Mortality In Rural China Declined As Health Insurance Coverage Increased, But No Evidence The Two Are Linked

The widespread provision of health insurance coverage helps countries improve the health of their populations and also protect people from catastrophic health spending. The United Nations endorses universal health coverage—to “ensure healthy lives and promote well-being for all at all ages”—as one of the Sustainable Development Goals, a set of international goals that aim to end poverty, tackle climate change, and foster sustainable social and economic development by 2030. ¹ Evidence on health insurance coverage and mortality from low- and middle-income countries is limited, although one recent study of the relationship between publicly financed health spending and mortality rates across 153 countries found that public spending reduces mortality, particularly among children in low-income countries. ²

In China over the past fifteen years, policies to improve risk protection for the rural majority and move toward universal health coverage built upon the legacy of the Cooperative Medical Schemes in place under central planning (before the 1980s). Lack of health coverage during the 1980s and 1990s had led to large amounts of out-of-pocket spending on health services and contributed to illness-induced poverty. ^3,4

Pilot programs for the government-subsidized voluntary insurance known as the New Cooperative Medical Scheme (NCMS) began in 2002–03. County-based health insurance risk pools were then implemented nationwide, with insurance coverage in rural areas increasing from 21 percent in 2003 to 97 percent in 2011. ³ At first, coverage was relatively thin (that is, reimbursement rates for health care expenses were low, and only a limited number of health services were reimbursed), but coverage generosity has increased over time. ⁴ Financed by individual contributions combined with subsidies from central and local governments, NCMS benefits and premiums vary across counties because local governments have had discretion—within broad guidelines set by the central government—in determining covered benefits, cost sharing, and premium contributions.

China’s dramatic expansion of health insurance coverage might have been expected to contribute substantially to the recent decreases in mortality, especially among the rural poor, and thus to the closing of the survival gap between rural and urban Chinese. However, few studies to date have found any impact of the insurance expansion on health or survival, despite some evidence of increased use of preventive services. ^5–9 Most studies of the NCMS have focused on the early years of implementation, a small sample of counties, or both. Xiaoyan Lei and Wanchuan Lin ⁶ found that the NCMS did not improve health as measured by self-reported health status or by injury or illness in the previous four weeks, although the NCMS did increase the likelihood of receiving preventive services. Similarly, Yuyu Chen and Ginger Jin ⁷ found no impact of NCMS coverage on child or maternal mortality. Using data from 2005 and 2008, Lingguo Cheng and coauthors ⁸ found no evidence that the NCMS had reduced mortality among previously uninsured elderly people. Karen Eggleston and coauthors ⁹ found that the NCMS had increased the use of preventive services and the likelihood of having hypertension under control, but they did not examine mortality.

In this study we examined the rollout of the NCMS in rural China, using data collected in the period 2004–12 on cause-specific mortality from seventy-two rural county-level administrative units (hereafter, counties) covered by the Chinese Center for Disease Control and Prevention’s Disease Surveillance Point (DSP) system. DSP surveillance sites are counties that are officially designated either as rural ( xian , or county) or urban ( qu , or district). We used differences across xian in the timing of the introduction of the NCMS to provide evidence on whether insurance expansion contributed to mortality reductions in rural China. We focused on the rural elderly (people ages sixty and older) as the group most likely to have improved survival because of the NCMS, and on the broader set of counties that adopted the NCMS after the initial pilot programs of 2003–05. Our study improves upon earlier estimates of the program’s effects by providing a longer-term and more geographically comprehensive assessment of how China’s progress toward universal health coverage has affected rural population health as measured by survival.

Study Data And Methods

Design And Data Sources

Our study used data from the DSP system, a rich data source on cause-specific mortality that has not yet been used to study the impact of insurance expansion (additional details about the primary data sources and analytic methods are available in the online Appendix). ¹⁰ The DSP system was established in 1978 with a pilot study in Beijing and expanded through the 1990s. ^11,12 In 2004 the number of surveillance sites was expanded again to 161 (97 counties and 64 districts), and all townships or subdistricts were sampled for each surveillance site—which raised the number of people covered to seventy-three million. ^11,13 The data are nationally and regionally representative ^11–14 and have been used in the Global Burden of Disease Study to measure mortality rates in China. ^15–17

The county-year population by age and sex is based on the registered population in the DSP system. Before 2008, information about each death in the DSPs was collected, cleaned, and compiled by local Centers for Disease Control and Prevention and reported to the Chinese Center for Disease Control and Prevention by electronic data file transfer, along with data on the DSP’s corresponding registered population ( hukou population, or huji renkou ).

However, since 2008 this reporting system has become incorporated into the China Information System for Disease Control and Prevention—the world’s largest Internet-based disease reporting system with a focus on infectious disease. Starting in 2008, the system has incorporated in real time all information about deaths in the DSP population catchment areas, as well as data on the DSP’s residential population ( changzhu renkou ). Because this population is considered more important in the control and prevention of infectious diseases than the formally registered population ( huji renkou ), the Chinese Center for Disease Control and Prevention obtains data on each DSP’s total residential population from the National Bureau of Statistics and enters the population data into the China Information System for Disease Control and Prevention, with yearly updating for the calculation of mortality rates.

To ensure the comparability of mortality rates before and after 2008, we adjusted the 2008–12 population denominators by age and sex as follows: First, we calculated the average ratio of the registered ( hukou ) to the residential ( changzhu ) population in 2004–07 for each DSP by age and sex. Then we multiplied that ratio by the DSP’s residential population in each year thereafter to obtain the estimated registered population for 2008–12 (that is, we weighted the population by the average hukou / changzhu ratio). The average ratio across age-sex groups for the counties in the DSP system was 1.016, which reflects the fact that many rural Chinese migrate to urban areas. We used the registered population as the population denominator, since insurance is based on the place of registration.

To construct our dependent variables, we age-standardized the age-specific mortality rates for each county using direct standardization. ¹⁸ The standardized population by sex and five-year age groups was obtained from China’s 2010 census. ¹⁹ We followed the methodology applied in the Global Burden of Disease Study 2013 ¹⁷ to categorize causes of death (for example, noncommunicable diseases and their subcategories).

To identify the year when the NCMS was implemented in each county in the DSP system, we searched local government websites. For the eighteen counties that did not report that information on their websites, we contacted the local Centers for Disease Control and Prevention. County-specific control variables were obtained from the National Statistics Bureau.

Statistical Analysis

Our data set of information at the county and year levels covers counties in the DSP system that adopted the NCMS during our 2004–12 study period: seventy-six counties in the unbalanced sample, and seventy-two counties in the balanced sample (that is, counties with data available for all nine years of the study period). The Appendix provides details about the construction of the balanced and unbalanced samples (Appendix Exhibit A1), compares the characteristics of these different samples (Appendix Exhibit A3), and discusses our statistical methods and sensitivity analyses in greater detail. ¹⁰

Our dependent variable was the county age-standardized mortality rate by sex and cause of death. We ran separate regressions for men and women and for each age group (ages 20–44, 45–59, and 60 and older) for the overall age-standardized mortality rate and the three primary categories of causes of death (communicable diseases, noncommunicable diseases, and injuries). We also studied an alternative category of mortality specifically designed to capture those causes of death most likely to be affected by health services—conditions amenable to health care, or “health-service-amenable mortality” ^20–24 (see the definitions in Appendix Exhibit A2). ¹⁰ For the population ages sixty and older, we also examined disaggregated mortality from noncommunicable diseases for selected prominent causes of death: cerebrovascular disease, cancer, diabetes, heart disease, and chronic obstructive pulmonary disease.

The independent variable of interest was the NCMS indicator measured with a one-year lag. In other words, if a county adopted the NCMS in 2006, it was coded as having the program starting in 2007, the first full year of enrollment. We estimated regressions that included time-varying county characteristics—per capita gross domestic product (GDP), number of hospital beds, and total population—as well as county and year fixed effects. Standard errors were clustered at the county level.

We analyzed the effects of the NCMS on multiple populations, separately analyzing males and females by age group. We also decomposed the all-cause mortality rates into different components. This approach raised the concern of falsely rejecting the null hypothesis because of multiple comparisons. However, because none of our main results were significant at conventional levels, correcting for multiple comparisons would not change our conclusions.

Limitations

Our study had several limitations. First, any difference-in-differences study is sensitive to changes in how the outcome is measured. A significant limitation in this case arose from our lack of data on the population denominator for the dependent variable: Our having to impute the population denominator for the observations for 2008–12 introduced measurement error.

Second, the county mortality records might not have fully captured China’s migrant worker population, although we mitigated this problem by focusing on mortality rates of people ages sixty and older (because migrant workers are usually younger than sixty).

Third, coding of cause-specific mortality rates inevitably evolved over the study period, especially for causes little known (that is, not well screened for, monitored, or coded) in rural areas previously, such as diabetes mellitus. ¹⁷

Limitations related to the data on explanatory variables—such as the lack of data on individual behavior, enrollment, or health insurance coverage benefits—also should be kept in mind when interpreting our results. We lacked data for county-year combinations on many other aspects of health and medical care that could confound the effects of the NCMS, such as increased government investment in county hospitals and township health centers and the separate financing stream for population health services.

Study Results

Consistent with national trends, the NCMS expanded rapidly in counties in the DSP system during 2004–08 ( Exhibit 1 ). The percentage of the population living in counties in the system that offered the NCMS increased from 18.4 percent in 2004 to 96.8 percent by 2008, with NCMS coverage available to all rural residents by 2012, the end of our study period. Over the same period, the age-standardized mortality rate for rural residents ages sixty and older declined 16 percent for both males and females, with most of the mortality reduction occurring in the period 2004–08 ( Exhibit 2 ). The majority of the survival gains were the result of a reduction in deaths due to noncommunicable diseases such as cerebrovascular disease and chronic obstructive pulmonary disease, rather than a reduction in deaths due to communicable diseases or injuries ( Exhibit 3 ). An alternative aggregation of cause-specific mortality, mortality amenable to health care, declined 8 percent for both males and females, which suggests that access to health care may have played a role in the reduction.

SOURCE Authors’ analysis of data from the Disease Surveillance Point (DSP) system of the Chinese Center for Disease Control and Prevention. NOTE “Exposure rate” is the percentage of the registered population residing in counties in the DSP system that resides in a county offering the NCMS.

SOURCE Authors’ analysis of data from the Disease Surveillance Point system of the Chinese Center for Disease Control and Prevention. NOTE The age standardization is explained in the text.

SOURCE Authors’ analysis of data from the Disease Surveillance Point system of the Chinese Center for Disease Control and Prevention. NOTES The age standardization and mortality amenable to health care are explained in the text. COPD is chronic obstructive pulmonary disease. NCD is noncommunicable disease.

To determine whether the expansion of health insurance coverage contributed to these improvements in survival, we conducted regression analyses that used the variation in NCMS adoption across counties in the period 2004–12. Each column of Exhibit 4 shows the results of a separate regression that includes year fixed effects to control for mortality trends that are similar across counties in China; county fixed effects to control for time-invariant, unobservable differences across counties; and county-year-specific variables to control for economic, demographic, and health care factors specific to that county and year (GDP per capita, number of hospital beds, and total population) that might otherwise confound the effects of the NCMS.

Exhibit 4 shows that the introduction of the NCMS to a county was not significantly correlated with reductions in mortality rates for any of the aggregated or specific causes of death among people ages sixty and older. Although the estimated NCMS coefficient on mortality from heart disease was negative for both men and women (as well as for all age and sex groups, as shown in the Appendix), ¹⁰ it was not significant. The estimated confidence intervals were relatively large. For example, the estimated coefficient for heart disease mortality was −0.508 with a standard error of 0.407 for men, and −0.438 with a standard error of 0.295 for women.

Our results were generally similar—that is, they were not significant, particularly after we accounted for multiple hypothesis testing—across a range of sensitivity analyses: the balanced sample of seventy-two counties with observations in all nine years (Appendix A4); with lagged effects of the NCMS (Appendix Exhibit A5); and with province and year fixed effects (Appendix Exhibit A6). ¹⁰ There were also no significant effects of the health insurance expansion on mortality rates among younger groups once we accounted for multiple hypothesis testing, although the estimated NCMS coefficients were consistently negative for all-cause mortality, mortality from noncommunicable diseases or injuries, and mortality amenable to health care (Appendix Exhibit A8). ¹⁰ The relatively large standard errors suggest that our study may have had insufficient power to detect effects of the NCMS on cause-specific mortality during the study period.

Discussion

Since the introduction of government-subsidized voluntary insurance, health insurance coverage for China’s rural population has increased dramatically. ³ Mortality has also declined substantially in rural China. ^15–17 This study is among the first to examine whether there is evidence that these two trends might be linked. While mortality rates declined among rural residents during our nine-year study period (2004–12), we found little evidence that the NCMS contributed to this decline.

However, we note that the standard errors of our estimated coefficients were large in many cases, relative to the coefficients. For example, the estimated coefficient for all-cause mortality was −0.871 among men ages 60 and older in the balanced sample, with a standard error of 1.586 (Appendix Exhibit A4), ¹⁰ and the estimated coefficient for all-cause mortality was −0.140 among men ages 45–59, with a standard error of 0.219 (Appendix Exhibit A8). ¹⁰ Thus, while we cannot reject the hypothesis that the NCMS had no effect on the mortality rates of rural residents, the relatively large standard errors leave open the possibility that the NCMS had effects on mortality that we could not detect because of limitations in the power of the study design. In addition, any mortality reductions associated with health insurance might arise only after many years of accumulated coverage. While we focused on older adults, hypothesizing that the potential for short-term mortality reductions would be greatest in that segment of the population, it is still possible that a longer follow-up period is necessary to detect changes in mortality.

Previous studies show that the NCMS has increased the use of health services and protection from risk. The literature has consistently documented a positive impact of the NCMS on preventive visits ^6,9 and other health care utilization. ¹⁷ For example, rigorous studies of populations in counties that adopted the NCMS at a relatively early date have found that it increased the use of outpatient and inpatient care ⁵ and of preventive services, ⁶ although some studies have found no impact on the use of formal curative medical services. ⁶ Research using the 2005 and 2008 waves of the Chinese Longitudinal Healthy Longevity Survey has found that NCMS coverage increased the use of medical services among elderly enrollees. ⁸

Studies report mixed evidence on whether the NCMS reduced enrollees’ out-of-pocket spending, with more recent studies more likely to document improved risk protection. For example, neither Adam Wagstaff and coauthors ⁵ nor Lei and Lin ⁶ —all of whom studied the population covered in the period 2003–06—found any reduction in out-of-pocket spending among NCMS enrollees. By contrast, using NCMS implementation data through 2008, Kimberly Babiarz and coauthors ^25,26 found evidence of improved risk protection as measured by whether out-of-pocket spending exceeded the ninetieth percentile of spending among the uninsured and whether medical care was financed by borrowing money or selling assets.

Despite the results regarding increased utilization and some risk protection, little evidence links the NCMS to any improvement in health. Our results are consistent with those of previous studies that have not found survival gains associated with insurance expansion in rural China. For example, using data for 2006 from the China Agricultural Census on 5.9 million people living in eight low-income counties, Chen and Jin ⁷ found that the NCMS had no effect on child mortality or maternal mortality. Cheng and coauthors ⁸ analyzed data for 2005 and 2008 and found some association of the NCMS with declines in mortality, but no evidence that it specifically reduced mortality among the rural elderly who had previously been uninsured.

Our study provides new evidence using data through 2012, covering the period when all of China’s rural areas adopted the NCMS. We did not find any statistically robust impact of the NCMS on mortality for rural Chinese during the study period. However, the relatively wide confidence intervals suggest the use of caution in interpreting our results. We lacked sufficient statistical power to detect some mortality effects that might have already occurred, and some mortality reductions might become manifest only after many years.

Our findings contribute to the growing literature on universal health coverage and health. In a 2013 review of the impact of such coverage, Ursula Giedion and coauthors conclude that “overall, the evidence suggests that [universal health care] schemes can indeed have a positive impact on health status…but that given their nature, impacts on health status are harder to achieve and/or detect. Several studies find mixed evidence or are inconclusive due to unresolved methodological challenges, important study limitations, and sometimes questionable relevance of the outcome variables that are chosen to evaluate impact.” ²⁷

Despite having a design that overcame several of the weaknesses of the previous literature and using the most detailed and representative data set of county-year variations in cause-specific mortality linked to NCMS implementation, our study also had many limitations that should be kept in mind when interpreting the results of our main analysis and those of the numerous sensitivity analyses described in the Appendix. ¹⁰ Future studies that are able to address these limitations with new data sets could provide valuable information about whether expanded coverage and generosity of health insurance have contributed to reductions in mortality and morbidity in rural China.

Conclusion

We found no evidence that the expansion of health insurance to hundreds of millions of Chinese under the New Cooperative Medical Scheme contributed to mortality reductions among rural Chinese adults in the period 2004–12, although continued research is warranted to investigate whether insurance may later show impacts on mortality, especially as coverage for catastrophic expenses increases.

Our findings suggest that population health policies remain central to China’s efforts to increase healthy life expectancy. Policies designed to enhance risk protection for rural Chinese and to integrate insurance programs across urban and rural areas, as emphasized by recent reforms, may be expected to reduce the disparities in health services access and financial risk faced by rural residents, compared to their urban counterparts. However, such policies cannot be viewed as a panacea for closing the survival gap between rural and urban Chinese.

ACKNOWLEDGMENTS

Preliminary results from this study were presented at the International Health Economics Association 2015 biennial congress, Milan, Italy, July 14, 2015, and at a Yale University seminar, New Haven, Connecticut, April 27, 2016. The authors are very grateful to Shuang Zhang, of the University of Colorado, for collaboration in the earlier stages of this research, and to Yingqiao He, a graduate student at Stanford University, for excellent research assistance in collecting the data on the implementation dates of the NCMS. The research presented in this article is that of the authors and does not reflect the official policy of the Chinese Center for Disease Control and Prevention.

NOTES

• 1 United Nations . Transforming our world: the 2030 agenda for sustainable development [Internet]. New York (NY) : UN ; [cited 2017 Jul 28 ]. Available from: https://sustainabledevelopment.un.org/post2015/transformingourworld Google Scholar
• 2 Moreno-Serra R , Smith P . The effects of health coverage on population outcomes: a country-level panel data analysis [Internet]. Washington (DC) : Results for Development Institute ; 2011 Dec [cited 2017 Aug 17 ]. (Working Paper). Available from: http://www.jointlearningnetwork.org/uploads/files/resources/Transitions_in_Health_Financing_Effects_of_Health_Coverage_on_Population_Outcomes.pdf Google Scholar
• 3 Meng Q , Xu L , Zhang Y , Qian J , Cai M , Xin Y , et 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
• 4 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
• 5 Wagstaff A , Lindelow M , Jun G , Ling X , Juncheng Q . Extending health insurance to the rural population: an impact evaluation of China's new cooperative medical scheme . J Health Econ . 2009 ; 28 ( 1 ): 1 – 19 . Crossref, Medline, Google Scholar
• 6 Lei X , Lin W . The New Cooperative Medical Scheme in rural China: does more coverage mean more service and better health? Health Econ . 2009 ; 18 ( Suppl 2 ): S25 – 46 . Crossref, Medline, Google Scholar
• 7 Chen Y , Jin GZ . Does health insurance coverage lead to better health and educational outcomes? Evidence from rural China . J Health Econ . 2012 ; 31 ( 1 ): 1 – 14 . Crossref, Medline, Google Scholar
• 8 Cheng L , Liu H , Zhang Y , Shen K , Zeng Y . The impact of health insurance on health outcomes and spending of the elderly: evidence from China’s New Cooperative Medical Scheme . Health Econ . 2015 ; 24 ( 6 ): 672 – 91 . Crossref, Medline, Google Scholar
• 9 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 LRLiu GG , editors. China’s healthcare system and reform . New York (NY) : Cambridge University Press ; 2017 . p. 321 – 34 . Crossref, Google Scholar
• 10 To access the Appendix, click on the Appendix link in the box to the right of the article online.
• 11 Guo K , Yin P , Wang L , Ji Y , Li Q , Bishai D , et al. Propensity score weighting for addressing under-reporting in mortality surveillance: a proof-of-concept study using the nationally representative mortality data in China . Popul Health Metr . 2015 ; 13 ( 1 ): 16 . Crossref, Medline, Google Scholar
• 12 Yang G , Zheng X , Zeng G , Wang L , Chen Y , Chen A , et al. [ Selection of DSP points in second stage and their representativeness ]. Zhonghua Liu Xing Bing Xue Za Zhi . 1992 ; 13 ( 4 ): 197 – 201 . Medline, Google Scholar
• 13 Zhou M , Jiang Y , Huang ZJ , Wu F . Adjustment and representativeness evaluation of national disease surveillance points system . Disease Surveillance . 2010 ; 25 ( 3 ): 239 – 44 . Google Scholar
• 14 Wang L , Wang L , Cai Y , Ma L , Mao L , Zhou M . Analysis of under-reporting of mortality surveillance from 2006 to 2008 in China . Zhonghua Yu Fang Yi Xue Za Zhi . 2011 ; 45 ( 12 ): 1061 – 4 . Medline, Google Scholar
• 15 Yang G , Wang Y , Zeng Y , Gao GF , Liang X , Zhou M , et al. Rapid health transition in China, 1990–2010: findings from the Global Burden of Disease Study 2010 . Lancet . 2013 ; 381 ( 9882 ): 1987 – 2015 . Crossref, Medline, Google Scholar
• 16 Lozano R , Naghavi M , Foreman K , Lim S , Shibuya K , Aboyans V , et al. Global and regional mortality from 235 causes of death for 20 age groups in 1990 and 2010: a systematic analysis for the Global Burden of Disease Study 2010 . Lancet . 2012 ; 380 ( 9859 ): 2095 – 128 . Crossref, Medline, Google Scholar
• 17 Zhou M , Wang H , Zhu J , Chen W , Wang L , Liu S , et al. Cause-specific mortality for 240 causes in China during 1990–2013: a systematic subnational analysis for the Global Burden of Disease Study 2013 . Lancet . 2016 ; 387 ( 10015 ): 251 – 72 . Crossref, Medline, Google Scholar
• 18 Curtin LR , Klein RJ . Healthy People 2000 statistical notes: direct standardization (age-adjusted death rates) [Internet]. Hyattsville (MD) : National Center for Health Statistics ; 1995 Mar [cited 2017 Jul 28 ]. Available from: https://www.cdc.gov/nchs/data/statnt/statnt06rv.pdf Google Scholar
• 19 Cai Y . China’s 2010 population by sex and age recovered from the census population pyramid [Internet]. Chapel Hill (NC) : University of North Carolina at Chapel Hill ; [updated 2012 Jan 5 ; cited 2017 Jul 28 ]. Available from: http://www.unc.edu/~caiyong/papers/sex.age.2010.pdf Google Scholar
• 20 Charlton JR , Velez R . Some international comparisons of mortality amenable to medical intervention . Br Med J (Clin Res Ed) . 1986 ; 292 ( 6516 ): 295 – 301 . Crossref, Medline, Google Scholar
• 21 Mackenbach JP , Bouvier-Colle MH , Jougla E . “Avoidable” mortality and health services: a review of aggregate data studies . J Epidemiol Community Health . 1990 ; 44 ( 2 ): 106 – 11 . Crossref, Medline, Google Scholar
• 22 Nolte E , McKee M . Measuring the health of nations: analysis of mortality amenable to health care . BMJ . 2003 ; 327 ( 7424 ): 1129 . Crossref, Medline, Google Scholar
• 23 Nolte E , McKee CM . Measuring the health of nations: updating an earlier analysis . Health Aff (Millwood) . 2008 ; 27 ( 1 ): 58 – 71 . Go to the article, Google Scholar
• 24 Nolte E , McKee M . Variations in amenable mortality—trends in 16 high-income nations . Health Policy . 2011 ; 103 ( 1 ): 47 – 52 . Crossref, Medline, Google Scholar
• 25 Babiarz KS , Miller G , Yi H , Zhang L , Rozelle S . China’s New Cooperative Medical Scheme improved finances of township health centers but not the number of patients served . Health Aff (Millwood) . 2012 ; 31 ( 5 ): 1065 – 74 . Go to the article, Google Scholar
• 26 Babiarz KS , Miller G , Yi H , Zhang L , Rozelle S . New evidence on the impact of China’s New Rural Cooperative Medical Scheme and its implications for rural primary healthcare: multivariate difference-in-difference analysis . BMJ . 2010 ; 341 : c5617 . Crossref, Medline, Google Scholar
• 27 Giedion U , Alfonso EA , Díaz Y . The impact of universal coverage schemes in the developing world: a review of the existing evidence [Internet]. Washington (DC) : World Bank ; 2013 Jan [cited 2017 Jul 31 ]. p. 72 . Available from: http://siteresources.worldbank.org/HEALTHNUTRITIONANDPOPULATION/Images/IMPACTofUHCSchemesinDevelopingCountries-AReviewofExistingEvidence.pdf Google Scholar