Self-reported symptoms of depression among Chinese rural-to-urban migrants and left-behind family members

IMPORTANCE There were an estimated 247 million rural-to-urban migrant workers in China in 2016, yet at a national level, there is scant evidence on the association of migration with mental health among migrants and their left-behind family members. OBJECTIVE To examine the association of rural-to-urban migration with symptoms of depression among migrants and left-behind family members aged 45 years and older. DESIGN, SETTING, AND PARTICIPANTS Using representative cross-sectional data of 14332 middle-aged and older adults from the 2015 China Health and Retirement Longitudinal Survey, regression analyses were conducted to examine the association of depressive symptoms with rural-to-urban migration status in urban areas and the association of depressive symptoms with left-behind status in rural areas. The statistical analysis was performed from January to August 2018. EXPOSURES Migration status (defined as having a rural hukou [household registration record]) in urban areas and left-behind status (defined as having a spouse or child living in another area) in rural areas. MAIN OUTCOMES AND MEASURES Depressive symptoms measured on the 10-item Center for Epidemiological Studies Depression (CES-D-10) scale. rural of rural family vs intact-family rural residents. Ethical approval for primary data collection Our involved secondary analysis of deidentified established data sets and ethical approval or informed consent Economics and Political Science 22 research ethics policy and procedures. Results were reported following the Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) reporting guideline for cross-sectional studies.


Introduction
Rural-to-urban migration in China has been a major driving force behind its economic growth in the past few decades. Beyond economic implications, there has been an increasing interest from research and policy communities to understand the association of rural-to-urban migration with various well-being outcomes in China. The 2018 World Happiness Report 1 suggests that rural-tourban migrants report lower happiness levels despite their higher incomes. The associations of migration could go beyond happiness, well-being, and life satisfaction: migration could also be associated with worse health outcomes, including mental health outcomes.
Despite increased interest, the literature on the association of rural-to-urban migration with mental health and the channels through which rural-to-urban migration and mental health are associated is limited, particularly for older adults. To our knowledge, most of the literature has focused on the migration and mental health nexus of younger migrants. 2 In addition, the literature offers limited evidence on the channels through which migration and mental health are associated.
To date, most studies in the literature focus on the hukou system, the Chinese household registration system, which is primarily based on place of birth and hence divided into 2 broad categories: urban and rural. Hukou affiliation is linked to the provisioning of social services (eg, schooling, types of jobs and employee benefits, housing opportunities). 3 This implies that rural-to-urban migrants are generally excluded from social welfare services, 4-6 with a potential impact on mental health, particularly for women and now-older migrants who moved during the earlier, more restrictive hukou environment. [7][8][9] For older rural-to-urban migrants who faced a more restrictive system, this may be associated with poorer mental health compared with their younger counterparts. A 2014 study 2 of a single city in China supports this notion, but nationally representative samples have not been explored, to our knowledge.
The large wave of rural-to-urban migration has also created a group of left-behind parents, spouses, and older adults. This issue has been explored in other Asian countries, but, to our knowledge, most of the studies in the Chinese context have been small and nonrepresentative. 10,11 The existing studies in China suggest that older left-behind or empty-nest parents have higher levels of depressive symptoms. Higher depressive symptoms may be attributed to lowered social capital, worse economic status, and increased loneliness. [12][13][14] More evidence from a nationally representative survey is needed to examine the mental health status of left-behind family members and the potential mitigation mechanisms, such as remittances or close emotional ties. 15,16 Additionally, further evidence is needed to distill policies relevant for strengthening the mental health of elderly rural-to-urban migrants. For example, more emphasis may be needed on community building and promotion of social support networks. 17 Studies have shown that increased social participation is associated with higher subjective well-being and can mediate the association of resources with well-being for retired people. [18][19][20] It is important to examine whether this is also the case for elderly migrant groups. This study had 3 objectives. The first was to examine the association of migration status with depressive symptoms in urban areas. The second was to examine the association of left-behind status with depressive symptoms in rural areas. The third was to explore factors that could contribute to or mitigate the disparities in depressive symptoms associated with rural-to-urban migration.

Migrants
We based our analysis of migrants on the urban subsample. Urban and rural area definitions were provided in the CHARLS according to the guidelines by the China National Bureau of Statistics. 25 Out of the concern that urban areas by this definition include the outskirts of city and town areas that were in fact rural in nature, 26 we excluded the outskirt areas and restricted the urban sample to the central areas of cities and towns. Migration status was defined by respondents' current place of residence and their current hukou status. Individuals were categorized as rural-to-urban migrants if they lived in an urban area and currently had rural hukou. Individuals were categorized as urban residents if they lived in an urban area and had nonrural hukou. In further analysis, we divided ruralto-urban migrants by sex.

Left-Behind Family Members
We based our analysis of left-behind family members on the rural subsample. Left-behind spouses were people living in a rural area who did not live with their spouse. Left-behind parents were people who had no children living nearby (in the same household or community) and at least 1 child who was not living in the same community. A left-behind family member was defined as being either a leftbehind spouse or left-behind parent. Other rural residents were categorized as intact-family rural residents. Left-behind spouses and left-behind parents were not necessarily mutually exclusive.

Covariates
Covariates were selected to control for potential confounding related to socioeconomic status and physical health. Socioeconomic variables included sex, age, marital status (currently married or not), education levels (4 categories: primary school and below, middle school, high school and above, and not reported), and a continuous variable of logarithm of per-capita household expenditure (coded as 0 if not reported). Physical health indicators included a binary variable of having fallen down in the last 2 years (to proxy for physical injury), a binary variable of experiencing any physical function difficulty in 10 daily tasks (eg, walking 1 kilometer), and a binary indicator of having been diagnosed as having 1 or more of 13 listed long-term conditions (eg, diabetes). A binary variable was also included to indicate if physical function difficulty information was not reported.

Additional Variables
We further considered variables about economic and social exclusion in parts of the analysis.

Descriptive Statistics
We started by presenting descriptive statistics. First, we showed regional variations in depressive symptoms among provinces by plotting the provincial CES-D-10 mean score on a map. We divided the provincial mean CES-D-10 scores into quartiles, then visualized the data by displaying the results on a thematic map. Second, we reported unconditional subgroup means in urban and rural areas separately. Specifically, for the urban subsample, we compared the mean CES-D-10 scores and the proportions of CES-D-10 scores of 10 or greater of rural-to-urban migrants with urban residents, and for the rural subsample, we compared mean CES-D-10 scores of left-behind with intact-family rural residents. We used t tests to test the differences.

Association of Migration or Left-Behind Status With Depressive Symptoms
Least squares regression analyses were conducted to assess the association of migration or leftbehind status with depressive symptoms. To allow for error terms to be associated within the household, we clustered standard errors at the household level to avoid overstating estimation precision. Survey sampling weights were applied in all regressions. Analyses were performed using Stata statistical software version 14.2 (StataCorp).
Next, we explored what factors might be associated with the potential disparities in depressive symptoms by migration status. We regressed CES-D-10 scores based on rural-to-urban migrant status, while also controlling for a range of socioeconomic and health status covariates. In recognition of regional disparities of economic and social development, we further controlled for province fixed effects. We further examined heterogeneity by replacing the rural-to-urban migrant variable with 2 variables, one indicating a male migrant and the other indicating a female migrant.
To assess potential migration-associated depressive symptoms pathways, we focused on the eco-  were left-behind family members, and the remaining 6523 participants (65.7%) were not.

JAMA Network Open | Global Health
Descriptive statistics for urban, rural, and all participants are shown in Table 1.  Table 2 shows the results of our analysis of the difference in depressive symptoms between rural-to-urban migrants and urban residents and the associated factors. Compared with urban residents, rural-to-urban migrants had higher CES-D-10 scores after adjustment for covariates (β = 0.74; 95% CI, 0.08-1.40; P = .03). Compared with intact-family rural residents, left-behind spouses had higher CES-D-10 scores after adjustment for covariates (β = 0.54; 95% CI, 0.05-1.03; P = .03). Being a woman and being in poor physical health (indicated by having recently fallen down, physical functional difficulty, or long-term disease) were associated with reporting higher depressive symptoms. Older age, being married, having higher education levels, and having higher living standards (indicated by household expenditure) were associated with reporting lower depressive symptoms. Estimated coefficients for covariates are reported in Table 1  Finally, we explored the mental health associations among the family members left behind because of spouse or adult child migration, as reported in Table 4

Discussion
Based on our analysis of data from the 2015 CHARLS survey, we report a number of important findings. Our analysis found that rural-to-urban migration was associated with higher depressive symptoms in middle and old age. This association was particularly strong among female migrants. We also found that migration was positively associated with economic deprivation (proxied by house ownership and savings) and social exclusion (proxied by social activities index), suggesting these factors were likely channels underlying the association of migration with depressive symptoms. We also found that community building (as proxied by community social facility index) could mitigate the negative association of migration with depressive symptoms, in that migrants living in communities with more social facilities had lower depressive symptoms than migrants living in communities with fewer social facilities. Additionally, we found that being a left-behind spouse was associated with higher depressive symptoms and that among left-behind parents, receiving more familial support from children financially or emotionally was associated with lower depressive symptoms than receiving less familial support.
Previous analyses have argued that female migrants have worse mental health status because of the nature of the hukou system and the marginalization of women within the system. Existing research argues that migrant women have multiple marginalized identities and thus are more likely to be exposed to social and economic inequality compared with their male counterparts. 8,9,28 Although we were not able to show causal effects, our findings support the notion that female migrants may be at higher risk of depression than their male counterparts.
We explored the roles of economic and social exclusions in understanding the association of migration with depressive symptoms. These 2 mechanisms are particularly relevant to rural-to-urban migrants in China where the hukou system acts as an institutional barrier to integration. 29 Previous studies have explored the urban economic and social exclusion of rural-to-urban migrants and the   While we did not find evidence that left-behind family members were associated with worse mental health status overall, we did observe that left-behind spouses were associated with higher depressive symptoms and left-behind parents were not. Among left-behind parents, our results suggest that familial support is important for lowering depressive symptoms. Because of the lack of a comprehensive pension system in rural China, family support is the primary source of support for elderly rural residents. 34 Within this context, intergenerational support structures, such as financial support and regular visits from children, play an important role for the mental health of the rural leftbehind elderly population.

Strengths and Limitations
There are a few strengths of our work. First, this analysis is the first analytical effort on the mental health and migration nexus of middle-aged and elderly Chinese people relying on a well-established nationally representative data set, to our knowledge. Second, in performing our analysis, we also explored some of the factors that could be associated with lower mental health outcomes among elderly migrants, such as socioeconomic exclusion. Relevant to policy making, we also explored whether the availability of community social facilities could help limit the negative association of migration with mental health. Third, we emphasized that migration was not only associated with the mental health of the migrants themselves but also with the mental health of family members that were left behind. We further explored the association of familial support with the mental health of left-behind family members.
This study also had limitations. First, although we tried to control for as many confounding covariates and regional variations as possible, there were unobservable factors that might be associated with migration and mental health that we were not able to account for. Thus, our interpretation of the results was not causal. Although we did not establish causality, we did show a pattern of association of migration with mental health status. Second, our data only included individuals aged 45 years and older, so we were unable to examine how the mental health of leftbehind children might be associated with the rural-to-urban migration of their parents.

Conclusions
Mental health policy in China should address the issue that migrants and some of their left-behind family member may be at higher risk of poor mental health. Effectively addressing this issue requires more broad policy coordination to deal with economic and social exclusions associated with ruralto-urban migration. While this will be a long-term task, our findings regarding the association of community-based social facilities with mental health provides support to the notion that communitybased social facilities could contribute to improving mental health of elderly migrants by facilitating opportunities for interaction with others and increasing feelings of self-determination. Community building projects should be continued, and access to community centers for rural-to-urban migrants should be facilitated. Additionally, familial support was associated with the mental health of leftbehind family members. Although emotional support may be irreplaceable, the lack of a formal social security system in rural areas could be a factor in why left-behind rural residents rely on their children for financial support. Establishing a social security system in rural areas could be a path to economic security and mental well-being for rural residents.