Article Text

Download PDFPDF

Life course socioeconomic position and incidence of mid–late life depression in China and England: a comparative analysis of CHARLS and ELSA
  1. Milagros Ruiz1,
  2. Yaoyue Hu2,
  3. Pekka Martikainen2,3,4,
  4. Martin Bobak1
  1. 1 Research Department of Epidemiology and Public Health, University College London, London, UK
  2. 2 Laboratory of Population Health, Max Planck Institute for Demographic Research, Rostock, Germany
  3. 3 Population Research Unit, Faculty of Social Sciences, University of Helsinki, Helsinki, Finland
  4. 4 Centre for Health Equity Studies (CHESS), Stockholm University and Karolinska Institutet, Stockholm, Sweden
  1. Correspondence to Dr Milagros Ruiz, Research Department of Epidemiology and Public Health, University College London, London WC1E 6BT, UK; m.a.ruiz{at}


Background Despite the growing prevalence of depression in the Chinese elderly, there is conflicting evidence of life course socioeconomic position (SEP) and depression onset in China, and whether this association is akin to that observed in Western societies. We compared incident risk of mid–late life depression by childhood and adulthood SEP in China and England, a country where mental health inequality is firmly established.

Methods Depression-free participants from the China Health and Retirement Longitudinal Study (N=8508) and the English Longitudinal Study of Ageing (N=6184) were studied over 4 years. Depressive symptoms were classified as incident cases using the Center for Epidemiologic Studies Depression Scale criteria. Associations between SEP (education, wealth, residence ownership and childhood/adolescent deprivation) and depression symptom onset were assessed using Cox proportional hazards models. In China, we also investigated children’s government employment status as a SEP marker.

Results Higher education and wealth predicted lower incidence of depression in both countries. The association with non-ownership of residence appeared stronger in England (HR 1.61, 95% CI 1.41 to 1.86) than in China (HR 1.11, 95% CI 0.95 to 1.29), while that with childhood/adolescent deprivation was stronger in China (HR 1.43, 95% CI 1.29 – 1.60) than in England (HR 1.33, 95% CI 0.92 to 1.92). Chinese adults whose children were employed in high-status government jobs, had lower rates of depression onset.

Conclusions Consistent findings from China and England demonstrate that SEP is a pervasive determinant of mid–late life depression in very diverse social contexts. Together with conventional measures of SEP, the SEP of children also affects the mental health of older Chinese.

  • ageing
  • china
  • depression
  • depressive symptoms
  • england
  • incidence
  • inequality
  • socioeconomic position

Statistics from

Request Permissions

If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.


Depression is the fourth leading cause of disability in China,1 where low mental health literacy coincides with grave stigma towards mental illness.2 Alarmingly high suicide rates prompted public discourse on mental health in China,2 especially for older Chinese, as adults aged 65+ formed 8.9% of the Chinese population yet accounted for 41.2% of all suicides.3 Since mid–late life depression prevalence rose steadily from 1987 to 2012,4 addressing its social determinants is particularly timely as elderly adults will reach almost one-fourth of the total Chinese population by 2050.3 There are two critical evidence gaps on the nature of mid–late life depression inequalities for the world’s largest ageing population. It remains unclear (1) how consistent socioeconomic position (SEP) differences in depression among older Chinese are and (2) whether these differentials mirror well-established inequalities observed in Western countries.

Previous studies report conflicting evidence of SEP and depression among older Chinese. Less education predicted a greater risk of major depressive disorder (MDD) among older Beijing residents5 and more depressive symptoms among older Chinese according to a meta-analysis,6 yet other studies have found weak and inconsistent associations with education in elderly samples from Hong Kong,7 Shanghai8 and selected rural areas.9 Mid–late life depression risk by material factors is also uncertain, as inverse associations between poorer circumstances and MDD5 10 were replicated in one of two studies investigating depressive symptomatology.8 11 Older Chinese in lower occupational statuses exhibited heightened depressive symptoms, but findings were statistically insignificant.7 8 Similar to these region-specific findings, nationally representative studies also provide mixed results on mid–late life depression inequalities by education and income.12 13 Most adulthood SEP studies were cross-sectional5 7–12 and, thus, cannot dismiss various selection biases. Longitudinal studies using nationally representative data do not exist, although depressive symptom trajectories increased at a higher rate for older Taiwanese with lower educational and occupational statuses.14 There are limited studies on childhood SEP in China, although nationally representative evidence suggests that early life exposure to the Great Chinese Famine predicted raised depressive symptoms in mid–late adulthood.15

While there is ample literature16 on SEP and depression in Europe and North America, few studies have discerned whether inequalities are consistent across more diverse contexts17 18 like China. It remains unknown whether effects of life course SEP on mid–late life depression in China, a postmarket transition society, are akin to those widely reported in high-income Western countries. Older Chinese have experienced striking socioeconomic developments. Many faced impoverished childhoods, came of age amidst rampant social turmoil and acquired much improved living standards as adults due to economic reforms of the 1980s and 1990s. Such life histories are more uncommon in Western ageing populations.19 The socioeconomic and political contexts may modify how socioeconomic factors influence mental health inequalities.18 This idea is supported by a cross-sectional WHO study which found that SEP inequalities in mental disorders were evident in North American and European countries, but not in the emerging economies.20

Although depression is estimated as the fourth leading cause of disability in both settings,1 depressive symptoms in later life vary between China and England. Whereas symptoms decline during early middle age, plateau at midlife and increase at older ages in English adults,21 this U-shaped age trajectory is not evident in Chinese adults, nor do symptoms increase linearly with age in later life.22 Although the burden of mid–late life depression appears higher in England, the consequences paradoxically appear graver in China. While suicide rates have declined among older adults in both countries, these declines have been more favourable in England23 than in China.3 These national differences reinforce the need for a cross-country investigation on depression inequalities. Therefore, we measured incident risk of mid–late life depression by childhood and adulthood SEP in China and England using two nationally representative ageing studies. To our knowledge, this is the first longitudinal study comparing these relationships in China with a Western country. Given the prominence of familial bonds and filial piety in China,2 24 we also assessed whether offspring SEP (adult children’s occupational status) influenced parental depression.


Study design

The China Health and Retirement Longitudinal Study (CHARLS) and the English Longitudinal Study of Ageing (ELSA) are nationally representative studies designed to monitor the long-term health of community-dwelling adults aged ≥45 (CHARLS) and ≥50 (ELSA) and their spouses. Baseline examinations conducted in 2011/2012 and 2002/2003 yielded response rates of 81% in CHARLS and 70% in ELSA.19 25 Original cohort participants (excluding spouses) comprised 17 140 and 11 391 Chinese and English adults, respectively, who were eligible for re-examination every 2 years.19 25 We employed 4-year prospective data from the first three biennial examinations (referred to as waves 1, 2 and 4 in CHARLS and waves 1, 2 and 3 in ELSA), plus retrospective data from the life history interview (referred to as wave 3—Life History Module in CHARLS and ELSA), and included participants with repeat assessments on depressive symptoms, and available SEP and covariate data.

Clinically significant depressive symptoms

Depressive symptoms were measured using the Center for Epidemiologic Studies Depression (CES-D) Scale, which detects depression risk in general populations, including older adults.26 27 The CES-D was administered in the three CHARLS and ELSA biennial examinations, using the 10-item and 8-item versions, respectively. The CES-D 10 asked ‘how often’ 10 depressive symptoms were experienced during the past week on a 4-point scale ranging from ‘less than 1 day’ (0) to ‘5–7 days’ (3); and summed to derive scores ranging 0–30 in CHARLS at waves 1, 2 and 4. The CES-D 8 ascertained whether eight depressive symptoms were experienced ‘for much of the time’ during the past week using yes (1) and no (0) response options, and added to generate scores ranging 0–8 in ELSA at waves 1, 2 and 3. CES-D scores were not calculated for participants with an item non-response of two or more, as scoring procedures recommend that CES-D data be at least 80% complete. If participants missed one CES-D item, the missing value was imputed with the participant’s mean response reported across the other items at each wave. Participants with CES-D 10 scores of ≥12 in CHARLS and CES-D 8 scores of ≥3 in ELSA were classified as having clinically significant symptoms, as these version-specific thresholds identify probable cases of depression.26 27

Socioeconomic position over the life course

Adulthood SEP data, collected at wave 1, were harmonised across both studies. Educational level and physical wealth (household assets) data were grouped to denote study-specific hierarchies ranging from low, medium and high. Housing tenure data were dichotomised to denote whether or not the participants (or other household members) owned their current residence. Full harmonisation details of the SEP markers are reported in the supplement.

Participants were asked if their family ‘ever lacked enough food to eat’ in CHARLS or if they ‘ever experienced severe financial hardship’ in ELSA, up to age 17 in each study (wave 3—Life History Module). Participants with affirmative responses were classified as having experienced severe childhood/adolescent deprivation.

The occupational titles of participants’ children were collected in CHARLS at wave 1. Government employment (cadre membership) marks occupational prestige in China,28 where cadres form an advantaged stratum of the Chinese population, distinguished by their political power and authority.28 29 Parents of adult children who worked at the administrative level for any government agency were classified as having a cadre child and were ranked according to whether the child held a low (‘Ke’ level and below) or high (‘Chu’ level and above) administrative position.28 Occupational data of children were not collected in ELSA; therefore, no measure of offspring SEP was available for English participants.


Age, gender, geographical indicator (CHARLS: urban vs rural area, ELSA: government office region), marital status and smoking status were obtained at wave 1, plus alcohol drinking frequency, self-rated health and number of limitations of activities of daily living (ADLs), for which data were harmonised as described in the supplement.

Analytical approach

We derived three study-specific samples for the analysis of adulthood (CHARLS: 8508, ELSA: 6184), childhood (CHARLS: 7637, ELSA: 3305) and offspring (CHARLS: 6831) SEP. Briefly, we restricted our analysis to non-cases by excluding participants with clinically significant depressive symptoms, plus those who reported any previous doctor diagnosis of emotional, nervous or psychiatric conditions at wave 1. The full selection criteria, including flow diagrams (online supplementary figures S1–S3), are detailed in the supplement.

Supplemental material

Figure 1

Study-specific incidence rates of depressive symptom onset per 1000 person-months by adulthood socioeconomic position in the China Health and Retirement Longitudinal Study (top pane) and the English Longitudinal Study of Ageing (bottom pane) analytic samples.

Incidence rates of clinically significant depressive symptom onset were calculated as the number of first events divided by follow-up time in person-months, specified as the number of months from the wave 1 interview date to the follow-up interview date where the first occurrence or censoring (the date of the participant’s last interview) took place. As events could occur once or twice over the 4-year period, the first occurrence was used to establish the time to onset.

Incidence rates were estimated by each SEP marker and compared across strata using log-rank tests. Associations between SEP and incident depressive symptom onset were examined using Cox proportional hazards models. For each SEP indicator, three study-specific models controlling for the following covariates were estimated: model 1 (age group, gender and geographical region), model 2 (model 1 covariates plus marital status, smoking status, alcohol drinking frequency, self-rated health and number of ADL limitations) and model 3 (model 2 covariates plus all measures of adulthood SEP). There was no evidence of interactions between SEP and gender or age, except for between physical wealth and age in ELSA (Wald test for interaction, p value=0.0293). Therefore, all models were pooled by gender and age group, and an additional age-stratified model for physical wealth was fitted in ELSA.

We conducted a 4-year comparative prospective study to match the number of waves completed in CHARLS. As five additional CES-D assessments in ELSA (waves 4–8) were excluded, we examined whether associations were robust over a longer follow-up with a sensitivity analysis that replicated the ELSA analyses over the 14-year period.


Incidence rates of depressive symptom onset were 5.35 per 1000 person-months in Chinese adults and were slightly lower at 4.46 in English adults (table 1). Chinese and English adults were 58 and 64 years old, respectively, at baseline and were equally distributed by gender. Adults in China had less education than those in England. Physical wealth was more polarised in England than in China, as 17.6% of English adults were in the highest wealth category, compared with 30.2% of Chinese adults. Living in a residence not owned by the household was more frequent in England. Severe childhood/adolescent deprivation was pervasive in China (69.7%) but uncommon in England (3.1%). Chinese adults with a cadre child were exceptionally rare at roughly 6%, of which only 1% had a high-ranking cadre child.

Table 1

Study characteristics of the CHARLS and ELSA analytical samples

Figure 1 shows strong gradients in incidence rates of depressive symptom onset from high to low education and physical wealth categories in both countries. Incidence rate differences by whether adults owned their current residence were sizeable in England (log-rank test, p<0.0001), but negligible in China (log-rank test, p=0.1726).

Incidence rates by childhood/adolescent and offspring SEP are depicted in the supplement (online supplementary figure S4). Adults who experienced severe childhood/adolescent deprivation had higher onset rates, but this appeared stronger in China (log-rank test, p=0.0005). The borderline difference in England (log-rank test, p=0.0593) may reflect small numbers and limited statistical power. A stepwise rate reduction was observed from Chinese adults with no cadre children to parents of low-ranking to high-ranking cadre children (log-rank-test for trend, p<0.0001).

Table 2 reports associations between adulthood SEP and rates of depressive symptom onset. Adults with higher levels of education and physical wealth had lower HRs compared with low groups for each indicator, which were remarkably similar between countries. Lower rates by higher education and physical wealth were slightly attenuated but remained robust after accounting for covariates and other adulthood SEP markers. In England, the beneficial effects of physical wealth against depressive symptom onset became stronger and larger with increasing age (online supplementary table S1). The onset rate for adults who did not own their current residence was smaller and weaker in China at 11% compared with 61% in England (model 1), which was reduced to 20% after adjusting for confounders and other adulthood SEP markers.

Table 2

Study-specific HRs of depressive symptom onset by adulthood SEP

Table 3 presents the results for the childhood/adolescent and offspring SEP analyses. Severe childhood/adolescent deprivation predicted higher rates in later life, but the evidence was stronger in China (HRmodel 1=1.43, 95% CI 1.29 to 1.60) than in England (HRmodel 1=1.33, 95% CI 0.92 to 1.92). In China, this long-term effect persisted after accounting for adulthood conditions (HRmodel 3=1.36, 95% CI 1.22 to 1.52). Compared with adults with no cadre children, parents of low-ranking and high-ranking cadre members, respectively, had lower rates of 23% (HRmodel 1=0.77, 95% CI 0.59% to 0.99%) and 62% (HRmodel 1=0.38, 95% CI 0.18% to 0.83%), which were marginally explained by their adulthood circumstances.

Table 3

Study-specific HRs of depressive symptom onset by childhood/adolescent and offspring SEP

The sensitivity analysis in ELSA found that associations between adulthood and childhood SEP and incident depressive symptom onset over 14 years (online supplementary table S2) were virtually identical to those reported over the 4-year follow-up in England.


Depression inequalities by education and wealth were remarkably similar between China and England. Inequalities by non-ownership of current residence appeared stronger in England, but home ownership may not aptly measure social stratification in China, given the large rural-to-urban migrant population, which often own properties in their rural hometowns. Another explanation is that many older Chinese live with their children, but additional analyses found that associations with residence ownership did not differ by living arrangements. Inequalities by severe childhood/adolescent deprivation were also considerable but perhaps greater in the postmarket transition setting of China. Large-scale agricultural disruptions in China during World War II, the civil war (pre-1949) and the Great Chinese Famine (1959–1961) led to widespread food insecurity and starvation.15 Born between 1900 and 1966, CHARLS participants who escaped severe deprivation amidst these challenging historical periods likely came from very privileged families. Similarly, weaker deprivation inequalities in England may be explained by limited study power as only 3% of ELSA participants reported severe financial hardship before adulthood.

Given the uncertainty in whether systematic, socially produced differences in mid–late life depression existed in China,5 7–12 we provided the first report on incident risk of mid–late life depression using various life course SEP markers simultaneously. Despite some variations in our findings, depression onset inequalities were remarkably congruent between China and England. Sociologists have described how economic development may encourage intergenerational social mobility by weakening the system of ascriptive role allocation, reducing cultural barriers to mobility and spurring internal migration to more developed regions.30 This is consistent with the observation that intergenerational social mobility is low in England31 but high in China, where industrialisation occurred more recently.32 Industrialisation is theorised to alter the social structure, but not the interactions, between strata.30 This may explain why inequalities were consistent in both countries. Our findings support the ‘fundamental causes’ theory, which postulates that SEP provides individuals with innumerable resources that are imperative for health irrespective of the social context.18

Despite limited evidence on mid–late life depression in China, studies on other health inequalities have recently emerged.28 33 Family, arguably the most fundamental institution in Chinese society, is central to one’s own personal identity and provides individuals with innumerable resources, irreplaceable social networks and social support. This is especially important for elders, given strong social emphasis on filial piety.2 24 Chinese culture has stimulated studies on the effects of adult children’s social characteristics on the health of elderly parents.28 33 We adopted this family-oriented perspective and explored the upward ‘intergenerational transfer of disadvantage’16 from adult children to parents’ risk of mid–late life depression in China. Our study found that depression inequalities by adult children’s cadre status were not only strong but also equivalent to those exerted by their own adulthood SEP. Socioeconomic and political advantages of cadres may ‘spill over’ to their nuclear and extended family, and may function as a channel of transferring health-related resources from children to parents. The upward intergenerational transmission of health inequality34 remains understudied by the predominantly Western literature on health inequalities that focuses on intergenerational transmission from parents to children.16 Child-to-parent transfers of disadvantage have been investigated in a small but emerging literature,28 33 35 and we encourage future work to promote knowledge and theoretical development on late-life health inequalities.

CHARLS and ELSA, designed in accordance with the US Health and Retirement Study, ensured high-quality harmonised measurements for this comparative analysis. Response rates were somewhat lower in ELSA, but each study is as representative as feasibly possible for each context. Incident depression was tracked over 4 years, but replicating the ELSA analyses using outcome data over 14 years (unavailable in CHARLS) showed that findings were consistent over the longer follow-up. This provides some support that SEP effects on rates of depressive symptom onset may be similar over the longer life course in China.

One important criterion for depression diagnosis is impaired role functioning,20 but it is unknown whether CES-D cases, based on symptoms alone,26 27 were hindered in normal daily activities. Nonetheless, screening scales are the most practical population-based tools to detect depression,26 27 and the CES-D adequately predicts psychiatric treatment, antidepressant use27 and suicidality in community samples.36 Many clinicians in China reluctantly diagnose depression for fear of stigmatising individuals and damaging doctor–patient relationships. Medical anthropologists suggest that Chinese people are inclined to articulate emotional pain through physical symptoms ranging from headaches, insomnia, chest pain and dizziness.24 Underdiagnosed depression is also more likely in China than in England owing to differences in universal healthcare coverage. While England achieved universal health coverage in 1948,37 China is working to achieve this by 2020.38 Therefore, comparing clinically diagnosed cases between the two countries could have biased the results. Despite these cultural and structural differences, onset rates of self-reported depressive symptoms were broadly equivalent between CHARLS and ELSA, and form a solid basis for comparative research on the social patterning of depression.

Although mid–late life depression inequalities exist in both countries, China faces additional challenges, including scarce public mental health resources and open discrimination against individuals with depressive disorders.24 Following the 2013 Mental Health Law of the People’s Republic of China, equity-oriented practice must now address socioeconomic vulnerabilities of older adults.

What is already known on this subject

  • There is limited and conflicting evidence of the relationship between socioeconomic position (SEP) and depression for China’s ageing population. Further doubt on this relationship in China is cast by early international comparisons, which suggested that inequalities in common mental disorders (including depression) appear less salient in emerging economies than in the West.

What this study adds

  • This first report on life course SEP and incident risk of depression in China shows that SEP from childhood/adolescence to adulthood has a pervasive impact on depression risk among older Chinese. By showing that mid–late life depression inequalities in China are similar to those in England, this study counters the notion that depression inequalities are chiefly a Western phenomenon. Additional analyses in China have found that adult children’s government employment status influences parental risk of depression at mid–late life and provides novel evidence on the upward intergenerational transmission of health inequality.


We thank the Gateway to Global Ageing Data and the UK Data Service for making the China Health and Retirement Longitudinal Study and English Longitudinal Study of Ageing study data sets, respectively, freely available to researchers.



  • MR and YH are joint first authors.

  • Contributors MR and YH, as joint first authors, developed the harmonised variables for the article, did the scientific literature search, prepared the data for analysis, analysed the data, interpreted the results, and prepared the first and final drafts of the article. PM and MB made substantial contributions to the analysis and interpretation of data, and critically reviewed the first and final drafts of the article for important intellectual content. MB made substantial contributions to the conception and design of the study.

  • Funding MR, PM and MB are supported by a European Commission Horizon 2020 grant, 667661, as part of the Promoting Mental Wellbeing in the Ageing Population: Determinants, Policies and Interventions in European Cities (MINDMAP) research project. PM is also funded by the Academy of Finland (#1294861 and 1308247), the Strategic Research Council PROMEQ project (#303615) and the C-LIFE project funded by NordForsk (#75970).

  • Competing interests None declared.

  • Patient consent for publication Not required.

  • Ethics approval Ethical approval was provided by the Ethical Review Committee of Peking University (IRB00001052-11015) and the London Multi-Centre Research Ethics Committee (MREC/01/2/91), respectively for the CHARLS and ELSA studies. All CHARLS and ELSA participants provided written informed consent.

  • Provenance and peer review Not commissioned; externally peer reviewed.

  • Data availability statement Data are available in a public, open access repository.