Introduction

Studies have shown a close relationship between education and health. For example, Conti et al. highlighted a causal effect in explaining differences in many adult outcomes and healthy behaviors, while emphasizing the heterogeneity [1]. Davies et al. showed that remaining in school causally reduced some health risks and mortality but highlighted genetic and socioeconomic confounding factors [2].

However, the relationship between education and health may be affected by historical socio-political contexts [3]. Specifically, the relative size of the effect of education on health may vary across countries due to different educational systems and their sociocultural implications. From a life course perspective, the level of education a person achieves reflects the far-reaching impact of their parents’ socioeconomic status and broader socioeconomic context on their socioeconomic status in young adulthood [4]. Education is also a strong predictor of important resources in adulthood, such as income and occupation [4].

This study compared educational gradients in self-rated health (SRH) in late adulthood in China and Japan, two countries that face an unprecedented pace of population aging [5]. We focused on cohorts aged 63 to 72 years old in 2018 (born between 1946 and 1955) in China and Japan. These cohorts shared similar life trajectories characterized by postwar construction, phenomenal economic growth, a rapid decline in total fertility rate, as well as reductions in child and adult mortality rates, leading to today’s ageing societies. The life experiences of our study cohorts differ from those developed in Western ageing societies, and they may affect how inequalities in educational attainment influence health status in late adulthood. This assertion is supported by a cross-sectional WHO study that identified socioeconomic inequalities in mental disorders in North American and European countries but not in emerging economies [6]. Despite their shared histories, China and Japan have distinct educational systems, with educational attainment in Japan generally being much higher than in China. This structural variation allows us to investigate the relative strength of the relationship between educational inequalities and healthy ageing.

Meanwhile, an increasing number of studies have focused on the underlying mechanisms linking education and health [7, 8]. Socioeconomic status (e.g., occupational status, income), health behaviors (e.g., smoking, physical activity, obesity), and social participation (e.g., volunteer work, community activities) are often considered potential mediators. However, the results have generally been mixed, partly because of the choice of health outcomes, mediator variables, and targeted age groups.

Several single-country studies have examined educational inequalities in health outcomes and health behaviors in China [9, 10] and Japan [11, 12]. Although some cross-country studies have compared Japan and Korea [13] and China and England [14], China and Japan have not yet been compared. Two recent comparative studies examined the relationship between socio-economic status and health across a range of countries, including China and Japan, but they did not investigate the mediators of educational inequalities in health [15, 16]. To our knowledge, this is the first study to compare educational inequalities in health outcomes in later life in China and Japan and examine the potential mediators underlying them.

Building on the existing literature, we hypothesized that a lower educational level would be associated with a higher risk of poor health status in late adulthood in China and Japan. However, the association may be weaker in Japan than China, due to potentially more prominent inequality in educational attainment among the birth cohort in China. All the socioeconomic and health behaviors mediators of interest could play a significant role in this association.

Materials and methods

Results

Descriptive analysis

Table 1 summarizes the distribution of educational attainment and ridit scores in China and Japan. In general, Japanese participants had higher educational attainment than their Chinese counterparts. In China, the proportion of those who had graduated from high school or above was 10.3% for men and 4.0% for women, compared with more than 80% for both sexes in Japan, where final educational attainment was heavily concentrated in high school. In both countries, men had higher educational attainment than women. In China, the illiteracy rate among female participants was 44.9%, which was strikingly higher than that of their male counterparts (12.2%). While this illiteracy rate among older Chinese women is surprising, it is consistent with the findings of previous studies [32, 33]. The ridit scores reflect variations in the distribution of educational attainment between men and women, as well as between China and Japan.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 1. Distribution of educational attainment and ridit scores.

https://doi.org/10.1371/journal.pone.0291661.t001

Table 2 compares the prevalence of poor SRH and potential mediators between China and Japan, with the 95% CI of the difference. Chinese participants rated their health as worse than their Japanese counterparts. In contrast to Japan, women in China assessed their health as poorer than their male counterparts did. While the prevalence of smoking among men was more than twice as high in China as in Japan, the prevalence of smoking among women was much lower than that among men and roughly comparable between the two countries. A higher proportion of Japanese participants did not engage in LTPA, whereas a higher proportion of Chinese participants did not participate in any social activity. It should be noted that CHARLS and LSMAEP used different definitions for LTPA and social participation.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 2. Proportions (%) of each health outcome and potential mediator.

https://doi.org/10.1371/journal.pone.0291661.t002

Regression and mediation analysis

Table 3 summarizes the results of Models 1–3 for men and women in each country. As shown in this table, when we only controlled for covariates in Models 1, RII was 1.69 (95% CI: 1.20–2.39) for men and 1.47 (95% CI: 1.06–2.05) for women in China, compared with 2.70 (95% CI: 2.21–3.28) for men and 2.60 (95% CI: 2.13–3.18) for women in Japan. The results indicated that in both countries, having a lower educational level was associated with a higher risk of poor SRH, but in China, the educational gradient was flatter. Although women in both countries showed somewhat lower educational inequalities in health, the sex difference was not statistically significant. To assess sex-based differences, we included an interaction term between the educational ridit score and females in the regression models for the entire sample. The estimated coefficient of the interaction term was not significant at the 5% significance level (not reported).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 3. Estimated associations among educational attainment, potential mediators, and self-rated health^a.

https://doi.org/10.1371/journal.pone.0291661.t003

The Model 2 results indicated that the educational gradient in each potential mediator was generally flatter in China than in Japan. Notably, in sharp contrast to the results for Japan, neither smoking nor LTPA was associated with educational attainment in China. In both countries, however, social participation was positively associated with educational attainment. Another noteworthy finding is that, for both men and women, reductions in the OR of the educational ridit score from Model 1 (which included only the educational ridit score) to Model 3 (which included both the ridit score and potential mediators) were substantially smaller in China than in Japan. We also found that, in China, only social participation was associated with both educational attainment and SRH; whereas in Japan, social participation and LTPA (and smoking for women) were both associated with educational attainment and SRH.

Based on the estimation results in Tables 3 and 4 presents the proportions of educational inequalities in health explained by each potential mediator, along with their bootstrap-estimated 95% CIs. Consistent with the results in Table 3, social participation functioned as a mediator in the association between educational attainment and health in China. However, income, smoking status, and LTPA did not. In China, the proportion of educational inequality partially mediated by social participation was as low as approximately 10% for both sexes. In Japan, LTPA and social participation mediated a substantial portion of educational inequalities in health. Together, these two factors accounted for approximately 37% and 40% of the educational inequalities for men and women, respectively, which is considerably higher than in China. Meanwhile, income did not mediate educational inequalities in health in both China and Japan. Smoking was identified as a mediator only for Japanese women. In both countries, social participation was a key mediator in the effect of educational attainment. Furthermore, we observed that the total proportion of the association mediated by the four mediators was higher in Japan than in China. Additionally, it is noteworthy that the total mediating effect was not significant among Chinese men.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 4. Estimated proportions (%) of the educational inequalities in self-rated health mediated by each potential mediator^a.

https://doi.org/10.1371/journal.pone.0291661.t004

For robustness check, we re-estimated Models 1 and 3 using SII as a measure of educational inequality and z-standardized SRH scores as the dependent variable. Likewise, we repeated the mediation analysis using alternative approaches to measure educational inequality and construct the health outcome variable. The results of these robustness checks, reported in S1 and S2 Tables, are largely in line with those in Tables 3 and 4, both in magnitude and significance, consistent with prior estimates. In addition, considering the observed importance of social participation as a mediator, we computed the z-score of the total number of social activities participated for each country, and did mediation analysis by replacing a binary variable of no social participation with a continuous variable of the z-score. The estimation results, which are reported in S3 Table, show that the results remained largely the same as those in Table 4.

Discussion

Although the scholarly literature on education and health is growing [1–3, 7–12], previous study did not directly compare educational inequalities in general health in China and Japan, nor have they examined potential mediators underlying this association. Given that both countries are experiencing rapid population ageing but have distinct distributions of educational attainment, we examined the association between educational inequalities and SRH in late adulthood. We summarize the key findings and their implications as follows.

First, our study found strong evidence for an educational gradient in health in both China and Japan, which is consistent with most previous studies [1–3, 9, 11]. Unsurprisingly, lower educational levels were associated with a higher risk of poor SRH in both countries.

Second, Japan had a stronger educational gradient in health than China. Given the remarkable concentration of final educational attainment in high school, employers in Japan have been more likely to recognize individuals with a bachelor’s degree or higher. This was especially true during Japan’s economic miracle era, from the end of WWII to the early 1990s, a period marked by the transition of well-educated Japanese people from universities to the workplace. From a life-course perspective, education in Japan could serve as a ladder of upward intra-generational mobility—a social movement from a disadvantaged socioeconomic position to an advantaged one during one’s lifetime [34]. In this process, the better-educated might also benefit from healthier work and living environments, more opportunities for social participation, and more LTPA, which mirrors the findings in Model 2. By contrast, most Chinese participants were not highly educated. After embracing marketization in 1978, China also experienced an economic miracle, but many jobs in the 1980s and 1990s were in low-skilled, labor-intensive sectors such as apparel and textiles. The major industries in Japan are automobiles, consumer electronics, and other types of skilled manufacturing. Therefore, formal education tends to have less influence on social mobility and health in China than in Japan.

It is intriguing to compare our findings in China and Japan with evidence from other countries. One study suggests that differences in socio-political contexts may explain observed educational variations in general health among European countries [35]. For example, in the 1980s, social policies focused on addressing unemployment and income inequalities rather than educational inequality, was proposed to explain why Nordic, egalitarian countries had greater educational variations in morbidity than other European countries [35]. Similarly, our findings of a flatter gradient in China than in Japan may also be explained by historical socioeconomic background, particularly the importance of educational attainment in employment.

It should be noted, however, that the difference in the educational gradient in health between China and Japan was more pronounced among women than among men. As seen in Model 1 results in Table 3, the 95% CI of RII for Chinese women was below that for Japanese women, while 95% CIs of RII for men were overlapped between the two countries. This finding suggests that Chinese women with higher educational attainment can enjoy more limited health benefit probably due to their employment-related issues, compared to the Japanese counterpart.

Third, the influence of potential mediators on educational gradients in health was more limited in China than in Japan. In China, only social participation mediated a small portion (approximately 10%) of educational inequalities; while household income, smoking, or LTPA did not. In Japan, LTPA and social participation together mediated a sizable portion (approximately 30–36%) of educational inequalities in health. After controlling for these mediators, the difference in the RII of poor SRH between the two countries was substantially reduced, suggesting the importance of mediators in explaining discrepancies in educational inequalities in late-life health between the two countries.

Fourth, our results underscore the importance of social participation as a mediator of educational inequality in health in both countries. As suggested by previous studies [7, 36–38], the mediating role of social participation can be explained by the enhanced quality of life, improved psychological well-being, expanded social support, and engaging in more protective health behaviors for staying physically active. The mediating effect of social participation has also been observed in European countries [35], suggesting that social participation may be a context-free, universal mediating factor.

Fifth, however, we observed that the mediating effect of social participation was more substantial in Japan than in China. It is worth noting that CHARLS and LSMAEP defined social participation in different ways. Previous studies suggested that social activities can vary significantly based on cultural and contextual factors and can be further differentiated into caring-based activities (e.g., grandparenting) and productive-based activities (e.g., dancing in public spaces and playing mahjong) [39]. These various types of social activities may indicate different socioeconomic backgrounds and generate differential health implications [39]. Our measures of social participation may capture variations in the caring and productive components of social activities, potentially leading to a greater mediating effect in Japan than China. More specifically, the Japanese list of social activities contained more caring components than the Chinese counterpart. Notably, the Chinese list omitted caregiving for grandchildren or adults in the same household. Hence, the results may suggest that the caring components of social activities may have stronger mediating effects on the association between education and SRH than the productive components.

This study has several limitations, in addition to the possibility of insufficient control for attrition biases. First, the existence of potential cohort effects warrants caution when generalizing results across cohorts. Our sample was restricted to those aged 63–72 years in 2018; however, the distribution of educational attainment may vary among cohorts. In China, for example, the national education system was restored after the Cultural Revolution ended in 1978 and has undergone profound reform since the Compulsory Education Law took effect in 1986 [40]. Second, potential endogeneity was not addressed. Although formal education is normally completed in young adulthood, making simultaneity bias implausible, the causal effects of health status on the mediators cannot be ruled out. For example, healthy individuals are more likely to engage in social activities. Third, other potential mediators, such as health literacy and dieting habits, might exist but were not examined in this study due to a lack of data. Fourth, due to the vastly different distribution of educational levels between the two countries in our sample, we were unable to harmonize the classification of educational levels. Instead, we applied the ridit score to focus on the relative position rather than the absolute level of educational attainment. Fifth, CHARLS and LSMAEP adopted different definitions of social participation with respect to types of social activities and how respondents participated (alone, with other people, or with an organization). The challenge of harmonizing different definitions between countries makes it difficult to interpret the differential magnitudes of the mediating effects of social participation on educational inequalities in health. Sixth, to estimate the indirect effects of educational inequalities on poor SRH via mediators, we imposed necessary assumptions, which included the absence of exposure-outcome, mediator-outcome, exposure-mediator confounders, or mediator-outcome confounder that is affected by the exposure [41]. However, it is impossible to control for all unobservable but plausible confounders, such as all kinds of childhood adversities that may affect both educational attainment and later-life health. Moreover, given the lengthy lag between educational attainment and social participation, the fourth assumption which requires short time between the exposure and the mediator is concerning [41].

Conclusion

The results underscore that one’s relative position in educational inequalities is a reliable predictor of subjective health in late adulthood in both China and Japan, although the educational gradient was flatter in China. We also found that social participation was a key mediator of educational inequalities in health in both countries. Policy measures to promote social participation are expected to help mitigate educational inequalities in later life health.

Supporting information