This study used data from the China Health and Retirement Longitudinal Study (CHARLS), which is a nationally representative survey in China. The objectives of this survey were to provide information about demographic characteristics, health status and functioning, health care and insurance, and socioeconomic conditions. Face-to-face computer-assisted personal interview (CAPI) was conducted every two years. Samples were obtained by the probability-proportional-to-size (PPS) sampling technique to ensure the representativeness of the sample. In the first stage, all counties (except Tibet) were stratified by region, urbanity and GDP per capita. Primary sampling units (PSUs) were chosen among each selected county using administrative villages in rural areas or neighborhoods in urban areas, which comprised resident committees. In each PSU, samples of dwellings were randomly selected using mapping software named CHARLS-GIS. In total, the survey was conducted in 28 provinces, 150 countries or districts, 450 villages or urban communities, consisting of people aged 45 and over living in households, but the baseline respondents who later entered into an institution were followed. The national baseline survey was conducted from May 2011 to March 2012. The total sample households included 23,422 dwellings, and the survey finally managed to contact 17,708 individuals in 10,257 households with an overall response rate of 80.5% (26). Further details of the sample are available elsewhere (26), and data can be accessed through its official website (http://charls.pku.edu.cn/).
According to a prior study (16), we chose the CHARLS baseline survey since it contained more sufficient information on individual migration experience, while migration indicators in surveys of younger cohorts were deduced based on the 2011 survey. From the 2011 wave, we excluded 2748 participants without information on migration and 1731 participants without information on depressive symptoms. A total of 1773 participants who were migrants were excluded, and 310 cases of missing covariates or mediators were also excluded. Our final sample contained 11,156 respondents with an average age of 58.91. Figure 1 presents a flowchart of the 2011 CHARLS study.
[Insert Fig. 1 here]
Depressive symptoms. Our outcome variable was a binary measure (i.e., whether depressive symptoms were present). This study used the 10-item Center for Epidemiologic Studies Depression Scale (CESD-10) short form (27), which has satisfactory validity and reliability among the Chinese older population (28, 29). The respondents were asked to rate their positive feelings, negative emotions and somatic symptoms experienced over the past week through the 10 following questions on a 4-point scale from rarely or none of the time (less than 1 day) to most or all of the time (5–7 days): (1) I was bothered by things that do not usually bother me; (2) I had trouble keeping my mind on what I was doing; (3) I felt depressed; (4) I felt everything I did was an effort; (5) I felt hopeful about the future; (6) I felt fearful; (7) My sleep was restless; (8) I was happy; (9) I felt lonely; (10) I could not get "going". The depressive symptoms index was obtained from the sum of the scores of the 10 questions ranging from 0 to 30, and a higher score indicated higher depression. According to previous studies, a cutoff point of 10 was of good validity among older Chinese respondents (27, 28); consequently, respondents who scored at least 10 in this study were considered to have depressive symptoms.
Residential status. Our dependent variable was categorized as urbanized rural residents, rural non-migrants and urban non-migrants. We used both the migration information and official household registration system information (called Hukou in Chinese) to measure residential status. In detail, we first removed migrants from the sample and took non-migrants as the targeted study population to study the exogenous effect of urbanization on depressive symptoms. To do this, we used data on their birthplace, current place of residence, age at migration and duration of migration to measure their experience of migration, excluding return migrants (having a migration experience of more than six months outside their birthplaces) and early-life migrants (the age at migration is younger than 16). We then used Hukou status to classify the non-migrant population into three groups: (1) urbanized rural residents, referring to local urbanized residents who have realized urbanization in their own towns and villages and hold rural Hukou. Its early life lives in the countryside, while its later life lives in the city. (2) rural non-migrants, referring to people who live in rural areas and hold rural Hukou; (3) urban non-migrants, referring to people who are born and live in urban areas and hold urban Hukou.
According to previous studies (13, 14, 20, 30–32), we considered social participation, healthcare utilization, income per capita and living conditions as mediators.
Social participation. In this study, social participation was a continuous variable with an aggregated count (0–11). Respondents were asked about whether they had taken part in any of the following types of social activities in the past month: (1) interacted with friends; (2) played mahjong, chess, cards or went to community clubs; (3) provided assistance to family members, friends or neighbors who do not live together for free; (4) went to a sport, social or other clubs; (5) took part in a community-related organization; (6) engaged in voluntary or charity work; (7) took care of a sick or disabled adult who does not live with the respondent for free; (8) attended an educational or training course; (9) invested in stock; (10) used the internet; (11) other, or (12) none of these. We then summed the total number of social activities one participated in from the above multiple-choice question, and the index ranged from 0 to 11, with higher scores indicating greater social participation.
Healthcare utilization. This study generated a binary healthcare utilization variable (1 = yes; 0 = no). In general, healthcare utilization includes outpatient and inpatient care. In this study, we used self-reported information on whether respondents had “visited a public hospital, private hospital, public health center, clinic, or health worker’s or doctor’s practice, or been visited by a health worker or doctor for outpatient care in the last month” or “received inpatient care in the past year”. Based on the characteristics of the data, this study generated a healthcare utilization variable conditioned upon the occurrence of outpatient visits or inpatient visits.
Income per capita. Income per capita is a continuous variable. We used the self-reported household income per capita in this study to measure the economic conditions of individuals.
Living conditions. Living conditions is a continuous variable measured as an aggregated count (0–7) of seven dichotomous indicators: (1) concrete and steel/bricks and wood, (2) flushable toilet, (3) running water, (4) shower or bath facilities, (5) coal or natural gas supply, (6) telephone connection, and (7) broadband internet connection. The summed score of these seven items ranged from 0 to 7, with higher scores indicating better living conditions.
Covariates. We included age (continuous variable), sex (female/male), educational attainment (primary school and below/junior high school and above), living arrangement (living without spouse/living with spouse), chronic diseases (1 = yes, 0 = no), smoking (1 = yes, 0 = no) and drinking (1 = yes, 0 = no).
In this study, descriptive statistics were used to present the characteristics of participants and the prevalence of depressive symptoms in the three groups. Logistic regression models were used to examine the relationship between urbanization and depressive symptoms in Models 1, 2, 3, 4, and 5. In Model 1, we controlled all covariates (age, sex, educational attainment, living arrangement, chronic diseases, smoking and drinking) to examine the joint results. In addition to the factors of Model 1, four mediators (social participation, healthcare utilization, income per capita, and living conditions) were included in Models 2 to 5 in turn. The analyses were performed with Stata/SE 15.0 for Windows (Stata Corp, College Station, TX, USA). In addition, to explore how some factors might mediate the association between urbanization and depressive symptoms, we applied the Structural Equation Modelling (SEM) approach to further investigate the causal pathways. SEM analyses were conducted using MPlus version 8.3 (Muthén & Muthén, Los Angeles, CA, USA). A CFI > 0.90 was considered an adequate model fit (33, 34), and a p value of less than 0.05 was considered statistically significant.