Decomposing the causes of socioeconomic-related health inequality among urban and rural populations in China: a new decomposition approach

Understanding the temporal trends in health inequalities as well as their determinants are important for informed policy decision making that may reduce such health inequalities [1]. Health inequalities are differences in health that are potentially avoidable, unjust and/or unfair [2, 3]. They are mainly related to demographic and socioeconomic determinants, including age, gender, income and education [4‐6]. Growing evidence indicates that lower socioeconomic status is associated with poorer health [7], and that health inequalities favor high-income groups [8‐10]. Socioeconomic-related health inequities may be influenced by socioeconomic factors directly, or may be explained by other socioeconomic-related factors such as lifestyle factors [11, 12]. Socioeconomic gradients in access to health care also lead to inequalities in health [13]. Access to health services is concentrated among those at the upper end of socioeconomic spectrum [14, 15].

China’s rapid transition from a planned economy to a more market led economy has resulted in dramatic economic growth [16]. However, this period of economic growth has not been associated with an equivalent improvement in health [17]. The growing disparity in health between urban and rural regions and between the rich and the poor has caused dissatisfaction with socioeconomic-related health inequality [18]. The Chinese government has adopted a number of measures to address health inequalities. For instance, the launch of the New Rural Cooperative Medical Scheme (NRCMS) in 2003 was designed to reduce inequity in access to health care for the rural population. Although these reforms have offered some relief to the (rural) poor, concerns remain [19‐21].

In order to achieve socioeconomic-related health equality, it is vital to determine the causes of the socioeconomic-related health inequality. Decomposing socioeconomic-related health inequality can help to uncover specific factors that are potentially modifiable by policy decision makers. The dominant decomposition approach, the Wagstaff decomposition method, was proposed by Wagstaff et al. [22] and has been used extensively in previous studies [9, 23]. However, there are potential concerns with this decomposition method. First, this decomposition method only explains the degree of variation in health rather than the covariance between health and socioeconomic rank [24, 25]. For example, a strict assumption of this method is rank ignorability [24]; Second, this decomposition method is only applicable to absolute inequality indices, such as the absolute concentration index (even though it was developed for the relative concentration index) [24, 26]; Third, this decomposition method imposes many restrictive assumptions, such as rank ignorability and weighting function ignorability, which in empirical practice often are unreasonable to impose [26]; Forth, it is unclear how to actually interpret the parameters, and the so called contributions, within these decompositions [24, 26]. In this paper, therefore, we used a new decomposition approach, called recentered influence function (RIF) regression decomposition method, as proposed by Heckley et al. [26] to decompose socioeconomic-related health inequality in China. RIF regression decomposition has a few important benefits. First, this method explains the causes of socioeconomic-related health inequality by directly decomposing the weighted covariance of health and socioeconomic rank [26]; Second, this method is able to decompose all forms of inequality measures, such as the Erreygers index (EI) [24], the Wagstaff index (WI) [22], the standard concentration index (CI), the absolute concentration index (AC), the attainment-relative concentration index (ARCI), and the shortfall-relative concentration index (SRCI) [26, 27]; Third, this method requires fewer, and less restrictive assumptions than the Wagstaff decomposition method. For example, this method simultaneously relaxes the rank and weighting function ignorability assumptions [26]; Fourth, this new decomposition approach is simple to estimate and the results are easy to interpret [26, 28]. To our knowledge, this is the first study to use this new method to decompose socioeconomic-related health inequality in China. Given the dual structure of urban and rural areas in China, it is necessary to distinguish the different potential causes of socioeconomic-related health inequality in urban and rural populations. Therefore, the objectives of this study were to examine changes in socioeconomic-related health inequality in China from 1991 to 2006; and to explore the potential causes of socioeconomic-related health inequality in urban and rural areas. These findings may provide a reference for the Chinese government to reduce socioeconomic-related health inequality and promote health in China.

Our research may make at least two contributions to the existing literature. First, we used a new decomposition approach to examine socioeconomic-related health inequality in China and compared our findings to previous studies that used the Wagstaff decomposition method. Second, using data from the China Health and Nutrition Survey (CHNS), one of the longest running panel studies, we assessed the changes in health inequality among urban and rural populations for the period 1991 to 2006, the period over which health status data were available.

China’s rapid economic development has led to major differences in social and economic life between urban and rural areas. Therefore, it is better to examine socioeconomic-related health inequality in urban and rural areas separately. Understanding the variation in health across different socioeconomic groups can help illuminate our understanding of the determinants of these inequalities. In the current study, we first compared the distribution of good health across different income groups. We then measured socioeconomic-related health inequality by calculating the conventionally employed CI between 1991 and 2006 in urban and rural areas. Finally, we analyzed the potential causes of these inequalities in urban and rural populations, separately. Several compelling new findings were revealed. First, our finding regarding the positive CI suggests that there was pro-rich health inequality in China. The rich had a greater share of good-health in China. Second, in the majority of the survey years, the CI in rural areas was slightly higher than that in urban areas, which suggests that health inequality in rural areas was slightly more pronounced than that in urban areas, and a higher degree of pro-rich health inequality existed in rural areas. Third, overall health inequality rose from 1991 to 2006. Fourth, through the decomposition of socioeconomic-related health inequality, we found that the potential causes of socioeconomic-related health inequality were different in urban and rural areas.

We found that the average level of health status declined from 1989 to 2006 for both urban and rural populations. This may be because urbanization triggers changes in occupational activities, social structures and socioeconomic status that may lead to illnesses such as cardiovascular disease, neuropsychiatric disorders, and other non-communicable chronic diseases [40]; or because environmental quality, including air and water pollution, contributes to lowering disease prevalence for both urban and rural areas [40, 41]; or because the transition to western style diets and sedentary lifestyles leads to a decline in health. One previous study reported that, today, China is faced with a set of health issues, such as the risks of hypertension, the impact of environmental pollution and smoking, and chronic and degenerative diseases [42]. Due to the absence of the relevant data, we were not able to examine the trend in various diseases or the trend in outdoor air pollution over the study period, however, we did find lifestyles and home environment had an impact on inequality in health.

Our finding that there existed pro-rich health inequality in China was consistent with previous studies on socioeconomic-related health inequality in China [43, 44]. This may be because that the rich can access more health resources. Zhang et al. [45] using the data from Chinese National Health Services Survey (NHSS) in 2008, found that there were substantial differences in health care utilization between the rich and the poor. Moreover, Zhou el al. [46] using the data from the Chinese NHSS in 1993, 1998, 2003 and 2008, found that with the same needs for health care, rich rural residents used more health care services than poor rural residents. In our study, we found that health inequality increased over the last two decades.

We also found that although in different years, there were small differences in the CI between rural and urban areas, the absolute value of the differences was less than 0.02. Consequently, these differences over the study period were small. In the majority of the survey years, the CI in rural areas was slightly higher than that in urban areas, which means that health inequality in rural areas was slightly more pronounced than that in urban areas. This finding was consistent with the findings by Xie [43]. Greater health inequality in rural China may be associated with the characteristics of rural areas. In China, health care systems were different for urban and rural populations. For the rural population, with the collapse of the Cooperative Medical System (CMS) in 1981, private health insurance was the only option for a long time [47, 48]. Private health insurance was expensive and unaffordable for most rural residents [49]. Financial barriers may widen health inequality for rural populations, especially for the rural poor. Despite the introduction of the New Rural Cooperative Medical Scheme (NRCMS) in 2003 for the rural population, it seems to have had a limited impact on improving access to formal care for the poor [50].

We also revealed some important causes of socioeconomic-related health inequality for urban and rural populations. Income was a common factor in reducing health inequalities in urban and rural areas in our study. In previous studies, income was often found to be associated with health and high-income groups tended to report good health [7, 12, 36, 51]. Zhou et al. [52] using data from the Chinese NHSS conducted in 2003 and 2008, reported strong pro-rich inequity in access to inpatient utilization in rural China. They found that income was the main determinant of this pro-rich inpatient utilization inequity, as wealthier people could pay for more services, and therefore, used more services regardless of need. Consequently, general improvements in income helped to reduce inequity in inpatient utilization. Another study also found that being poor prevented people from seeking care [53]. Thus, in order to promote health equality, it is important to provide support for the poor.

The level of education had a significant effect on health inequality for both urban and rural populations. However, secondary education helped to reduce inequality, while tertiary education promoted inequality in our study. Previous studies on the impact of education on health have yielded different findings. Cott et al. [54] found that those with lower level of education were less likely to report excellent or very good health. Prus [55] compared the determinants of SRH across the United States and Canada and found that the education gradient was steeper in the U.S compared with Canada. Yang and Kanavos [44] reported that education made an important contribution to total health inequality. However, because Yang and Kanavos used the Wagstaff decomposition method, it is unclear how to actually interpret the contribution. Moreover, this method only focuses on the impact of education on health rather than the impact of education on socioeconomic-related health inequality. Chen et al. [9] assessed income-related health inequality and health achievement in children in China and found middle school enrollments could promote health achievement, but primary school enrollments showed no influence.

In our study, respondents who were currently working were more likely to positively impact health inequality than those who were not working, which is consistent with the finding by Xie [43]. It is possible that people who were engaged in work were richer compared with those without a job, so people who had a job were healthier [43]. Thus, work status further strengthened socioeconomic-related health inequality. Previous studies also found that having a job increased the likelihood of reporting excellent/good health; and work status also made a positive contribution to total health inequalities [43, 44]. In our study, the impact of work status on health inequality was greater in urban areas compared with that in rural areas. As was shown in the result section, the rural respondents who were currently working included those who were engaged in farm work and non-farm work. This results in a higher proportion of respondents reporting that they were currently working (including farm and non-farm work) than those in urban areas. Because the majority of the rural respondents reporting that they were currently working, were actually engaged in farm work, which leads to a low socioeconomic status. Thus, the differences in socioeconomic status between those who were currently working and those who didn’t work may be small for rural population; while the differences may be great for urban population. This helps explain why health inequality in rural areas caused by work status was relatively smaller compared with that in urban areas (as was shown in Table 4, the coefficient on work status was smaller in rural areas).

It is worth noting that health insurance in rural and urban areas may also affect health inequalities. Health insurance helped to reduce health inequality for the urban population; while it increased health inequality for the rural population. It is expected that health insurance provides support for urban populations to seek care. The Urban Employee Basic Medical Insurance (UEBMI) scheme for urban workers and the Urban Residents Medical Insurance (URMI) scheme for children and non-working urban residents are the main forms of health insurance in urban areas. However, for the rural population, as we mentioned above, they were uninsured for a long period following the collapse of CMS in 1981, and rural populations only had recourse to private insurance until the introduction of the NRCMS in 2003 [47, 48]. The NRCMS, introduced between 2003 and 2008, aims at providing insurance to rural residents. NRCMS is a voluntary program that covers only those who join. However, evidence has shown that implementation of NRCMS has not improved the health of the rural population. Sun et al. [56] found that the NRCMS had little effect on reducing household health spending. Yang [50] found that the impacts of the NRCMS on improving access to formal care for the poor were limited. Similar findings were reported by Yu et al. [57], who found that under NRCMS inpatient service utilization has increased for high income groups, but there was no significant change for middle and low income groups, and people with higher incomes tended to benefit more than those on lower incomes. Several reasons may explain why NRCMS has had a limited impact on improving access to health services for the poor in rural areas. First, the low reimbursement rate and the high co-payments were an important limit to the success of the NRCMS. Under the NRCMS, the average level of national reimbursement for outpatient care was approximately 10% of overall expenses [58]. Similarly, the use of preventive care was unequally distributed and related to the unequal distribution of income level [50]. Second, low income participants were already burdened with a premium, while substantial co-payments due to the limited coverage further aggravated inequity in health care access [50]. The low reimbursement rate of the NRCMS and the associated financial burden from co-payments prevented the rural poor from seeking care [52, 59‐61]. Thus, it is necessary to develop more comprehensive forms of coverage for outpatient and preventive care; and implement a more comprehensive insurance package to effectively target the rural poor and to provide low income participants with better financial protection. For services not being covered by the NRCMS, using commercial health insurance to improve equity may be an option for the rural populations [50]. Evidence of adverse selection may also help to explain why health insurance in rural areas did not contribute to reducing inequality; i.e., having sick household members increased the likelihood of enrollment in NRCMS [21, 62].

Finally, it should be noted that unhealthy lifestyles could increase health inequality for urban and rural populations. This may be because that people with different socioeconomic status have different lifestyles, thus influencing health [55]. Poor lifestyles were reported to result in ill health [63, 64]. The better home environment contributed to the reduction of health inequality in rural areas, but had no influence on the health inequality in urban areas.

This study had some potential limitations. First, there may exist other potential factors affecting health inequality, which we were unable to examine due to the limitations of the survey. For example, the distance to the medical institution may influence the willingness of people to seek medical care, thus resulting in disparities in health. However, we have incorporated as many potential factors as possible. Second, the data we used were drawn from nine Chinese provinces, representing 42% of total population of China [9]. However, the CHNS is one of the longest running panel studies, which provides an excellent opportunity to examine changes in health inequality for urban and rural Chinese residents over the past two decades. The nine provinces included in the survey are broadly representative of the economic and regional make-up of China. Third, SRH is a subjective measure, which may suffer from potential response bias. However, this limitation cannot be avoided in the absence of an objective measure of health.

In conclusion, our study provides a reference for the development of policies to reduce socioeconomic-related health inequalities. First, the key to addressing socioeconomic-related health inequalities is to increase the income of those with lower socioeconomic status, thereby improving their access to health services. Second, health insurance coverage could be expanded and vulnerable groups (such as older people and rural poor) could be included in the health care security net. Third, based on our findings, the overall health of the population declined over the past two decades and lifestyles and home environment played roles in this process; this finding suggests that more attention could also be paid to improvements in the lifestyles and home environment of residents while striving to improve income and enhancing access to the health care system. It is worth noting that an emphasis on empowering older people and improving healthy lifestyles from a very early age may be also result in cost-savings and reduce health inequality. Finally, some specific suggestions are stressed in order to reduce health inequality for the rural and urban populations. For the rural populations, the reform of health insurance and improvement of flush toilets and access to tap water require particular attention; while for urban populations, it is necessary to advocate for healthy lifestyles in order to prevent obesity and the occurrence of chronic disease.