Health Inequalities of Intra-Family Expertise Accessibility: Evidence From China


 Objectives

To investigate whether unequal exposure to health-related expertise of intra-family is the root of health inequality in China, and to explore the underlying mechanisms through which health-related expertise shapes health outcome.
Methods

In a representative sample of Chinese adults ages over 18 from the 2017 Chinese General Social Survey (CGSS) (n = 3,047 respondents), we use multiple linear regression model and the two-stage least-squares model to analyze the correlation between health-related expertise of intra-family and self-rated health.
Results

The presence of a health professional (HP) in the family is associated with better self-rated health (SRH), and the effect is more important in rural areas than urban areas. An increased chance of exercising appears to explain a part of the association between HP and SRH.
Discussions

Health professionals doing for their family members would have the potential to make a substantial dent in population health and reduce health inequality. Future work will need to understand the patterns of intra-family expertise in health (and other) domains, and the potential replicability of this transmission by public policies.


Introduction
There is a widespread belief that the de ciency in health information and medical knowledge is the main barrier to achieving high-quality health outcomes, have been observed in many countries (Frakes, Gruber, & Jena, 2020). Health-related expertise knowledge is key for people to change the stages of behaviors: pre-contemplation, contemplation, preparation, action, maintenance and termination (Eibich & Goldzahl, 2020;Prochaska & Velicer, 1997). The stage from pre-contemplation to contemplation is the most important in part, because an individual evolves from being uninformed or underinformed about health behavior to intend to change behavior, prove awareness and balance the pros and cons of health behaviors. For example, information disseminated by screening programs leads to an increase in various testing (Callaghan,

Measurement Health Outcome
A single item self-rated health (SRH) measure with ve response categories was used in the CGSS. Respondents were asked 'In general, would you say your health is excellent, very good, good, fair or poor?' We coded this variable as 1 = poor, 2 = fair, 3 = good, 4 = very good and 5 = excellent.

Intra-family Expertise Accessibility
We are interested in whether unequal exposure to health-related expertise affects individuals' health outcomes. We use the presence of a health professional in the family to measure intra-family expertise accessibility. It is reasonable to think that family members and friends of a health professional have greater access to such expertise in daily informal communication, which in turn may improve their own health knowledge, change health behavior and thereby improve their health outcomes. HP is a binary variable based on a question: "Do you know anyone who works as a nurse?" and there are 5 options, which are "family or relative, close friends, someone else, no one, and can't chose.(Sapin, Joye, & Wolf, 2020)" We coded this variable as 1 = family or relative or close friends, 0 = someone else or no one.

Mechanism variables
Exposure to expertise may affect individuals through multiple mechanisms. For example, they can transmit health knowledge about the costs and bene ts of healthy behaviors and health investments, improving the assessment of the marginal e ciency of health care consumption and prevention (L. Chen, Fan, & Chu, 2020; Y. Chen, Persson, & Polyakova, 2019; Eibich & Goldzahl, 2020); On the other hand, they can remind, nudge or corroborate existing knowledge, enable individuals to better internalize the new health information in their decision making process and guide them to seek formal care and improve their health.
According to the question in the survey, we test two potential mechanism variables in total: taking exercise and having relax regularly. The two variables were based on the two questions "do you take exercise regularly?" and "do you have relax regularly". We coded these two variables as 1 = never, 2 = less, 3 = average, 4 = often and 5 = very often, respectively.

Control variables
Regression analyses adjust for sociodemographic characteristics that are associated with self -rated health (Grossman, 1972 , and minzu (han vs minority). We also control for individual's educational attainment (continuous variables), income level (we took the natural log of self-reports of personal income last year) and spouse's educational attainment (continuous variables), which relates to individual's health outcomes. Additional control variables re ect formal health services accessibility that could be associated with self-rated health. These include whether or not participate in endowment insurance (yes vs no) and medical insurance (yes vs no).
Analytic Strategy Stata 15.0 was used to analyze the impact of health-related expertise accessibility on the health outcome in China. First, self-rated health was regressed on the presence of a health professional in the family, controlling for sex, age, age square, minzu, hukou, educational attainment, spouse's educational attainment, endowment insurance, and medical insurance. Second, two-stage least-squares model was used to resolve the endogenous problems of health-related expertise accessibility and health outcome.
It seems plausible that the effect of health-related expertise differs by hukou. For example, the household registration (hukou) system divided China into two separated societies, with the majority of the population con ned in the rural areas and entitled to have fewer chance to access to health resources compared with urban residents. To investigate differences in the explanatory contribution by hukou, we conducted separate heterogeneity analyses by groups.
To further investigate the potential mechanisms linking health-related expertise accessibility to respondent's self-rated health, we conduct Sobel-Goodman mediation tests (Sobel, 1982), which assess whether the inclusion of mechanisms variables to the baseline model explains a signi cant amount of the relationship between health-related expertise and health outcome in China. Footnote: [1] A third of the subsample was randomly selected from full sample over 10,000 respondents and asked for detailed social network information, leading to smaller sample size for our analysis. Table 1 presents descriptive statistics for all variables by the presence of a health professional in the family. Respondents having a health professional in the family, report a better SRH (Mean = 3.18 SD = 1.21) than their counterparts. Compared with those having no health professional in family, respondents having a health professional in the family are more likely to be young age, rural hukou. Moreover, respondents having a health professional in the family have slightly more years of education and higher income level, but have no difference by formal health services accessibility, like participating in endowment insurance and medical insurance. The Effect of the Presence of a Health Professional in the Family on SRH As is shown in Table 2, the presence of a health professional in the family has a positive and signi cant effect (β = 0.127, p = 0.008) on SRH, that is, unequal exposure to health-related expertise is an underlying causal mechanism that contributes to better self-rated health. This effect is robust to several changes of the model speci cation, e.g., controlling for additional sociodemographic characteristics (sex, age, age square, hukou, minzu, educational attainment, income level, spouse's educational attainment), health insurance (endowment insurance and medical insurance).

Descriptive Statistics
Compared with female, male (β = 0.127, p = 0.005) had a higher self-rated health. The age effect shows that self-rated health decreases rst and then increases with age, following a U-shaped curve over age. In addition, individual and spouse's educational status and self-rated health showed a positive and signi cant association: individual's higher educational level (β = 0.023, p = 0.000) and spouse's higher educational level (β = 0.018, p = 0.006) were associated with reporting better self-rated health, respectively. Further, individual with a higher income (β = 0.028, p = 0.001) had a better self-rated health outcome. In contrast, having health insurance was associated with higher elf-rated health outcome, but not signi cantly. This revealed the importance role of informal health accessibility other than social insurance and formal access to healthcare in sustaining health inequality in China.

Robustness Test
In the above benchmark model, it is likely the key explanatory variable the presence of a health professional in the family is endogenous, leading to potential bias in the estimation. Theoretically, the endogeneity may arise due to three reasons: (1) omitted variables affecting key explanatory variable and self-rated health; (2) reverse causality between the presence of a health professional in the family and self-rated health; (3) measurement error in individuals' reported health variables (Eriksson, Pan, & Qin, 2014; Y. Liu, Duan, & Xu, 2020). In our model, it is not likely that (3) will be the case, as the self-rated health does not usually suffer from substantial recalling errors. Thus, possible endogeneity would most likely arise from unobserved factors and reverse causality.
We used the instrumental variable method to resolve the endogenous problems. A valid instrumental variable must satisfy two conditions: (1) Correlation, where the instrumental variable is related to the informal health accessibility; (2) Independence, where the instrumental variable is not related to the error term (the random error that affects self-rated health). We selected "number of contacts per day" as instrumental variables. The motivations are as follows. First, there is a strong correction of social network and social resources, particularly in developing countries. In general, the more people an individual has access to, the more likely it is to have expose to health-related expertise. Second, the number of contacts per day should (arguably) not directly impact the individual's self-rated health, because it is unlikely to be correlated with the unobserved individual's heterogeneity (such as genetic formation). Table 3 shows the regression results using the instrumental variables. The regression results of the rst stage show that number of contacts per day signi cantly increase the likelihood of people espouse to health-related expertise controlling for other characteristics (β = 0.027, p = 0.000). After dealing with the endogenous problems of informal health accessibility and self-rated health, it was found that informal health accessibility still had a larger positive effect on their self-rated health (β = 2.627, p = 0.007), comparing to the previous result.  Liu et al., 1999). The persistent urban-rural inequality in health raises concerns regarding the already segregated and unbalanced development paths between the two sectors. Thus, it is important to examine the role of health-related expertise in accounting for the health inequality, justifying separate analyses on urban and rural populations.
These results indicate that exposure to health-related expertise can have differing effects for urban and rural residents, and might therefore contribute differently to the explanation of self-rated health inequalities. Speci cally, among rural hukou, exposure to health-related expertise has a positive and signi cant effect (β = 2.253) on health outcome, while the coe cient of urban hukou is not signi cant.
Taken together, Table 4 shows that the more importance of health-related expertise as a key explanatory factor in rural areas. Note: t statistics in parentheses; * p < 0.1, ** p < 0.05, *** p < 0.01

Potential Mediating Pathways
We turn to examining underlying mechanisms through which engagement in health-related expertise shapes respondent's self-rated health. Before a formal test of mediation, we examine whether healthrelated expertise is associated with proposed mechanism variables. Columns 1 and 3 of Table 5 presents the associations between health-related expertise and the two mechanism variables. Consistent with earlier studies, results show that health-related expertise is associated with the mechanism variables of healthy behavior in the expected directions. Furthermore, Columns 2 and 4 present results for the mechanisms linking health-related expertise to self-rated health, suggesting that taking exerciser regularly is important mediator connecting HP to SRH, while having relax is not a valid mechanism variable. Note: t statistics in parentheses; * p < 0.1, ** p < 0.05, *** p < 0.01

Discussion
Growing evidence across various disciplines reveals stark correlations between formal access to healthcare throughout the course of life and health outcome. Although China is becoming a universal formal access to healthcare and a well-developed social safety net, mortality and morbidity across a range of ages and conditions still persist. Hence, identifying informal access to health care that affect health outcome is of a great public health concern to improve overall population health and reduce health inequalities. This study examined the association between informal health accessibility and self-rated health status and their underlying mechanisms in China.
We nd that exposure to health-related expertise in a family is associated with better self-rated health outcomes. To examine whether the association between health-related expertise and health outcome was confounded by unobserved family background and reverse causality, we estimated models with instrument variable. The estimated result of the instrumental variable documents a robust association between health-related expertise and self-rated health, and the estimate effect of health-related expertise in a family increased. In general, these ndings build on previous studies about the roots of health inequality.
In addition, consistent with previous research on rural-urban differences, our results reveal the effect of health-related expertise differs by hukou status. Health-related expertise is even more important in rural areas than urban areas. Furthermore, we also apply mediation analysis to uncover the mechanisms connecting health-related expertise to health outcome. Our results suggest that an increased chance of exercising appears to explain a part of the association between health-related expertise and self-rated health.
Yet, the ndings should also be considered in light of several limitations. First, this study used crosssectional survey data from 2017 and cannot identify a causal relationship between health-related expertise and self-rated health outcome in China, even we use the instrument variable. Second, self-rated health is easily measured and has been shown to be a stable predictor of subsequent mortality in different populations (Bopp et al., 2012). However, the use of self-rated health can be in uenced by cultural or mood backgrounds. Moreover, it might not be very sensitive to variation in health especially in younger age groups. Third, compared to the main models, the elaborated model of mediation analyses is less rigorous and insu cient informed from a causal point of view.
Despite these limitations, the results of this study offer information that may in uence public policy. As the rst study to examine the effect of informal health accessibility in China, distinguishing the different effects of urban and rural exposure to the health-related expertise, our study adds new empirical evidence to the literature on the origin's health inequality. With the improvement of the access to formal health resources, such as medical and health service system and social endowment insurance, policy-makers and practitioners should take a more holistic and multifaceted approach to know about the important in uence of non-institutional health resources on individual health, instead of solely emphasizing the importance of institutional resource acquisition as their strategy. A public health policy that should be able to "mimic" what health professionals do for their family members would have the potential to make a substantial dent in population health and reduce health inequality. In light of the intra-family transmission of Covid-19, understanding the patterns of intra-family expertise in health (and other) domains, and the potential replicability of this transmission by public policies, remains an important area for current and future work.

Conclusion
The presence of a health professional (HP) in the family is associated with better self-rated health (SRH), and the effect is more important in rural areas than urban areas. Health professionals doing for their family members would have the potential to make a substantial dent in population health and reduce health inequality. Future work will need to understand the patterns of intra-family expertise in health (and other) domains, and the potential replicability of this transmission by public policies.