Avoidable Hospital Admissions From Diabetes Complications In Japan, Singapore, Hong Kong, And Communities Outside Beijing

Diabetes has recently been at the center of an urgent call for global action, with “beat diabetes” the theme of 2016 World Health Day on April 7 and the World Health Organization (WHO) releasing its first Global Report on Diabetes.¹ The corresponding 2016 Lancet report showed an approximately fourfold increase in global diabetes prevalence over the past thirty-five years.² The prevalence of diabetes increases with population aging, obesity, and urbanization—all of which are key demographic and epidemiological transitions under way in East Asia.³ The rapid increase in diabetes prevalence implies increased morbidity and premature mortality in addition to large increases in health care spending on complications such as cardiovascular disease, diabetic nephropathy, retinopathy, and neuropathy. In 2015 an estimated 12 percent of global health expenditures (US$673 billion) was related to diabetes control, management, and complications.⁴ Thus, diabetes is a major health challenge as countries confront the growing burdens of chronic noncommunicable diseases and population aging.

A key component of health expenditures is the cost of inpatient hospital care. Many inpatient hospitalizations can be avoided with appropriate ambulatory care for conditions that are best treated in an ambulatory setting. Although not all hospital admissions are avoidable, the concept of avoidable admissions has been used in quality metrics such as the Health Care Quality Indicators of the Organization for Economic Cooperation and Development (OECD), on the assumption that proper ambulatory care can manage the health of patients with chronic diseases such as diabetes and prevent the onset of future diabetes-related illnesses or complications.⁵ Given the high cost of inpatient care relative to outpatient care, and the fact that resorting to inpatient care usually signals ineffective management and worsening severity of diabetes, reducing avoidable hospital admissions rates could improve the quality of care for patients with diabetes while reducing expenditures (for a list of diagnosis codes for avoidable hospital admissions for diabetes, see online Appendix Exhibit A1).⁶ Avoidable hospital admissions have been used as a metric to assess a range of important clinical and economic topics, from international comparisons of the quality of primary care to the impact of health insurance expansion on primary care and the health costs of the US economic downturn in 2008—which was associated with increases in nonelective (urgent and unscheduled) hospital and emergency room visits for preventable conditions.^7–9

We used avoidable hospital admissions as a metric to determine the quality of primary care and an indicator of resource use that could be avoided by improved primary care management for people with diabetes. Improving access to primary care may increase the use of and spending on outpatient care. But it could also reduce episodes of hyper- or hypoglycemia that require hospital admissions and the sequelae of uncontrolled diabetes, such as coronary heart disease and stroke. In this case, outpatient visits and subsequent avoidable admissions would be negatively correlated. If people learn to control diabetes and stay healthy enough to avoid any need for hospitalizations for diabetes-related issues, outpatient management would be preferable to inpatient admissions, using fewer resources to achieve the same or better health outcomes.

Policy interventions have sought to reduce avoidable hospital admissions by increasing access to outpatient care through multiple means, such as reducing financial barriers via subsidies, improving out-of-hours care, and offering a wider range of outpatient services and telemedicine.¹⁰ These policy interventions targeted at outpatient spending may increase the net value of care if savings in inpatient spending more than offset the increase in outpatient spending. One study found that charging patients more for drugs and doctor visits led to a substantial “offset” in terms of increased hospitalizations, with those offset effects concentrated among patients with diabetes, hypertension, and other chronic diseases.¹¹

However, outpatient visits and inpatient admissions could be positively correlated if the number of outpatient visits increased with the severity of the condition—so that a higher number of outpatient visits foreshadowed a subsequent hospital admission instead of preventing one. For example, one study found that greater outpatient spending was associated with more hospital admissions: A $100 increase in outpatient spending was associated with a 1.9 percent increase in the probability of having any inpatient spending, and a 4.6 percent increase in such spending.¹² Thus, the correlation between outpatient visits in a given year and the probability of an avoidable admission in the subsequent year is an empirical question that we examine in this article.

Our objective was to analyze avoidable admissions for diabetes-related complications and related spending patterns in Japan, Singapore, and Hong Kong, with comparisons to rural and peri-urban areas under the jurisdiction of Beijing—one of China’s three megacities (independent cities that are not part of a province). To do this, we estimated the avoidable admission rates among people diagnosed with diabetes in these four study sites in East Asia, and we compared the rates to those in other middle- and high-income countries. We also estimated the medical spending associated with avoidable admissions. Finally, we conducted an empirical analysis of whether outpatient visits and inpatient admissions were positively or negatively correlated in our study samples, and whether that pattern differed across study sites.

Study Data And Methods

Three of the study sites (Japan, Singapore, and Hong Kong) are high-income areas, while the rural and peri-urban territory under the jurisdiction of Beijing is considered a middle-income area. Appendix Exhibit A2 provides summary statistics about the four study sites.⁶

Data Sources

We analyzed a uniquely comprehensive collection of international data sets that includes detailed data on medical care utilization, spending, and clinical outcomes—drawn from different sources for each study location. A summary table of the different data sets is provided in Appendix Exhibit A3, and a more detailed description of the study sites appears in Appendix Exhibit A4.⁶

Japan:

Analyses for Japan were based on administrative claims data from the Japan Medical Data Center. This data set consists of claims of workers employed at large corporations. Residents of Japan receive insurance coverage from employers or participate in a national health insurance program administered by local governments. Health care services are provided through a mixture of public and private hospitals and clinics. Data used in this analysis are only for people with employer-sponsored insurance and thus cannot be considered nationally representative.

This data set has been used in several other analyses of Japan’s medical system.^13,14 It includes detailed information on medical care and prescription drug use in both outpatient and inpatient settings, all associated medical spending, and the results of annual health screening checkups for over 500,000 Japanese employees in the period 2005–14. We focused on people with diabetes, defined as people who had medical claims for diabetes (that is, claims with codes E10, E11, E13, or E14 from the International Statistical Classification of Diseases and Related Health Problems, Tenth Revision [ICD-10]), in two consecutive years.

Singapore:

Data for Singapore were obtained from the administrative data sets of the Ministry of Health. Singapore has a mixed public-private health care system, with public hospitals providing the majority of inpatient care (85.5 percent of total bed days and 76.0 percent of admissions) and private-sector general practitioners providing up to 80.6 percent of primary care.^15,16 The administrative data sets are comprehensive in their coverage of visits to providers and admissions to public health care institutions. They also capture all hospitalizations and day surgeries in private hospitals, as well as selected outpatient admissions that are covered under Medisave (the national medical savings account program) and MediShield (the national health insurance program). Data for outpatient care at private specialist hospitals and primary care visits at private primary care providers are generally limited, except for primary care visits covered under the Community Health Assistance Scheme (CHAS). Introduced in 2012, CHAS is a program that provides subsidies to lower- and middle-income households. Patients with diabetes in the administrative data sets available to us were defined by diagnosis codes from the International Classification of Diseases, Ninth Revision (ICD-9), (for diagnoses before 2012) and ICD-10 (for 2012 and after). Except for CHAS, which started in 2012, data for Singapore were available for the period 2008–13.

Hong Kong:

The electronic health records of Hong Kong’s Hospital Authority (using its clinical management system) contain information about all of the 6.4 million unique individuals who used Hospital Authority services in the period 2006–14. Like Singapore, Hong Kong has a mixed public-private health care system. Public-sector services, provided by the Hospital Authority, account for the majority of inpatient care (90 percent of total bed days and 80 percent of admissions), 50 percent of specialist outpatient care, and 30 percent of first-contact outpatient services.^17,18 The public system is highly subsidized, with the government providing 94.2 percent of the funding. People diagnosed with diabetes were identified using the WHO’s 2011 guidelines on hemoglobin A1c (HbA1c), fasting plasma glucose, and glucose tolerance tests; the American Diabetes Association’s 2015 guidelines on random plasma glucose; recorded diagnosis codes (from the ICD-9 and International Classification of Primary Care, second edition [ICPC-2]); and filled prescriptions.¹⁹

Beijing:

The claims data for people under Beijing’s jurisdiction covered by the New Cooperative Medical Scheme (NCMS) capture all residents of rural and peri-urban areas living close to one of China’s megacities (2.4 million people in 2014). The NCMS is a government-subsidized voluntary insurance program that was implemented nationwide in the mid-2000s, with coverage rates increasing from 21 percent of the rural population in 2003 to 97 percent by 2011.²⁰ Initially, the benefits covered by the NCMS were relatively thin, with low reimbursement rates and a limited number of reimbursed services. For example, being designed to protect rural people from catastrophic inpatient expenses, the NCMS did not initially reimburse outpatient services. However, benefits have increased over time, varying across counties because local governments have discretion in such matters—within the broad guidelines set by the central government.^21,22

The Beijing NCMS claims data include information about all of Beijing’s NCMS-insured population in 2008–14—that is, 2.4 million residents, of whom about 100,000 have been diagnosed with diabetes. The data set includes details about inpatient and outpatient utilization and associated health expenditures (both reimbursed and out-of-pocket). However, it lacks clinical metrics. ICD-10 coding of diabetes and comorbidities varied in completeness and accuracy, with improvement during the six-year study period.

Avoidable Hospital Admissions

Avoidable hospital admissions were identified using the criteria set by the 2014–15 OECD Health Care Quality Indicators.²³ This source represents an international quality metric to capture variations in health service utilization patterns across countries. Cases were defined as hospital admissions that were neither maternal nor neonatal and that had a principal ICD-9 or ICD-10 diagnosis code for a diabetes-related complication in a specified year (for the list of diagnosis codes, see Appendix Exhibit A1).⁶ We excluded outpatient admissions, patients transferred from another acute care institution, and any admission with an ICD code related to pregnancy, childbirth, the puerperium or newborn period, or neonates.

Statistical Analysis

After we estimated the avoidable hospital admission rate using the criteria set by 2014–15 OECD Health Care Quality Indicators,²³ we compared these rates to those of other middle- and high-income countries, adjusting for age and sex according to the OECD standard population. To account for the differing prevalence of diabetes across study sites, we also estimated the avoidable admissions rate for people diagnosed with diabetes (that is, the number of diabetes-related avoidable admissions in given year, divided by the number of people in the same site and year who had been diagnosed with diabetes). Because defining avoidable admissions according to the OECD definition requires consistent coding of four-digit ICD-10 codes, we could not apply the detailed exclusion criteria to patient samples in Japan or Beijing. As many as 75 percent of inpatient observations for diabetes in the Japan Medical Data Center data are missing the fourth digit, so we used three-digit codes and considered E10, E11, E13, and E14 as indicating avoidable hospitalizations. For Beijing, we included any hospital admission for which diabetes was listed as one of the primary diagnoses. We used complete case analysis, with no attempt to impute missing data.

Spending on diabetes-related avoidable admissions was measured in 2013 purchasing power parity–adjusted US dollars. For comparisons of utilization and spending, estimates were age- and sex-standardized to the Hong Kong 2014 population ages 15–79, as the Japanese data set was limited to people younger than age eighty.

We also estimated the correlates of an avoidable admission, specifically examining whether the number of outpatient visits in 2012 was positively or negatively correlated with the probability of an avoidable admission in 2013 at the level of the individual patient. To perform this analysis, we used a probit regression model that controlled for age, age squared, sex, smoking status, HbA1c (a marker of diabetes control), and Elixhauser comorbidities (for the list of comorbidities, see Appendix Exhibit A5).⁶ We also included the Charlson Comorbidity Index (a marker of severity) in the regression model, as a sensitivity analysis.

Ethics Approval

Our analysis of data for Japan was reviewed and approved by the Stanford University Institutional Review Board (IRB). IRB approval was not required for Singapore, as all of the data we used had been deidentified. Our analyses of data for Hong Kong were approved by the seven Hospital Authority cluster IRBs, and of data for Beijing were approved by the Peking University Health Science Center IRB.

Limitations

Despite its numerous strengths—such as large population samples, with longitudinal follow-up across diverse health systems—our study had several limitations. First, the study sites differed in the scope of the data available and the representativeness of the sample. For example, the Japanese data came from insurers of workers employed at large corporations, thus underrepresenting women and completely missing the population of Japanese with diabetes ages eighty and older. Therefore, to enhance international comparability, we did not use the worker claims data to estimate the national avoidable admissions rate but rather cited the OECD measure of the avoidable hospital admission rate for Japan, which captures appropriate age- and sex- standardization for the entire adult population. Although all of our analyses were conducted at the level of the individual patient, there may be differences in data collection and sampling of individuals across national (in the case of Singapore), municipal (Hong Kong), and regional (rural and peri-urban Beijing) or sectoral (corporate insurers in Japan) data sets. Outpatient visits in Singapore and Hong Kong may also be underestimated, as the data did not fully capture private outpatient visits. Data from Beijing included all rural residents insured by the NCMS, but data for urban residents were not available—and rural and urban residents may be very different.

Second, sample representativeness may also have varied over the study period. Thus, trends must be interpreted with caution.

Third, differences in coding practices may have affected the comparability of our results. There was a sharp increase in avoidable hospital admission rates and associated spending after Singapore switched to ICD-10 codes in 2012.

Finally, differences in avoidable admission rates may reflect access to care and the level of diagnosed diabetes. Japan, Singapore, and Hong Kong offer essentially universal access to highly subsidized specialist care, while screening and treatment coverage in China vary widely by locality and insurance plan. For example, one study estimated that nationally in mainland China, only 25.8 percent of people with diabetes received treatment, and only 39.7 percent of those treated achieved adequate glycemic control.³

Study Results

To account for differences in the population prevalence of diabetes and the rate of diagnosis of diabetes across our four study sites, we used age- and sex- standardized avoidable hospital admission rates and health care utilization among people diagnosed with diabetes. The standardized avoidable admissions rates for diabetes-related complications per thousand people with diabetes varied across study sites, from 22.8 in Singapore in 2009–11 to 126.2 in rural and peri-urban Beijing in 2008 (Exhibit 1). Among our sample of working-age Japanese adults diagnosed with diabetes, there was a large decrease in the rate of avoidable admissions from 2006–08 (58.9) to 2012–14 (23.3). This trend was also observed in Beijing, where rates fell from 126.2 in 2008 to 82.8 in 2012–13. In contrast, the avoidable admissions rates in Singapore and Hong Kong increased over time.

In concordance with the trends in the avoidable admissions rates, the percentages of people with diabetes who experienced an inpatient admission were halved in Japan (from 10.8 percent in 2006–08 to 5.3 percent in 2012–14) but fell only slightly in Singapore (from 20.2 percent to 18.5 percent) and Hong Kong (from 24.1 percent to 23.4 percent) in the same period. People with diabetes in Japan had the highest number of yearly outpatient visits throughout the study period, followed by rural and peri-urban Beijing after 2008. Compared to other study sites, Japan had a low proportion of people with diabetes who required an inpatient admission and a high rate of outpatient visits. Inpatient length-of-stay increased slightly over time in Japan (from 4.1 to 5.1 days), remained stable in Singapore (at 6.6 or 6.7 days), and decreased in Hong Kong (from 6.7 to 5.3 days).

Age- and sex-standardized mean annual medical expenditures for people with diabetes increased over time in Japan, Singapore, and rural and peri-urban Beijing but fell in Hong Kong (Exhibit 2). Mean annual expenditures on inpatient care for people with diabetes followed the same pattern.

Mean annual spending on avoidable hospital admissions increased in Singapore (from $173 to $356) and Hong Kong (from $262 to $284) but fell in Japan (from $431 to $226). As expected, annual medical expenditures for people with an avoidable admission were vastly greater than those for people without such an admission: about six to nine times greater in Japan and Singapore and more than twenty times greater in Hong Kong.

The OECD-standardized avoidable hospital admission rates for diabetes-related complications per 100,000 people over the period 2008–13 estimated from our data ranged from 213.8 to 380.2 in Singapore, from 129.0 to 154.1 in Hong Kong, and from 211.1 to 271.6 in rural and semi-urban Beijing (Exhibit 3). The rate in Hong Kong was consistently lower than the rates in rural and peri-urban Beijing and Singapore during the study period, and comparable to the rates for both Japan and the OECD average. The avoidable admission rates in rural and peri-urban Beijing and in Singapore were higher than the OECD average rate, though different sites might not be directly comparable due to differences in data sources and versions of ICD codes.

In Japan, Singapore, and Hong Kong, the number of primary care outpatient visits in 2012 was negatively correlated with the probability of an avoidable admission in 2013, though the correlation was not significant for Japan (Exhibit 4). A positive correlation for specialist outpatient visits was observed in Japan, Singapore, and Hong Kong. Our sensitivity analysis that controlled for the Charlson Comorbidity Index yielded consistent results (Appendix Exhibit A8).⁶

Discussion

Our study shows how policy makers might be able to measure dimensions of health care access and quality for chronic disease management in diverse economies that have different institutional contexts but strikingly similar challenges as a result of population aging and increased health care spending.

As we expected, health care use varied considerably across the four study sites. Annual avoidable hospital admission rates for diabetes-related complications per thousand people in the latest part of the study period ranged from 23.3 in Japan to 82.8 in rural and peri-urban Beijing. The average number of outpatient visits and outpatient spending were higher in Japan than in Singapore, Hong Kong, or rural and peri-urban Beijing. Among the high-income study sites, inpatient spending was lower in Japan than in Singapore or Hong Kong, which may reflect the fact that the Japanese sample included only employed individuals—who may be in better health than nonemployed individuals within the same age range. The average level of outpatient spending was higher than that of inpatient spending in Japan, while the converse was observed in Singapore, Hong Kong, and rural and peri-urban Beijing.

Our results provide empirical evidence to support the hypothesis that a higher number of primary care outpatient visits in a given year is correlated with a lower likelihood of a diabetes-related avoidable hospital admission in the following year. In Japan, Singapore, and Hong Kong, we found that outpatient primary care visits were negatively associated with avoidable admissions, while in rural and peri-urban Beijing overall outpatient visits were negatively associated with avoidable admissions.

These findings are consistent with the view that outpatient visits can reduce avoidable inpatient admissions. In contrast, specialist outpatient visits were positively correlated with a subsequent avoidable hospital admission in Japan, Singapore, and Hong Kong, which indicates that the number of specialist outpatient visits is a predictor of an avoidable admission. This correlation probably arises because visits to outpatient specialists provides indications about the severity of the condition not captured by HbA1c levels and Elixhauser comorbidities.

Policy Implications

These findings suggest that interventions to improve primary care management of diabetes and reduce avoidable hospital admissions in these four East Asian sites, and elsewhere, hold promise to save health care resources without compromising—and perhaps even improving—the quality of care. Various policy interventions have sought to reduce avoidable admissions in this way. Interventions that improve access (such as providing more out-of-hours care), provide greater continuity of care with a general practitioner, and increase adherence to outpatient management regimens may be associated with lower rates of avoidable hospital admission. However, some studies have yielded mixed results, and most have been in Western settings.¹⁰ Policy tools to increase primary care access include lowering patient copayments to reduce financial barriers. In East Asia, where primary care is often perceived to be weak, and many patients exhibit a preference for specialist care, improving the quality and status of primary care—or better, integrating primary and secondary care—may help address the cultural and system-driven barriers to effective primary care management of chronic disease. Other ways to increase outpatient management and reduce avoidable admissions might be to reduce the direct and indirect costs of adherence through the use of telemedicine; physician “nudges” that remind patients of the importance of adherence; medication reminders (for example, text messages); and monitoring patients with wearable devices, perhaps specifically targeted to high-risk populations. Such policy interventions may improve the net value of care if savings in inpatient spending more than offset the increase in outpatient spending.

Conclusion

We found significant variations in diabetes-related avoidable hospital admission rates across four study sites in East Asia. Spending on these admissions increased from 2006 to 2014 and is substantial: Annual medical expenditures for people with an avoidable admission were six to twenty times those for people without such an admission. We also found empirical evidence that a higher number of outpatient visits (for primary care in Japan, Singapore, and Hong Kong, and for all outpatients in rural and peri-urban Beijing) in a given year was negatively correlated with the probability of an avoidable admission in the following year. In East Asia, as in other parts of the world, policy interventions that increase access and adherence to effective outpatient management potentially improve health outcomes while realizing cost savings from reduced avoidable hospital admissions.

ACKNOWLEDGMENTS

Brian Chen, Karen Eggleston, and Toshiaki Iizuka gratefully acknowledge funding for this research from the Kikawada Foundation. The authors also thank the Japan Medical Data Center for access to its proprietary data for this research and the Hospital Authority (Hong Kong) for its assistance in providing data.

NOTES

• 1 World Health Organization. Global report on diabetes [Internet]. Geneva: WHO; 2016 [cited 2017 Aug 30]. Available from: http://apps.who.int/iris/bitstream/10665/204871/1/9789241565257_eng.pdf?ua=1 Google Scholar
• 2 NCD Risk Factor Collaboration (NCD-RisC). Worldwide trends in diabetes since 1980: a pooled analysis of 751 population-based studies with 4.4 million participants. Lancet. 2016;387(10027):1513–30. Crossref, Medline, Google Scholar
• 3 Xu Y, Wang L, He J, Bi Y, Li M, Wang Tet al. Prevalence and control of diabetes in Chinese adults. JAMA. 2013;310(9):948–59. Crossref, Medline, Google Scholar
• 4 International Diabetes Federation. IDF diabetes atlas—7th edition [Internet]. Brussels: IDF; [cited 2017 Aug 30]. Available from: http://www.diabetesatlas.org/ Google Scholar
• 5 Marshall M, Klazinga N, Leatherman S, Hardy C, Bergmann E, Pisco Let al. OECD Health Care Quality Indicator project. The expert panel on primary care prevention and health promotion. Int J Qual Health Care. 2006;18(Suppl 1):21–5. Crossref, Medline, Google Scholar
• 6 To access the Appendix, click on the Details tab of the article online.
• 7 Organization for Economic Cooperation and Development. Health at a glance 2015 [Internet]. Paris: OECD; 2015 Nov 4 [cited 2017 Aug 30]. Available from: http://www.oecd.org/health/health-systems/health-at-a-glance-19991312.htm Google Scholar
• 8 Kolstad JT, Kowalski AE. The impact of health care reform on hospital and preventive care: evidence from Massachusetts. J Public Econ. 2012;96(11–12):909–29. Crossref, Medline, Google Scholar
• 9 Currie J, Tekin E. Is there a link between foreclosure and health? Am Econ J Econ Policy. 2015;7(1):63–94. Crossref, Medline, Google Scholar
• 10 Purdy S. Avoiding hospital admissions: what does the research evidence say? [Internet]. London: King’s Fund; 2010 Dec [cited 2017 Aug 30]. p. 7–8. Available from: https://www.kingsfund.org.uk/sites/files/kf/Avoiding-Hospital-Admissions-Sarah-Purdy-December2010_0.pdf Google Scholar
• 11 Chandra A, Gruber J, McKnight R. Patient cost-sharing and hospitalization offsets in the elderly. Am Econ Rev. 2010;100(1):193–213. Crossref, Medline, Google Scholar
• 12 Kaestner R, Lo Sasso AT. Does seeing the doctor more often keep you out of the hospital? J Health Econ. 2015;39:259–72. Crossref, Medline, Google Scholar
• 13 Fukushima K, Mizuoka S, Yamamoto S, Iizuka T. Patient cost sharing and medical expenditures for the elderly. J Health Econ. 2016;45:115–30. Crossref, Medline, Google Scholar
• 14 Iizuka T. Physician agency and adoption of generic pharmaceuticals. Am Econ Rev. 2012;102(6):2826–58. Crossref, Medline, Google Scholar
• 15 Ministry of Health Singapore. Health facilities [Internet]. Singapore: MOH; [last updated 2017 Apr 5; cited 2017 Aug 30]. Available from: https://www.moh.gov.sg/content/moh_web/home/statistics/Health_Facts_Singapore/Health_Facilities.html Google Scholar
• 16 Ministry of Health Singapore. Primary care survey 2010 [Internet]. Singapore: MOH; [last updated 2016 Feb 4; cited 2017 Aug 30]. Available from: https://www.moh.gov.sg/content/dam/moh_web/Publications/Reports/2014/Primary%20Care%20Survey%202010%20report.pdf Google Scholar
• 17 Census and Statistics Department. Thematic household survey [Internet]. Hong Kong: The Department; 2010 Dec [cited 2017 Aug 30]. (Report No. 45). Available from: http://www.statistics.gov.hk/pub/B11302452010XXXXB0100.pdf Google Scholar
• 18 Tin KYK, Tsoi PKO, Lee YH, Chong DSY, Lam DWS, Yeung AYTet al. Hong Kong domestic health spending: financial years 1989/90 to 2011/12. Hong Kong Med J. 2015;21(3, Suppl 3):1–24. Medline, Google Scholar
• 19 Quan J, Li TK, Pang H, Choi CH, Siu SC, Tang SYet al. Diabetes incidence and prevalence in Hong Kong, China during 2006–2014. Diabet Med. 2017;34(7):902–8. Crossref, Medline, Google Scholar
• 20 Meng Q, Xu L, Zhang Y, Qian J, Cai M, Xin Yet al. Trends in access to health services and financial protection in China between 2003 and 2011: a cross-sectional study. Lancet. 2012;379(9818):805–14. Crossref, Medline, Google Scholar
• 21 Eggleston K, Bundorf MK, Triyana M, Wang Y, Zhou S. Health insurance and chronic disease control: quasi-experimental evidence from hypertension in rural China. In: Burns LR, Liu GG, editors. China’s healthcare system and reform. Cambridge (UK): Cambridge University Press; 2017. p. 321–34. Crossref, Google Scholar
• 22 Meng Q, Fang H, Liu X, Yuan B, Xu J. Consolidating the social health insurance schemes in China: towards an equitable and efficient health system. Lancet. 2015;386(10002):1484–92. Crossref, Medline, Google Scholar
• 23 Organization for Economic Cooperation and Development. Definitions for health care quality indicators: 2014–2015 HCQI data collection [Internet]. Paris: OECD; [cited 2017 Sep 28]. Available for download from: http://stats.oecd.org/fileview2.aspx?IDFile=62f94ae6-180c-4e4b-9a22-b030ddadfd35 Google Scholar