...... 37
Table 4.3 Results of DID Analysis Using 1993 and 1997 Waves for Healthcare Utilization ...... 42
Table 4.4 Results of DID Analysis Using 1993 and 1997 Waves for Healthcare Costs ...... 44
Table 5.1 DID Analysis Results for Formal Care Utilization and Outpatient Utilization ...... 54
Table 5.2 Test Results for DID Analysis of Formal Care Utilization and Outpatient Utilization ...... 55
Table 5.3 Multivariate Analysis Results for Inpatient Care Utilization ...... 59
Table 5.4 Test Results of Disparities for Inpatient Care Utilization ...... 60
Table 5.5 Test Results of Change in Disparities for Inpatient Care Utilization ...... 62
Table 5.6 DID Analysis Results of Formal Care and Outpatient Utilization (Controlling for Insurance Status) ...... 65
Table 5.7 Test Results for DID Analysis of Healthcare Utilization (Controlling for Insurance Status) ...... 66
Table 5.8 Multivariate Analysis Results for Inpatient Care Utilization (Controlling for Insurance Status) ...... 67
Table 5.9 Test Results of Disparities for Inpatient Care Utilization (Controlling for Insurance Status) ...... 69
Table 5.10 Test Results of Change in Disparities for Inpatient Care Utilization (Controlling for Insurance Status) ...... 70
Table 5.11 DID Analysis Results for Formal Care and Outpatient Utilization (Dropping the Richest Province) ...... 73
Table 5.12 Test Results for Formal Care and Outpatient Utilization (Dropping the Richest Province) ...... 74
Table 5.13 DID Analysis Results for Formal Care and Outpatient Utilization (Dropping the Poorest Province) ...... 75
Table 5.14 Test Results for Formal Care and Outpatient Utilization (Dropping the Poorest Province) ...... 76
v Table 5.15 Multivariate Analysis Results for Inpatient Utilization (Dropping the Richest/Poorest Province) ...... 77
Table 5.16 Test Results of Disparities in Inpatient Utilization (Dropping the Richest/poorest Province) ...... 78
Table 5.17 Test Results of Change in Disparities for Inpatient Care Utilization (Dropping the Richest/poorest Province) ...... 79
Table 5.18 DID Analysis Results for Formal Care and Outpatient Utilizations (Including Interaction Term with Household Income) ...... 82
Table 5.19 Test Results for Formal Care and Outpatient Utilizations (Including Interaction Term with Household Income) ...... 83
Table 5.20 DID Analysis Results for Inpatient Care Utilization ...... 84
Table 5.21 Test Results for Inpatient Care Utilization (DID Analysis) ...... 85
Table 6.1 Multivariate Analysis Results for OOP Exceeding Certain Percentage of Household Income ...... 93
Table 6.2 Multivariate Analysis Results for Total Healthcare Costs ...... 95
Table 6.3 Test Results of Disparities for OOP Exceeding Certain Percentage of Household Income ...... 100
Table 6.4 Test Results of Changes in Disparities for OOP Exceeding Certain Percentage of Household Income ...... 101
Table 6.5 Bootstrap Results for Disparities in Total Health Costs ...... 103
Table 6.6 Multi‐variate Analysis Results for OOP Exceeding Certain Percentage of Household Income (Controlling for Insurance) ...... 104
Table 6.7 Test Results of Disparities for OOP Exceeding Certain Percentage of Household Income (Controlling for Insurance) ...... 105
Table 6.8 Test Results of Changes in Disparities for OOP Exceeding Certain Percentage of Household Income (Controlling for Insurance) ...... 106
Table 6.9 Bootstrap Results for Disparities in Total Health Cost (Controlling for Insurance) ...... 107
Table 6.10 Multi‐variate Analysis Results for OOP Exceeding Certain Percentage of Household Income (Dropping the Richest Province) ...... 109
Table 6.11 Test Results of Disparities for OOP Exceeding Certain Percentage of Household Income (Dropping the Richest Province) ...... 110
vi Table 6.12 Test Results of Changes in Disparities for OOP Exceeding Certain Percentage of Household Income (Dropping the Richest Province) ...... 111
Table 6.13 Multi‐variate Analysis Results for OOP Exceeding Certain Percentage of Household Income (Dropping the Poorest Province) ...... 112
Table 6.14 Test Results of Disparities for OOP Exceeding Certain Percentage of Household Income (Dropping the Poorest Province) ...... 113
Table 6.15 Test Results of Changes in Disparities for OOP Exceeding Certain Percentage of Household Income (Dropping the Poorest Province) ...... 114
Table 6.16 Bootstrap Results for Disparities in Total Health Costs (Dropping the Richest Province) ...... 115
Table 6.17 Bootstrap Results for Disparities in Total Health Cost (Dropping the Poorest Province) ...... 116
Table 6.18 Multi‐variate Analysis Results for OOP Exceeding Certain Percentage of Household Income (Low‐income Families) ...... 118
Table 6.19 Test Results of Disparities for OOP Exceeding Certain Percentage of Household Income (Low‐income Families) ...... 119
Table 6.20 Test Results of Changes in Disparities for OOP Exceeding Certain Percentage of Household Income (Low‐income Families) ...... 120
Table 6.21 Multi‐variate Analysis Results for OOP Exceeding Certain Percentage of Household Income (Medium‐income Families) ...... 122
Table 6.22 Test Results of Disparities for OOP Exceeding Certain Percentage of Household Income (Medium‐income Families) ...... 123
Table 6.23 Test Results of Changes in Disparities for OOP Exceeding Certain Percentage of Household Income (Medium‐income Families) ...... 124
Table 6.24 Multi‐variate Analysis Results for OOP Exceeding Certain Percentage of Household Income (High‐income Families) ...... 126
Table 6.25 Test Results of Disparities for OOP Exceeding Certain Percentage of Household Income (High‐income Families) ...... 127
Table 6.26 Test Results of Changes in Disparities for OOP Exceeding Certain Percentage of Household Income (High‐income Families) ...... 128
Table 6.27 Bootstrap Results for Disparities in Total Health Costs (Low‐income Families) ...... 129
Table 6.28 Bootstrap Results for Disparities in Total Health Costs (Medium‐income Families) ...... 130
vii Table 6.29 Bootstrap Results for Disparities in Total Health Costs (High‐income Families) ...... 130
Table 6.30 DID Analysis Results for OOP Exceeding Certain Percentage of Household Income ...... 132
Table 6.31 Test Results for OOP Exceeding Certain Percentage of Household Income (DID Analysis) ...... 132
Table 6.32 Bootstrap Results for Disparities in Total Health Costs (DID Analysis) ...... 133
viii Figures Figure 2.1 Health Insurance Coverage in Urban and Rural Areas in China, Selected Years 1993‐2008 ...... 15
Figure 2.2 Health Service Utilization in Urban and Rural Areas in China (2003) ...... 16
Figure 2.3 Healthcare Spending in China, by Source and Year ...... 17
Figure 2.4 Per Capita Out‐of‐Pocket Health Expenses as a Percentage of Income ...... 18
Figure 4.1 Updated Structure of Anderson Model ...... 31
Figure 5.1 Probability of Formal Care Utilization in 4 Weeks by Rural and Urban Residences and Registrations ...... 48
Figure 5.2 Probability of Outpatient Care Utilization in 4 Weeks by Rural and Urban Residences and Registrations ...... 49
Figure 5.3 Probability of Inpatient Care Utilization in 4 Weeks by Rural and Urban Residences and Registrations ...... 50
Figure 5.4 Predicted Probability of Formal Care Utilization in 4 Weeks by Rural and Urban Residences and Registrations ...... 56
Figure 5.5 Predicted Probability of Outpatient Care Utilization in 4 Weeks by Rural and Urban Residences and Registrations ...... 57
Figure 5.6 Predicted Probability of Inpatient Care Utilization in 4 Weeks by Rural and Urban Residences and Registrations ...... 63
Figure 6.1 Probability of Having Out‐of‐pocket Medical Expense Exceeding 20% of Household Income by Rural and Urban Residences and Registrations ...... 8 9
Figure 6.2 Probability of Having Out‐of‐pocket Medical Expense Exceeding 40% Household Income by Rural or Urban Residences and Registrations ...... 90
Figure 6.3 Total Healthcare Costs by Rural and Urban Residences and Registrations ...... 91
Figure 6.4 Predicted Probability of Having OOP Exceeding 20% of Household Income by Rural and Urban Residences and Registrations ...... 97
Figure 6.5 Predicted Probability of Having OOP Exceeding 40% of Household Income by Rural and Urban Residences and Registrations ...... 98
Figure 6.6 Predicted Total Healthcare Costs by Rural and Urban Residences and Registrations ...... 98
ix
Abstract China’s economic success during the past 30 years was not mirrored in its health care system. As a result, the rural‐urban disparities in health insurance coverage and the related health care areas became prominent. Since the late 1990s, China has been expanding insurance coverage, in order to provide accessible and affordable health care to all residents. My study analyzes whether the insurance expansion reduces rural‐urban disparities in terms of health care utilization and financial protection. To my knowledge, this is the first study to address the disparity issue by examining China’s health care reform policies over an extended 18‐year period (1993‐2011). It is also the first study to address the dynamic phenomenon of rural‐urban migration during the study period by separating the study group into 4 subgroups in terms of respondents in residential areas versus household registration type.
Drawing on seven waves of data from the China Health and Nutrition Survey and applying multivariate analysis techniques, such as difference‐in‐difference analysis and generalized linear model, I find that rural‐urban disparities in formal care and outpatient utilization were significantly reduced by the expanded health insurance coverage in rural area in 2003. The rural‐urban disparity in total health costs is also significantly reduced.
However, no evidence shows that the policy changes in health insurance coverage had impact on disparities in inpatient utilization or having high out‐of‐pocket payments. By conducting several sets of sensitivity analyses, my study also finds that the expanded health insurance coverage impacted richer province more than poorer provinces, and impact high‐income families more than medium‐ and low‐income families.
xi The study findings have important policy implications for China’s ongoing health care reform. First, China’s policy makers should provide better health care coverage and more health care resources to rural areas to further reduce the rural‐urban disparity.
Second, since prior policy changes affected rich province more than poor province, new policy should target specifically poor provinces. Third, given the finding that the positive impact on health care utilization of policy change in 2003 happening mainly in high‐income groups, new policy change should focus more on medium‐ and low‐income group.
xii Acknowledgements I am grateful for the support provided by my wonderful dissertation committee: Dr.
Hao Yu, Dr. Gema Zamarro, and Dr. Emmett Keeler. The successful completion of this dissertation was a consequence of their excellent guidance. I am especially thankful for mentorship of my Committee Chair, Hao. His timely feedbacks on our weekly meetings were crucial to keep me on the right track. I would also like to thank Gema and Emmett for their insightful and constructive advices on the policy context and methodological issues. I also want to thank my outside reader Teh‐wei Hu, Professor Emeritus of Health Economics,
University of California, Berkeley, for his helpful and responsive comments on my dissertation.
I would also like to thank my research mentor Nelson Lim. He taught me how to do research and how to write, and provided me with advices and encouragement during my dissertation work. I would also like to thank the PRGS faculty, staff and students for their help during my dissertation writing.
The dissertation would not have been possible without the generous financial support provided by the Rosenfeld Dissertation Award.
Lastly, I would like to extend special thanks to my parents for their trust and encouragement, and to my husband, Yong Fu, for his love and support.
xiii
Chapter 1 Introduction China experienced rapid economic growth in the past two decades, benefiting many sectors of the economy. However, the economic success was not mirrored in the healthcare system. Instead, the transition from a centrally planned economy to a market‐oriented economy has caused problems in the public health arena. For example, after the economic reforms started in 1978, the existing health insurance providers faced increased operational challenges, and as a result, many residents lacked any form of health insurance.
The condition was especially troublesome in rural areas, revealing sharp rural‐urban disparities in health insurance coverage and related healthcare services and costs. Since the late 1990s, there have been attempts to expand public health insurance coverage to both rural and urban residents in order to provide accessible and affordable healthcare to all residents. Another goal of the healthcare reforms was to provide healthcare to the poor and disadvantaged populations. As of the end of 2011, three health insurance programs, called schemes, were established, covering most of the rural and urban residents with some form of health insurance. However, the performance of the current health insurance schemes has not been well examined. Mixed findings have been presented regarding this issue. My dissertation focuses on the role of health insurance in reducing the rural‐urban disparities in terms of healthcare utilization and financial protection, in the context of the current health insurance schemes.
The dissertation is organized as follows: Chapter 2 provides the background of the policy change. The chapter briefly reviews the history of the Chinese health insurance system reform, including the collapse and re‐establishment of the systems. I also provide
1 statistics of the trends and current status of rural‐urban healthcare disparities. Chapter 3
reviews existing literature on the topic of rural‐urban healthcare disparities and
summarizes the research questions. Chapter 4 presents the study design, including data used, conceptual framework, and analytical approach. Chapters 5 and 6 present the results of the study. In Chapter 7, I conclude the study and present policy implications.
2 Chapter 2 Background The great economic reform in China brought changes to all areas of the economy,
including the healthcare system. Unfortunately, as a result, many residents lost health
insurance coverage. The existing health insurance schemes experienced difficulties in
providing sufficient healthcare to insured residents. The cooperative medical scheme (CMS)
providing rural health insurance experienced the greatest damage. In response to the
emerging problems in its healthcare system, China has made numerous attempts to rebuild universal coverage system since the late 1990s. Through decades of effort, the Chinese government has developed three systems, in both urban and rural areas, which provide coverage for more than 90% of the population. During the launch of each new health insurance scheme, the government also proposed other measures to provide more healthcare resources to the targeted population. These measures work together with the health insurance systems to provide sufficient and affordable healthcare to all residents.
Although there has been great progress, the health insurance system is far from perfect.
The health insurance reform is still underway, and the effect of the expanded insurance coverage in China is still under debate.
2.1 Health insurance reform in China In this section, I review the history of health insurance reform in China. The health
insurance system collapsed in the late 1970s, and a great number of residents left
uninsured. Starting from the late 1990s, the government established three new health
insurance systems in both rural and urban areas. In 2009, the government started a new
round of healthcare reform. In the new round of reform, the major goal was to provide
3 universal coverage to all residents, and to target on disadvantage population to improve the healthcare service for them and reduce disparities.
2.1.1 Collapse of health insurance schemes in the 1970s and 1980s Since the late 1970s, the Chinese economic reforms have led to a period of prosperity. However, the economic success was not mirrored in the healthcare system.
Instead, the economic transition caused problems in the public health arena.
Prior to the economic reforms, there were three basic forms of insurance, which covered almost all Chinese citizens. The Government Insurance Scheme (GIS) covered government employees. The Labor Insurance Scheme (LIS) covered state‐owned enterprise
(SOE) workers. Finally, the cooperative medical scheme (CMS) covered the rural agricultural workers. The economic reforms brought changes to the healthcare sector, weakening all three forms of insurance to some extent. First, the government‐run hospitals under the GIS experienced financial difficulties and thus were hard pressed to provide sufficient healthcare service to those insured under GIS. One reason for the financial crisis was that the economic reforms led to relaxation of price controls, and as a result, the costs incurred by the government‐run hospitals increased. Another reason is that the government contributed less to public hospitals: Government contributions shrank from 50% in the 1980s to less than 10% in 2000 (Wang 2004). Second, during the reform, financial autonomy was granted to the SOEs. As a result, a large number of SOEs closed, and many employees lost their jobs. Thus, the number of those insured by the LIS was reduced. Even those who kept their jobs found that their SOE employers faced difficulties in financing health insurance for workers (Li 2008). Finally, in the rural areas, the basic production unit
4 became households as the collective farms were dismantled. The CMS also collapsed with
this change. In the 1990s, the vast majority of the rural population lacked any form of
health insurance coverage (Hsiao 1984; Liu 2004).
As mentioned, all three major health insurance systems experienced damages as a
result of the changes brought by the economic reforms, and among them, the rural health insurance scheme CMS faced the biggest challenge. By 1998, the percentage of rural residents with any form of health insurance coverage had dropped to 13%, compared to 56% for residents covered in urban areas (China Ministry of Health, 2004). As the urban‐rural gap widened, the urban‐rural disparity in health insurance started to draw more attention.
2.1.2 Early efforts in the 1980s and early 1990s Before the major health reforms began in the late 1990s, there had been attempts to
improve the existing health insurance systems. Even since the 1980s, actions had been
taken in urban areas to relieve the financial burden on the health insurance systems. By
introducing demand‐ and supply‐side cost sharing, the attempts in the 1980s focused on
reducing costs. These actions curbed the rapid healthcare cost growth, but they were not
able to solve the fundamental financial problems (Liu 2002). Beginning in the early 1990s,
the government introduced more actions to increase the level of risk pooling. In 1995, the
government introduced a new model combining individual responsibility and social
protection with city‐wide risk pooling. However, pilot programs of the new system were
launched in only two cities and were not spread nationwide until the late 1990s.
In rural areas, debate and research has focused on how to maintain the collapsing
corporative insurance scheme from the 1980s and 1990s. The central government’s effort
5 mainly focused on urban area; the local governments were advised to develop and
complete the current CMS systems based on local economic conditions. However, the local
actions only slightly increased the health insurance coverage in rural areas. Most of the
coverage concentrated only on developed provinces and cities, such as Shanghai, Jiangsu,
Guangdong, and Shandong. By the end of 1990s, most of the rural residents were left
uninsured.
2.1.3 Health insurance reform since the late 1990s In response to the emerging problems in its healthcare system, China has made
numerous attempts to rebuild universal coverage since the late 1990s. The goal of
universal coverage is to provide safe, effective, convenient, and affordable basic medical
services to all urban and rural residents (State Council, 2009). One of the most important
components of universal coverage is health insurance. Before this goal of universal
coverage was officially introduced in 2009 with the Chinese government’s announcement
of the blueprint for health system reform, health insurance reforms in both urban and rural areas had resulted in greater health insurance coverage. Three major health insurance schemes were established. The Urban Employees Basic Medical Insurance was launched in
urban areas in 1998, and the Urban Residents Basic Medical Insurance was launched in
2007. In rural areas, the New Rural Cooperative Medical Insurance (NRCM) was
established in 2003. In 2008, the two urban health insurance schemes covered about 65%
of urban residents, and the rural scheme covered about 90% of rural residents (National
Health Services Survey, 2008). The three major health insurance schemes are discussed in
detail in the next section.
6 The expanded health insurance coverage provided residents with more financial protection and encouraged residents to use healthcare when needed. However, the utilization of healthcare was also subjected to medical resources available. Instead of only providing health insurance coverage to residents, the healthcare reform was a comprehensive system with other measures and actions. These measures and actions worked together with health insurance expansion, providing residents with more healthcare resources and granting them adequate healthcare access.
First, the medical service system with basic facilities was constructed in rural areas.
In 2003, together with the launch of NRCM, the State Council announced other measures designed to rebuild the rural medical system (State Council, 2002). One of the measures was to construct the medical service system with basic facilities. In order to achieve this goal, central and local governments increased their financial support to the medical system each year. From 2003 to 2010, the increased funding was partially used on the construction of the medical system. Local governments at the county level were responsible for the operational cost of the local medical facilities. The central government and local governments at the province level provided undeveloped areas with subsidies for infrastructure construction.
Second, a medical assistance program was established in both rural and urban areas.
In rural areas, the program was launched in 2003. The program was to provide financial assistance to low‐income households. The assistance could either be used to treat catastrophic disease or be used as premiums to join the local NRCM. Government subsidies for the program have been increasing since the program was launched. In urban areas, the
7 program was launched in 2005. The targeted populations were (a) urban residents living
below the poverty line who did not join the Urban Residents Basic Medical Insurance; and
(b) urban residents who joined the URBMI but were still carrying heavy financial burdens.
The program was designed and funded by local governments. The central government also
provided assistance through government transfers.
Third, training of medical professionals was enhanced in rural areas. In its 2002 document No. 13, the State Council announced measures to improve the quality of medical professionals in rural areas. In post‐secondary medical schools, the Council introduced a 5‐ year program after middle school and a 3‐year program after high school, in an effort to produce more medical professionals, especially for rural areas. Medical graduates and retired medical professionals from urban areas were encouraged to go back to work in rural areas (State Council, 2002). As a reflection of ongoing progress, measures to improve education and training of medical professional were introduced again in a new round of health reform (State Council, 2009). Healthcare workers were encouraged to attend formal education programs and obtain official licenses. The training of general practitioners for rural areas was included in the Ministry of Education 2010 work plan. The government provided the training costs (Meng and Tang 2010).
Finally, the government undertook other actions to refine the whole medical system, such as regulation of drug policy, allocation of medical funding, and strengthening of administration and supervision system. All the measures worked as a whole to improve the medical service for both rural and urban areas.
8 2.1.4 Healthcare reform after 2009 As mentioned in the previous section, the goal of universal coverage was brought up by the State Council in 2009. The goal was published in the Opinions on Deepening the
Reform of the Healthcare System (State Council, 2009), which marked a new era of health care reform in China. In this round of healthcare reform, the State Council set up the goal of the universal coverage for the first time. It was also the first time for the Chinese governments to break the urban‐rural dichotomy and to provide equivalent public healthcare service to both urban and rural residents.
In order to achieve the goal of universal coverage, all three existing health insurance programs were to be improved. In addition to extending insurance coverage to the uninsured population, the benefit coverage of the insured was to be increased and expanded to cover catastrophic illnesses and outpatient visits. Another goal of the new round of health insurance reform was to provide better healthcare coverage to vulnerable population, such as rural residents, low‐income families, unemployed former SOE employees, senior population, the retired, the disabled and children. The rural‐urban gap of benefit coverage was expected to be closed, and the medical assistant programs were going to be strengthened.
In addition to improving the health insurance system, the State Council also launched other initiatives to change the health care system (State Council, 2009). The first was to provide equivalent public healthcare service to both rural and urban residents. The public healthcare service included preventative care, healthcare education, as well as health service for women and children. The second was to establish basic drug supply system. In order to ensure the supply of affordable basic drugs, the central government
9 established a list of essential drugs, and guaranteed the supply of the listed drugs to all levels of medical facilities. Moreover, the health insurance programs provided more coverage for these basic drugs. The third was to strengthen the grass root level medical service system. In rural areas, a comprehensive medical system, including medical facilities in county, town and village levels, was to be established, in order to provide medical service at each local level. In urban areas, community medical facilities were to be strengthened. Training for medical professionals were also improved at local levels. Finally, pilot programs for public hospital reform were started by the central government after
2009.
2.2 Three Major Health Insurance Schemes As discussed in the last section, China is now implementing ambitious reforms of the health insurance system, and three types of health insurance schemes have been launched.
These three schemes were launched in different years targeting different population groups. Two insurance schemes cover the urban residents, and the third one covers the rural residents.
2.2.1 The Basic Medical Insurance for Urban Employees In 1998, the Chinese State Council issued the Decision of the State Council on
Establishing the Urban Employees’ Basic Medical Insurance System. This was the first step in re‐establishing the health insurance system in urban areas. The Urban Employees Basic
Medical Insurance (UEBMI) is compulsory based on employment. It provides basic medical insurance coverage for urban employees in both the public and private sectors (State
Council, 1998). Local governments, mainly at the municipal level, set the level of deductibles, copayments, and reimbursement caps according to local economic levels.
10 The policy was launched in early 1999, and in late 1999, it was expanded
nationwide. By the end of 2002, about 94 million people participated in the UEBMI. In
order to further expand the coverage, the Ministry of Human Resources and Social Security
issued Notification of Further Expanding the Coverage of the Urban Employees Basic
Insurance Coverage in 2003. By the end of 2008, the number of insured totaled 200 million.
The UEBMI is financed by premiums from both employers and employees. In their
decision, the State Council suggested that the employers’ contribution be 6% of the
employee’s salary and the employees’ percentage be 2%. The revenue collected from
premiums is distributed evenly into two independent accounts: the Medical Savings
Account (MSA) and the Social Pooling Account (SPA). All employees’ contributions and
about 30% of employers’ contributions go into the MSA, and the remainder of the employers’ contributions goes to SPA. The two accounts are managed separately and pay for different services: the MSA covers outpatient and emergency services and drug expenses, and the SPA covers inpatient services.
2.2.2 The Basic Medical Insurance for Urban Residents In 2007, the State Council issued guidelines to launch the Urban Residents Basic
Medical Insurance (URBMI). According to the guidelines, the URBMI covers primary and
secondary school students who are not covered by the UEBMI (including students in
professional senior high schools, vocational middle schools, and technical schools), young
children, and other unemployed urban residents on a voluntary basis (State Council, 2007).
The main purpose of the guidelines is to provide coverage for urban residents without
11 formal employment with the intention of eliminating impoverishment resulting from chronic or fatal diseases, which can lead to catastrophic medical expenditures.
The URBMI was piloted in 79 cities, including two to three cities in each of the provinces that were able to participate, and expanded to more cities in 2008 and 2009, with the objective of covering 80% of all cities in the participating provinces. In 2010, this insurance scheme was expanded nationwide and gradually extended to all unemployed urban residents. The number of insured was about 43 million by the end of 2007 and increased to 118 million by late 2008 (China Ministry of Health, 2010).
The financing of this insurance program mainly comes from participants’ premiums.
The government also provides a smaller amount of subsidies, compared to the premium contributions. The premium of the policy is determined by the local government, according to the local economic level, the medical care expense level, and the participants’ household income level. When the policy was launched, the government contribution was at least 40
Yuan per participant. From this amount, the central government transfers 20 Yuan to central and western areas residents. There are extra government subsidies for low‐income families, disabled students, and young children (State Council, 2007). The URMBI mainly targets people with chronic and fatal diseases; therefore, it covers more expenses for inpatient services. In 2008, the URMBI covered 45% of expenses from inpatient service related to chronic and fatal diseases, which equaled 1436 Yuan per inpatient stay (State
Council Evaluation Group for the URBMI Pilot Program, 2008).
12 2.2.3 The New Rural Cooperative Medical Insurance In 2003, the State Council issued the Decision to Further Enhance the Rural Health
Care System, aimed at re‐establishing the Rural Cooperative Medical Insurance (NRCM).
The NRCM scheme covered the rural residents on a voluntary basis in order to avoid impoverishment caused by catastrophic expenses from infectious and endemic diseases.
The NRCM was piloted in 2003 in selected counties. In 2006, coverage increased to 40% of all counties, and about 60% in 2007. In 2010, the NRCM covered more than 90% of all rural residents.
The NRCM was funded by premiums from both the insured and by subsidies from the local and central governments. In 2003, the central government provided a subsidy of
10 Yuan for each insured resident. The Council’s 2003 decision also required local governments to provide no less than 10 Yuan. In 2011, the subsidized amount was raised to a total of 200 Yuan. The NRCM provides partial coverage for all kinds of medical expenses, excluding some outpatient expenses and drug expenses. The reimbursement caps vary by local economic development levels.
2.3 Trends and Current Status of Healthcare Disparities China is a vast country with uneven economic development. Rural and urban residents are categorized separately according to the household registration system. The government financing systems for rural and urban sectors are also separate. Most of the government revenue comes from the urban economy, and most is spent on urban economy as well. This is especially true in public service areas, resulting in the urban‐rural disparity.
13 As mentioned before, by 1998, the urban‐rural disparity in health insurance coverage had become prominent. The coverage gap persisted in subsequent years. For example, in 2003, the urban health insurance coverage rate was still more than 50%, while only about 20% of the rural residents were covered by some form of health insurance coverage, and about half of the 20% was covered by pure commercial health insurance.
This is shown in Figure 2.1, which presents the percentage of residents covered by health insurance in both urban and rural areas over time. During the selected period, public health insurance coverage was reduced year by year in both rural and urban areas until 2003.
However, the percentage of coverage had always been much lower in rural areas than in urban areas.
Then, in 2008, there was a large increase in insurance coverage, especially for rural areas. Coverage increased to more than 90%, and a larger portion of rural residents was covered by health insurance at this time, compared to the portion of urban residents. We can also observe the shift in the urban‐rural ratio (the green line). Before 2003, the urban‐ rural ratio of health insurance coverage was extremely high; however, in 2008, the ratio decreased to less than 1, indicating more coverage in rural areas. Between the two time points, there were several policy changes that affected health insurance coverage. In the urban areas, the basic medical insurance for urban employees was launched in 1998, and in
2007, the basic medical insurance for urban residents was established. In the rural areas, in
2003, the government started to rebuild the cooperative health insurance system (NRCM), which influenced a very large population. Most of the rural coverage in 2008 was from
NRCM. Therefore, I believe the initiation and expansion of the NRCM diminished the disparities in health insurance coverage; however, it is still unknown whether the
14 expansion helped reduce disparities in other healthcare areas, such as healthcare
utilization and cost.
Figure 2.1 Health Insurance Coverage in Urban and Rural Areas in China, Selected Years 1993-2008
Disparity was also observed in other healthcare issues related to health insurance coverage, such as in healthcare utilization and out‐of‐pocket cost, especially before the year
2003. On one hand, the urban‐rural disparity on healthcare utilization decreased from
1993 to 2003. For example, in 1993, the percentages of hospital outpatient service use in
the two weeks prior to the survey for urban and rural residents were 19.9% and 16.0%,
respectively; in 2003, the percentages became 11.8% and 13.9%, respectively (China
Ministry of Health, 2004). On the other hand, in 2003, about half of the residents in rural
areas who sought outpatient services went to informal healthcare institutions instead of to
formal hospitals, while the percentage in urban areas was only about 25%. The shrinkage
15 of the urban‐rural gap of healthcare utilization was due to the reduction in informal healthcare institutions in urban areas (China Ministry of Health, 2004). Moreover, the percentage of unmet needs was highest among the low‐income population in rural areas
(China Ministry of Health, 2004).
The healthcare utilization disparity was most prominent in the health service area.
Figure 2.2 shows the percentage of pregnancy healthcare utilization and the percentage of women who gave birth in hospital in 2003. We can see that rural women used less of these services, especially low‐income women. By 2008, the disparity in health service utilization had been relieved but still existed. The percentage of pregnancy healthcare utilization had risen to 93.7% for rural women. Compared to the 97.6% ratio for urban women, the rate of healthcare utilization was still lower but the gap between urban and rural had become narrower.
Health Service Utilization in Urban and Rural Areas in China, by Income (2003) 100% 90% 98% 80% 70% 85% 81% 60% pregnancy 50% health care 40% 30% 45% give birth in 20% hospital 10% 0% lowest highest lowest highest percentile percentile percentile percentile Urban Rural Source: China Ministry of Health, The Third National Health Services Survey Report (in Chinese), 2004, http://www.moh.gov.cn/publicfiles///business/cmsresources/mohwsbwstjxxzx/cmsrsdocument/doc9908.pdf (accessed Aug. 28, 2012)
Figure 2.2 Health Service Utilization in Urban and Rural Areas in China (2003)
16
Driven by limited health insurance coverage and rapidly growing healthcare costs, high out‐of‐pocket expenses comprised a major challenge for those seeking healthcare.
China became one of the Asian countries with the highest ratio of out‐of‐pocket cost to total healthcare costs in 2002 (Yip and Hsiao 2008). At that time, the out‐of‐pocket ratio was 60%
(Hu, Tang et al. 2008), and rural residents bore an even higher ratio. The trend of health spending is shown in Figure 2.3. The percentage of out‐of‐pocket payments by individual patient rose steadily from 1980 to 2001. This trend indicates that the financial burden of healthcare shifted more and more to the individual patients during that period. However, after 2001, the government and social programs started to take on more of the cost, and this resulted in a downward influence on individual out‐of‐pocket payments.
Healthcare Spending in China, by Source and Year 70
60 Individual Patient, 50 38.2
40 Social Programs, 34.6 30
Percentage Government, 27.2 20
10
0 1978 1980 1982 1984 1986 1988 1990 1992 1994 1996 1998 2000 2002 2004 2006 2008
Source: China Ministry of Health, China Health Statistics Yearbook(in Chinese), 2010, http://www.moh.gov.cn/publicfiles/business/htmlfiles/zwgkzt/ptjnj/year2010/index2010.html
Figure 2.3 Healthcare Spending in China, by Source and Year
17 Per Capita Out‐of‐pocket Health Expenses as a
30.0% Percentage of Income
25.0%
20.0%
15.0%
10.0% 1993 5.0% 1998 2003 0.0% lowest middle highest lowest middle highest percentile percentile percentile percentile Urban Rural Source: China Ministry of Health, The Third National Health Services Survey Report (in Chinese), 2004, http://www.moh.gov.cn/publicfiles///business/cmsresources/mohwsbwstjxxzx/cmsrsdocument/doc9908.pdf (accessed Aug. 28, 2012)
Figure 2.4 Per Capita Out-of-Pocket Health Expenses as a Percentage of Income
Figure 2.4 shows the per capita out‐of‐pocket health expenditure as a percentage of income by urban and rural areas. Rural residents paid for medical service with a larger portion of their incomes than did urban residents. Among the poorer rural residents, out‐ of‐pocket payments for healthcare services constituted 26.7% of their total income in 2003, a large increase from the percentage ten years earlier.
18 Chapter 3 Literature Review and Study Objectives
3.1 Existing Research Two research areas inform my study. The first area comprises research on
healthcare disparities. As discussed, urban–rural disparities in health and healthcare have
drawn attention in China in recent years. Many studies have provided empirical evidence
on the conditions, trends, and associated factors of such disparities in health status, healthcare utilization, healthcare costs, and related issues such as health insurance coverage. Other research in this area has focused on examining the determinants of the disparities. The second area of research includes assessments of the insurance schemes in
China in terms of impact on healthcare utilization, out‐of‐pocket cost, and health outcomes.
Although these studies are usually not focused on healthcare disparities, I viewed them as a good foundation for my research. I also found these studies helpful in terms of data and methodology. In the next section, I review some of the key research.
3.1.1 Literature on Rural–Urban Disparities in Healthcare Utilization Recent studies have provided empirical evidence on the conditions and trends of
rural–urban healthcare disparities (Liu, Hsiao et al. 1999; Zhao 2006; Tang, Meng et al.
2008; Meng, Zhang et al. 2012). Liu, Hsiao, and Eggleston (1999) examined the changes in
disparity in health status and healthcare utilization in China from 1985 to 1993 and found
that the gap in health status and healthcare utilization between urban and rural residents
widened during the transitional period when the Chinese economy was shifting from a
command economy to a market economy. The authors concluded that the trends were
correlated with the reduction of rural health insurance coverage. Zhao (2006) provided
evidence for later years, showing that the rural–urban disparities in morbidity and
19 mortality levels were associated with disparities in healthcare access. Meng, Zhang et al.
(2012) provided similar evidence on disparities in maternal and under‐five mortality rates.
Tang, Meng et al. (2008) pointed out that there were rural–urban disparities in a set of child health indicators, including infant mortality rate, level of malnutrition, child stunting, and underweight status. However, the researchers believed that China has the ability to carry out the necessary reforms to improve health equity.
Several researchers specifically examined disparities in healthcare access and utilization to identify the determinants of healthcare utilization. (Gao, Tang et al. 2001;
Wang, Yip et al. 2005; Gao, Raven et al. 2007; Liu, Zhang et al. 2007; Fang, Chen et al. 2009;
Jian, Chan et al. 2010; Long, Zhang et al. 2010; Feng, Guo et al. 2011; Xu and Short 2011; Liu,
Tang et al. 2012; Meng, Zhang et al. 2012). Among these studies, researchers presented
mixed findings. Generally, the authors agreed that most healthcare resources were being
allocated to urban areas and that urban residents use more formal healthcare than do rural
residents. However, Fang, Chen et al. (2009) examined the evolution of rural–urban
disparities in healthcare utilization from 1997 to 2006 and concluded that rural residents
actually visit physicians more often than do urban residents when they are ill. Some of the
researchers pointed out that better insurance coverage was associated with increased
healthcare utilization. Liu, Zhang et al. (2007) noted that hospital utilization was lower
among the uninsured.
Some of the studies focused on certain subpopulations and reached similar
conclusions. Gao, Raven et al. (2007) examined the trend of inpatient utilization among the
elderly in urban China, and they found that within this subpopulation, the insured were
20 more likely to use inpatient care. Jian, Chan et al. (2010) analyzed changes in the rural–
urban gap for patients with chronic disease, drawing on data collected between 2003 and
2008. They concluded that the gap between urban and rural residents was narrowed in
terms of hospital admission rates; however, there was no change in terms of early self‐
discharge from hospital. Liu, Tang et al. (2012) analyzed the impact of health insurance on utilization of outpatient and inpatient services. They concluded that having health insurance coverage had no significant impact on outpatient service utilization; however, inpatient service utilization increased.
Some of the researchers found that changes in disparities and the impacts of health
insurance coverage were different among different income groups. Gao, Tang et al. (2001)
concluded that from 1993 to 1998, healthcare access for low‐income groups shrank more
than did healthcare access for high‐income group. Liu, Tang et al. (2012) pointed out that
the effect of insurance coverage on inpatient service utilization was greatest for high‐
income groups, while low‐income group enjoyed fewer benefits.
3.1.2 Literature on Disparities in Out‐of‐Pocket Expenditure and Healthcare Costs Several studies focused on the disparities and determinants of out‐of‐pocket
expenditures and healthcare cost (Pan, Dib et al. 2009; Sun, Jackson et al. 2009; Long,
Zhang et al. 2010). The researchers generally agreed that rural residents tended to be at
increased risk for high and catastrophic medical payments; the current insurance schemes in rural areas offer limited financial protection. Pan, Dib et al. (2009) concluded that hospitalization costs were higher among insured patients because the insured generally stayed longer in hospital than did the uninsured. Long, Zhang et al. (2010) found that
21 participating in the NRCM reduced out‐of‐pocket expenditures on average, but the rural poor were still faced with high payment problems. Sun, Jackson et al. (2008) pointed out that out‐of‐pocket payments remained a burden for rural residents after the initiation of
NRCM.
3.1.3 Literature on Disparities in Health Insurance Coverage Research has focused on the trends of disparities in health insurance coverage (Akin,
Dow et al. 2004; Xu, Wang et al. 2007; Xu and Short 2011). Akin, Dow & Lance (2004) examined changes in health insurance coverage from 1989 to 1997 and concluded that the overall coverage decreased slightly, from 26% in 1989 to 23% in 1997. They further pointed out that urban areas (cities and towns) experienced reductions in health insurance coverage, while rural area coverage increased. However, the changes were very small, and the rural–urban disparity in health insurance coverage persists. Xu, Wang et al. (2007) used data from the National Health Services Surveys of 1998 and 2003 to examine the impact of the reform on population coverage, and they concluded that the overall health insurance coverage stayed almost the same among urban residents. Xu and Short (2011) examined the trends of health insurance coverage from 1997 to 2006. They pointed out a sharp increase of coverage in 2006 in rural residents, which resulted in a smaller gap in health insurance coverage between rural and urban residents.
3.1.4 Methodological Issues
3.1.4.1 Definition of Rural and Urban Two definitions are used to determine rural and urban status in China. The first definition classifies residents by geographical residential areas, which are officially divided into urban and rural areas by the National Bureau of Statistics of China, according to
22 China’s administrative divisions. The second definition is by household registration type.
China classifies people as either agricultural (rural) or nonagricultural (urban). These categorization data are recorded by the household registration (Hukou, 户口) system.
These two definitions of rural and urban status are not entirely consistent.
Different definitions of rural areas can lead to different results when studying health policy, because the definition of rural areas affects the resources to which people have access (Hart, Larson et al. 2005). However, few existing studies address the definition specifically. For most of the studies, I identified the authors’ definitions of rural/urban areas only by the terminology used. For example, if the authors used terms such as residents, areas, or geographic regions, I viewed these terms as being consistent with the first definition. If the authors mentioned household registration or used the term population,
I viewed these terms as consistent with the second definition. In all of the cited papers, the researchers adopted the first definition except for one study assessing NRCM. Lei & Lin
(2009) adopted both the first and second definitions when they evaluated NRCM. However, they restricted their sample by only including people who lived in rural areas and were with rural household registration.
3.1.4.2 Modelling In terms of methodology, most of the studies mentioned were descriptive, and some of the papers used cross‐sectional data to fit logit/probit models. The researchers emphasized the problem of urban–rural disparities in healthcare in China and clarified the trends and current conditions, as well as provided direction for further study of this issue.
However, no research has provided a complete picture of how the disparities in health
23 insurance coverage, healthcare utilization, and healthcare cost change over time. Little research has focused on the role of health insurance coverage on closing the rural–urban gap in healthcare utilization and healthcare costs, while considering all major insurance changes.
As discussed before, some researchers have evaluated NRCM, and this type of research provided me with methodological help. Wagstaff & Lindelow (2009) drew on multiple data sources to study the insurance and financial risk in China before 2003. They applied fixed‐effect models for two panel datasets and an instrumental variable (IV) technique for a cross‐sectional dataset, and they concluded that having health insurance in
China does not always reduce financial risk. They explained this curious phenomenon by adverse selection, i.e., people with higher risk of high medical expense tend to join the insurance scheme. The advantage of this research is that it used panel data and advanced analysis techniques. However, there were still drawbacks in this study’s methodology.
Their longest panel had only four waves, and these waves covered a time period before the
NRCM was launched. As discussed before, all health insurance systems had experienced changes to some extent at that time. It would be more comprehensive and convincing to extend the research by incorporating the most recent data.
More recently, three other papers addressed the NRCM using different data and methodologies, reaching mixed conclusions (Lei and Lin 2009; Yu, Meng et al. 2010; Lu, Liu et al. 2012). In the first study, Lei & Lin (2009) concentrated on evaluating the healthcare service and health outcome after the initiation of NRCM. They used panel data from the
China Health and Nutritious Survey to estimate fixed‐effect and IV models, and they also
24 applied a difference‐in‐differences estimation with propensity score matching. The researchers found no evidence that the NRCM decreased out‐of‐pocket expenditures or increased utilization of healthcare service. Therefore, they concluded that the impact of the
NRCM was limited. In their study, they included only three waves of data, one before NRCM was launched and two waves after. This panel could still be expanded to include richer information.
In the second study, Yu, Meng et al. (2010) used data from six counties in two provinces to conduct a cross‐sectional study to examine whether the launch of NRCM increased healthcare utilization. They found that NRCM did not significantly increase outpatient service utilization in rural areas, while inpatient service in general increased.
Further, they pointed out the association between household income and healthcare utilization. The authors concluded that the increase happened only among the most affluent. For people with middle and lower incomes, the increase was not significant.
In the third study, Lu, Liu et al. (2012) used data from the 2001 China Health
Surveillance Baseline Survey to investigate whether the launch of NRCM led to an increase in healthcare utilization and a decrease in possible catastrophic medical expense for rural residents. Similar to the method used by Lei & Li (2009), Lu, Liu et al. also used propensity score matching, and applied the IV method. They found that NRCM did not decrease out‐of‐ pocket expenses. However, unlike Lei & Li (2009), they found that NRCM did significantly increase healthcare utilization.
25 3.2 Gap in the Existing Literature To sum up, current research provides empirical evidences on the rural–urban disparities in health insurance coverage, healthcare utilization, and healthcare costs.
However, current research could be improved in several ways. First, in current studies, researchers have examined rural–urban disparities in different time periods, but have not provided a complete picture of the trends in rural–urban disparities. Second, the determinants of rural–urban disparities have not been well examined. The impact of health insurance status, which can be a very important policy intervention to reduce disparities, has not been well studied. Third, in the papers on health insurance or healthcare disparities, the authors have not drawn consistent conclusions; the studies could be improved in terms of data quality and methodology. Fourth, the papers on the impact of health insurance usually focus on certain population groups. For example, when studying the effects of
NRCM, researchers usually focus only on rural residents.
The first possible expansion to existing literature is to include more waves of data to show a more complete picture of the trends of change in rural–urban disparity in health insurance coverage, healthcare utilization, and healthcare cost. The second possible expansion to these studies is to include more waves of data and to use advanced techniques to examine the determinants of the disparities and thus provide policy suggestions on ways to further relieve the disparities. In addition, among the factors associated with the disparities, health insurance is an important issue to study. The third area of expansion is to include urban areas as a control group when examining the impact of health insurance expansion. To address these gaps in the existing literature, I explored all possibilities in my research.
26 3.3 Objectives and Research Questions The objectives of my research were to examine the status and trends of rural–urban disparities in healthcare utilization and costs, to analyze the role of health insurance coverage in reducing these disparities, and to provide evidence and suggestions to policy makers about how to further reduce rural–urban healthcare disparities.
My research questions were:
1. What do the rural–urban disparities in healthcare utilization and costs look like?
How do the disparities change along with major health insurance policy changes?
2. Does more health insurance coverage in rural area reduce the rural–urban
disparities in healthcare utilization?
3. Does more health insurance coverage in rural area reduce the disparities in high
out‐of‐pocket healthcare expenditure and total healthcare costs?
4. Does the impact of health insurance on disparities differ by income group and by
region?
27 Chapter 4 Study Design
4.1 Data For this study, I drew on the detailed individual‐level longitudinal data from the
China Health and Nutrition Survey (CHNS), which is a collaborative project between the
Carolina Population Center at the University of North Carolina at Chapel Hill and the
National Institute of Nutrition and Food Safety at the Chinese Center for Disease Control
and Prevention. As a panel survey, CHNS started in 1989 and has been conducted roughly
every other year. I used the most recent seven waves of data (1993, 1997, 2000, 2004,
2006, 2009, and 2011) in the analysis. The 1989 and 1991 datasets were not used because
these datasets did not contain health insurance information or household registration
information.
CHNS used a multistage, random cluster‐sampling approach, and was conducted in
nine provinces,1 which are mostly representative of Central and Eastern China and vary
substantially in geography, economic development, public resources, and health indicators.
Counties in the nine provinces were stratified into three layers by income, and a weighted
sampling scheme was used to randomly select four counties in each province. Villages and
townships (the CHNS definition of communities) within the counties and urban and suburban neighborhoods within the cities were then selected randomly into primary
sampling units (PSUs). The same households were surveyed over time whenever possible
and newly formed households were included beginning in 1993. In the sample, rural
communities had populations ranging from 125 to 14,964 people, and urban communities
1 The nine provinces are Guangxi, Guizhou, Heilongjiang, Henan, Hubei, Hunan, Jiangsu, Liaoning, and Shandong. In the 2011 wave, three municipalities (Beijing, Shanghai and Chongqing) were added into the sample.
28 had populations ranging from 167 to 86,733 people. In this study, I included all respondents who responded to the health insurance section. This final sample included more than 90,000 respondents. The sample sizes are shown in Table 4.1.
CHNS was a good data source for the research because it provided detailed information on insurance coverage, medical providers, health services use, and healthcare costs. Therefore, CHNS allowed me to look at how insurance coverage affects health service use and health financing. Questions about healthcare accessibility, time and travel costs to health facilities, and perceived quality of care were also asked.
Table 4.1 Sample Size by Rural and Urban Residences and Registrations
Rural Urban Residents Residents
Rural Urban Rural Urban Wave Registration Registration Registration Registration Total
1993 7,663 2,253 1,433 2,470 13,819
1997 7,255 2,492 1,661 2,801 14,209
2000 7,956 2,601 1,563 3,015 15,135
2004 6,016 2,081 1,188 2,858 12,143
2006 5,774 2,059 1,228 2,679 11,740
2009 5,931 2,064 1,241 2,688 11,924
2011 6,489 2,874 1,420 4,717 15,500
Total 47,084 16,424 9,734 21,228 94,470
29 4.2 Study Periods For this analysis, I classified the study period of 1993–2011 into four periods:
1. 1993–1997, a period before the major health insurance expansion in China
2. 2000, a period after the initiation of UEBMI in 1998
3. 2004–2006, a period after the initiation of NRCM in 2003
4. 2009–2011, a period after the initiation of URBMI in 2007
4.3 Conceptual Model and Variable Selection The variable selection was based on the Andersen model (Andersen 1968). The model focused on the individual as the unit of analysis and, when first developed, was used to explain why people use healthcare services. After several generations, the model grew to include other endpoints of interest, such as healthcare quality and health outcomes
(Andersen 1995).
Figure 4.1 shows the most recent Andersen model. This figure depicts the interaction between environment, population characteristics, health behavior, and health outcomes. Specifically, the healthcare system includes policy, resources, and organizations; predisposing characteristics include demographic characteristics, health beliefs, and social structure; enabling resources includes income, health insurance, and other resources for healthcare services. All these characteristics can impact the decision to use health services and further influence healthcare outcomes. Health behavior can influence enabling resources; health outcomes can affect enabling resources and health behaviors (Andersen
1995). Therefore, by including personal demographic information, family and social structure, income, insurance status, health conditions, and policy change in the model, I was able to examine how these factors affected peoples’ healthcare‐seeking behaviors and
30 healthcare costs. The variables of health insurance coverage and types of coverage are
viewed as enabling factors in the model. By including location information about urban
versus rural areas, I also controlled the impact of the external environment.
Figure 4.1 Updated Structure of Anderson Model
Moreover, Andersen assigned a degree of mutability to the model components when
he developed the model. According to Andersen, the most mutable population
characteristic component was enabling resources, which included insurance coverage. In
my analysis, status of health insurance was affected by policy changes. Therefore, when interpreting the results, I focused on the impact of health insurance coverage on healthcare utilization and costs, and the resulting policy implications.
4.3.1 Dependent Variables The analysis focused on urban–rural disparities in healthcare utilization and
healthcare costs. All the healthcare utilization questions in CHNS focused on a four‐week
period right before the interview. For healthcare utilization, I constructed three variables:
31 formal care utilization, outpatient care utilization, and inpatient care utilization. Formal
care utilization is a binary variable indicating whether the respondent sought formal
medical care from a hospital or clinic in the four weeks before the interview. The formal
care utilization variable was constructed from several raw variables: (a) whether the
respondent was sick or injured or suffered from a chronic or acute disease, (b) whether the
respondent sought care from a formal medical provider, and (c) what the respondent did
when he or she was ill or injured. If the answer to the first question was “yes,” the
respondent was asked the second and third questions. If the answer to the second question was “yes” or the answer to the third question was “saw a doctor (clinic, hospital)”, I
considered the respondent to have sought formal medical care in the past four weeks.
There were some inconsistences in the question setting and wording across waves. In
waves 1993 to 2000, CHNS only asked the second question, and repeated the question for a
second facility. In the latter waves, CHNS asked both questions.2 The outpatient and
inpatient utilization were also binary variables. They were constructed from the raw
variable of whether the visit was an inpatient or outpatient visit.
For healthcare expenses, I constructed two types of variables. The first type of
variable involved the amount of total healthcare costs. The second type contained several
binary variables indicating whether the out‐of‐pocket healthcare costs were more than a
certain percentage of the household income. I used two cut‐off points for the percentage:
20% and 40%. The amount of total healthcare costs was derived from the raw variables
underlying the treatment costs. The amount of out‐of‐pocket costs was constructed from
2 There has been a jump of percentage of people who use formal medical care since the 2004 wave. However, the change is not a result from the setting of the questions.
32 the total treatment costs and percentage of treatment costs paid by insurance and other
cost of treating the illness or injury. These variables were also constrained to the four‐week
period before the interview. I inflated the amounts to 2011 values using the index from
CHNS data. In the survey, the question about household income referred to a time period of
one year. Therefore, I multiplied the out‐of‐pocket healthcare expenses by 12 to match the
two time frames. The healthcare costs variables measured the costs within 4 weeks before
the interview, thus the costs could be from acute illness and be overestimated when transported to costs in one year. Therefore, I did not pick a lower cut‐off point for high out‐ of‐pocket costs.
4.3.2 Independent Variables
4.3.2.1 Key Independent Variable: Dummies Indicating the Respondents’ Residence and Household Registration Type My key independent variable was a set of dummies indicating the respondents’
resident area and household registration type. There are two definitions of rural and urban
in China. The first consists of geographic residential areas, which are officially divided into
urban and rural areas. The National Bureau of Statistics of China officially assigns these
levels. This variable was directly created from the primary sampling units of CHNS, which
drew samples from cities, suburbs, towns, or villages. The first two designations—cities and suburbs—are considered urban areas; the latter two are classified as rural areas. The
second definition of rurality is by type of household registration. China classifies people as
either agricultural (rural) or nonagricultural (urban) population, recorded by the
household registration (Hukou, 户口) system. These two definitions are not completely
consistent, for three possible reasons: (a) there are areas in China called urban–rural mixed
33 areas (城乡结合部), but they can only be classified as either urban or rural area; (b) increasing numbers of people with rural household registration migrate to urban areas to work, but their household registrations do not change; and (c) some people with urban household registration, especially in recent years, have chosen to live in rural areas. Most of the agricultural population resides in rural areas. In my CHNS sample, 75% of people with agricultural household registration lived in rural areas, and 67% of people with nonagricultural household registration lived in urban areas. These percentages stayed relatively consistent across waves; therefore, my assumption was that the sample covered few migrating rural workers. If this were not the case, there should be greater numbers of rural workers migrating to urban areas as the economy develops and the control of residency relaxes.
As discussed in the literature review, most of the studies on the disparity issue used residential area to define rurality, while most studies evaluating NRCM used the household registration system to define rurality. In my research, I sought to examine the changes in disparities, as well as to establish a link between insurance and disparity. Therefore, I used both of the two classifications to divide people into four categories: rural residents with rural registration (Group RR), rural residents with urban registration (Group RU), urban residents with rural registration (Group UR) and urban residents with urban registration
(Group UU). I used Group UU as the reference group and compared the three other groups with it.
By adopting the four categories, I was able to track all three health insurance policy changes that expanded health insurance coverage to people with certain household
34 registration types and to people living in certain areas. I was also able to examine how the disparity levels changed with the residing environment. As discussed, the policy changes also included construction of healthcare facilities, training of medical service workers, and drug policy changes. These are all applied to the residing environment and can affect the residents’ healthcare utilization and costs.
4.3.2.2 Descriptive Statistics of Independent Variables Other independent variables included basic demographic characteristics, family size and wealth, health measures, and health insurance status. Table 4.2 shows descriptive statistics of all the independent variables. In order to reflect the difference between rural and urban residents, I report the statistics separately for rural and urban residents. From the descriptive statistics, rural and urban residents were substantially different. In my sample, rural residents contained a slightly larger portion of males and minorities than urban residents. Rural residents were younger than urban residents, on average, although I observed aging trends in both groups. More urban residents were married, but rural residents usually had larger household sizes. Urban residents had higher education levels and incomes than did rural residents.
4.3.2.3 Equivalence Scale for Adjusting Household Income In order to provide a more accurate measure of household income, I used the equivalence scale to adjust the size of household and then computed the per‐capita household income using the adjusted household size. I chose to apply one of the most commonly used scales, the square‐root scale, which involves dividing household income by the square root of household size. This scale was adopted by some recent OECD publications on income inequality and poverty (e.g., OECD 2011).
35 4.3.2.4 Missing Value Imputation for Independent Variables I performed basic imputation for missing values. For marital status, I replaced the
missing values with “never married” if the respondent was younger than 18. According to
China’s marriage law, the youngest age to get married is 18. For household size and
household income, I imputed the missing values using other household members’ answers.
For household registration type, if the value was missing in one wave, but the previous and
post waves had the same values, I assigned this value to the missing wave.
For missing values in education years, I assigned 0 to the variable if the respondent
was younger than seven. If the values in the previous and post waves were equal, I assigned
the same value to the missing wave. If the values in last two waves did not change, I
assigned the same value to the missing wave. If the respondent was older than 30, I
assigned the value from the previous wave to the missing wave. I used the value from the
variable indicating years of formal education to impute the missing values in highest level
of formal education, which was used in the analysis. For missing values for the variable of
whether the respondent was still in school, I replaced the value with 0 if the respondent was younger than seven or older than 30.
For missing values in the variable of having any medical insurance coverage, I assigned 1 to the variable if the respondent claimed to have any type of medical insurance.
After the basic imputation, there were still a few missing values. The percentage of
missing values was generally less than 1%. In order to better use the information in the
dataset, I created additional categories in each variable indicating whether the value was
missing and included the categories in my analysis.
36 Table 4.2 Descriptive Statistics of Independent Variables by Rural and Urban Residences and Registrations Group RR Group RU Group UR Group UU N=47,084 N=16,424 N=9,734 N=21,228 gender male 0.496 0.514 0.482 0.485 female 0.504 0.486 0.518 0.515 Ethnicity Han 0.843 0.873 0.826 0.945 Minority 0.156 0.121 0.173 0.049 unreported 0.001 0.006 0.002 0.006 age age equal or below 5 0.063 0.044 0.057 0.035 age between 6 and 17 0.179 0.141 0.177 0.114 age between 18 and 60 0.621 0.632 0.642 0.625 age equal or above 61 0.136 0.183 0.123 0.225 unreported 0.001 0.001 0.002 0.001 marital status married 0.602 0.640 0.602 0.665 never married 0.332 0.281 0.328 0.251 other(divorced, widowed, or separated) 0.059 0.071 0.063 0.076 unreported 0.007 0.008 0.007 0.008 education level primary school 0.623 0.381 0.560 0.329 middle school 0.295 0.293 0.307 0.251 high school 0.074 0.256 0.116 0.287 college and above 0.003 0.063 0.011 0.126 unreported education status 0.005 0.006 0.006 0.007 whether still in school whether still in school 0.141 0.136 0.157 0.118 not in school 0.839 0.847 0.831 0.873 unreported whether in school 0.020 0.017 0.012 0.009 income groups low income group 0.384 0.261 0.314 0.173 medium income group 0.345 0.351 0.325 0.305 high income group 0.270 0.387 0.359 0.517 unreported 0.001 0.001 0.002 0.006 Note: 1. Income was adjusted for inflation to 2011 value 2. Adjusted per‐capita household income was used
37 4.4 Analytic Approach Difference‐in‐differences (DID) analysis comprised my main analyzing technique.
DID analysis assumes parallel trends in control and treatment groups before the policy intervention. For the variables for which the parallel trends did not hold, I performed multivariate models, controlling for existing trends. I also performed several sensitivity analyses, each of which had different focuses, as discussed in the next section.
4.4.1 Difference‐in‐Differences Analysis with Multiple Groups and Multiple Time Periods Using the longitudinal data collected in seven waves between 1993 and 2011 enabled me to take a DID approach in my empirical analysis. This approach has become increasingly popular in the empirical literature on the effects of public policy interventions.
The DID estimation is based on the simple idea of comparing the difference in outcomes before and after an intervention for groups affected by it to the difference for unaffected groups. The great appeal of DID estimation comes from its simplicity as well as from its potential to mitigate biases in the comparison between the treatment and control group that could be the result of permanent differences between those groups, as well as to mitigate biases from the pre‐post comparison of the treatment group that could be the result of secular trends unrelated to the intervention (Card and Krueger 2000; Athey and
Imbens 2002; Bertrand, Duflo et al. 2004; Abadie 2005; Conley and Taber 2005). My research focused on the change in disparities. Further, the setting of the research questions made DID the most suitable approach.
The DID analysis can be expanded to include more than two time periods (Bertrand,
Duflo et al. 2004; Hansen 2007). As discussed, there have been three major policy changes
38 in health insurance in China. I included all three major policy interventions on health
insurance in my model. My main hypothesis was that the second policy change, which
expanded insurance coverage in rural areas in 2003 helped reduce rural–urban disparities
in healthcare utilization and costs. However, it was important to take the other two policy
changes in urban areas into consideration and separate the effects from different policy changes.
After the DID model, I interpreted the results using the whole sample to make predictions for different residence and registration groups in each period. The results are presented in bar graphs. Using the adjusted outcome variables, I was able to observe the trends in disparities.
4.4.1.1 Econometric Models In this section, I elaborate on how I built econometric models to perform the analysis based on the conceptual framework. For different outcomes, I applied different techniques.
Considering the dichotomous variables, such as whether a person used outpatient care, I applied logistic regression model and a general framework considered by Bertrand,
Duflo et al. (2004) and Hansen (2007). Empirically: