Intergenerational contact and depressive symptoms in midlife and older adults: an analysis of six cohort studies across 33 countries

Summary

Background

Population aging has heightened global concern about mental health in later life. Intergenerational contact, a crucial component of social support, is a modifiable social factor. However, existing evidence is largely derived from single-country studies and there is limited cross-national longitudinal research exploring how these relationships vary across diverse sociocultural and welfare contexts. This study examined the association between intergenerational contact and depressive symptoms among adults aged 50 years and older across 33 countries to address these gaps.

Methods

We conducted a cross-national analysis using harmonized data from six population-based cohort studies to ensure cross-national comparability: the Health and Retirement Study (HRS), English Longitudinal Study of Ageing (ELSA), Survey of Health, Ageing and Retirement in Europe (SHARE), China Health and Retirement Longitudinal Study (CHARLS), Korean Longitudinal Study of Aging (KLoSA), and Mexican Health and Ageing Study (MHAS), collected between 2010 and 2019. Intergenerational contact with children was categorized into three categories: co-residence; living apart with frequent (at least weekly) contact; and living apart with infrequent contact. Depressive symptoms were measured by the Centre for Epidemiologic Studies of Depression scale (CES-D) or the Euro-Depression scale (Euro-D), as appropriate. Logistic regression models with generalized estimating equations were used to estimate associations, adjusting for demographic characteristics, socioeconomic status, health status, and health related behaviors.

Findings

The analytic sample included 189,036 participants. The prevalence of depressive symptoms, defined using scale-specific cut-offs, ranged from 20·3% in England to 33·3% in China. Compared with co-residence, living apart with infrequent contact was consistently associated with higher odds of depressive symptoms in most cohorts, including China (odds ratio [OR] 1·19, 95% confidence interval [CI] 1·11–1·27), the US (OR 1·18, 95% CI 1·10–1·28), continental Europe (OR 1·10, 95% CI 1·05–1·16), South Korea (OR 1·24, 95% CI 1·14–1·35), and Mexico (OR 1·43, 95% CI 1·23–1·66). By contrast, living apart with frequent contact was associated with lower odds of depressive symptoms in European cohorts, including England (OR 0·88, 95% CI 0·81–0·96) and continental Europe (OR 0·93, 95% CI 0·91–0·96).

Interpretation

The association between intergenerational contact and depressive symptoms among midlife and older adults is context dependent. Residential separation combined with infrequent contact may reflect reduced emotional support and is linked to an increased burden of depressive symptoms across various settings. Conversely, frequent intergenerational contact can help mitigate this risk in certain cultural contexts. Future research is needed to clarify causal pathways across diverse settings.

Funding

This study is funded by the National Natural Science Foundation of China.

Research in context.

Evidence before this study

We searched PubMed and Web of Science from database inception to Dec 27, 2025 independently, using combinations of the terms (“intergenerational contact” OR “intergenerational support” OR “living arrangement” OR “co-residence”) AND (“depression” OR “depressive symptoms”) AND (“older adults” OR “ageing”). Previous studies have examined associations between intergenerational co-residence, contact with children, and depression among midlife and older adults, but findings remain inconsistent. Most studies have examined living arrangements and contact frequency separately rather than conceptualizing intergenerational contact as a multidimensional construct combining co-residence status and interaction frequency. In addition, existing evidence is largely derived from single-country studies, limiting the ability to evaluate whether associations vary across sociocultural contexts. To our knowledge, no prior longitudinal cross-national study has examined intergenerational contact patterns and depressive symptoms using harmonized population-based cohort data across multiple regions.

Added value of this study

The present study integrates harmonized longitudinal data from six large ageing cohorts—the Health and Retirement Study (HRS), English Longitudinal Study of Ageing (ELSA), Survey of Health, Ageing and Retirement in Europe (SHARE), China Health and Retirement Longitudinal Study (CHARLS), Korean Longitudinal Study of Aging (KLoSA), and Mexican Health and Ageing Study (MHAS)—representing 33 countries and 189,036 participants aged 50 years and older. By jointly characterizing intergenerational contact through living arrangements and contact frequency, we provide cross-national evidence on how intergenerational contact relates to depressive symptoms in midlife and later life. Living apart from children without frequent contact was consistently associated with higher odds of depressive symptoms across most cohorts, whereas living apart with frequent contact was associated with lower odds of depressive symptoms in several European settings. These findings underscore the importance of intergenerational contact for health and aging, while also highlighting contextual differences in how intergenerational relationships relate to mental health outcomes.

Implications of all the available evidence

The association between intergenerational contact and depressive symptoms among midlife and older adults appears to vary across family, cultural, and welfare contexts. Residential separation is not consistently associated with higher depression risk when frequent contact and social support systems are available to support independence and autonomy. Given the observational nature of the evidence and the possibility of reverse causation, causal inferences cannot be drawn. However, these findings highlight the potential importance of maintaining intergenerational connections and strengthening community and welfare resources that complement family-based support. Future research is needed to determine whether such approaches can improve mental health outcomes in ageing populations.

Introduction

Global population aging is accelerating at an unprecedented pace. The global population aged 65 years or older is projected to reach 1·5 billion by 2050.¹ However, gains in longevity have not been matched by comparable improvement in healthspan.² Mental health conditions, particularly depressive symptoms, constitute a major and growing component of the disease burden in later life. By 2050, the number of older adults worldwide living with major depressive disorder is projected to reach 97·04 million,³ with substantial consequences for quality of life, functional independence, and health-care systems. Ensuring healthy ageing has therefore become a central global public health priority.⁴

Across societies, family members remain the cornerstone of care and social support for midlife and older adults.⁵ Yet family structures and living arrangements are undergoing profound transformation. Although increased life expectancy has raised the likelihood of three or more generations being alive simultaneously, older adults are now far more likely to live apart from their children than in previous decades.⁶ This transition toward ‘empty nest’ households has increased markedly; for instance, 44% of households in the US aged 65 and above are empty nesters, and 25·7% of Canadian households are couples living without children. More recently, this transition has accelerated across Asian societies, with countries such as China, Japan, and India reporting sharp increases in empty nesters.⁷ These demographic changes may diminish opportunities for intergenerational contact that were more prevalent in traditional co-residential living arrangements, potentially heightening the risk of depression for parents with children.

Intergenerational contact—typically defined as interactions between midlife or older adults and their children—encompasses both living arrangement (such as co-residence) and the frequency of contact among those living apart. Previous studies have linked intergenerational contact to psychological well-being and reduced mortality risk,⁸ but inconsistent findings have also been reported.^9,10 Some studies have found that co-residence is associated with higher levels of depressive symptoms,11, 12, 13 while others have indicated it can have protective effects, particularly in contexts where multigenerational living reflects cultural norms rather than economic necessity.^14,15 Similarly, although more frequent intergenerational contact is generally associated with reduced loneliness and improved well-being,¹⁶ these benefits may diminish with increasing geographical distance or weaker emotional closeness.¹⁷ Cross-cultural evidence further suggests that both the patterns and implications of intergenerational contact vary substantially between individualistic societies, where welfare-oriented support for older adults is anticipated, and societies with strong familism traditions, where family caregiving is the primary expectation for caring for older adults.¹⁸ Given that women often assume primary caregiving and emotional labor roles, theoretical frameworks suggest they may be more psychologically reactive to intergenerational contact than men, yet empirical findings are inconsistent—some studies report greater benefits for women, while others find no gender differences or even that men are more responsive.¹⁹

Despite this growing body of research, important gaps remain. Most studies have examined co-residence and contact frequency as separate exposures rather than considering them jointly as dimensions of intergenerational contact. Moreover, few studies have systematically examined whether the mental health implications of intergenerational contact differ across socio-cultural contexts using comparable longitudinal data. Clarifying these associations is essential for understanding how evolving family structures are associated with mental health outcomes over time in aging populations.

To address these gaps, we conducted a longitudinal, cross-national analysis using harmonized data from six large cohort studies of aging, spanning 33 countries with diverse socio-cultural contexts. We examined the association between intergenerational contact—jointly characterized by living arrangements and contact frequency—and depressive symptoms among adults aged 50 years and older, and assess heterogeneity across countries and population subgroups.

Methods

Study design and data sources

This study drew on data from six large, population-based aging cohorts comprising adults aged 50 years and older from 33 countries: the China Health and Retirement Longitudinal Study (CHARLS) from China,²⁰ the Health and Retirement Study (HRS) from the US,²¹ the English Longitudinal Study of Ageing (ELSA) from England,²² the Survey of Health, Ageing and Retirement in Europe (SHARE) from continental Europe,²³ the Korea Longitudinal Study of Aging (KLoSA) from South Korea,²⁴ and the Mexican Health and Ageing Study (MHAS) from Mexico.²⁵ Data collected between 2010 and 2019 were included. Specifically, we used Waves 1–4 (2011–2018) from CHARLS, Waves 10–14 (2010–2018) from HRS, Waves 5–9 (2010–2018) from ELSA, Waves 4–8 (2011–2019) from SHARE, Waves 3–7 (2010–2018) from KLoSA, and Waves 3–5 (2012–2018) from MHAS. The cross-sectional response rates across the six cohorts during 2010–2019 ranged from 18·1% in SHARE to 83·8% in CHARLS.26, 27, 28, 29, 30, 31 Detailed information on sampling, field procedures, and response rates for each cohort has been reported in the respective study's cohort profile papers as summarized by the Gateway to Global Aging Data platform.³² All participants were retained in the analysis regardless of missing follow-up data.

Participants were excluded if they were younger than 50 years, reported having no children, or had missing data on depressive symptoms. After applying these exclusion criteria, the final analytical sample included 20,541 participants with 54,301 observations from CHARLS; 23,965 participants with 84,004 observations from HRS; 11,047 participants with 37,187 observations from ELSA; 105,966 participants with 232,006 observations from SHARE; 8972 participants with 36,480 observations from KLoSA; and 18,545 participants with 39,707 observations from MHAS. In total, the pooled sample consisted of 189,036 participants with 483,685 observations across all cohorts (Fig. 1).

Fig. 1.

Flow diagram of participant selection across six longitudinal aging cohorts. The panels show the stepwise exclusion of participants from each cohort: (a) CHARLS, (b) HRS, (c) ELSA, (d) SHARE, (e) KLoSA, and (f) MHAS. After excluding participants aged <50 years, those without children, and those with missing outcome data, the final analytical sample comprised a total of 189,036 participants with 483,685 observations across all cohorts (CHARLS: 54,301; HRS: 84,004; ELSA: 37,187; SHARE: 232,006; KLoSA: 36,480; MHAS: 39,707). CHARLS, the China Health and Retirement Longitudinal Study; HRS, the Health and Retirement Study; ELSA, the English Longitudinal Study of Ageing; SHARE, the Survey of Health, Ageing and Retirement in Europe; KLoSA, the Korea Longitudinal Study of Aging; MHAS, the Mexican Health and Ageing Study.

Measures

Intergenerational contact

Intergenerational contact was operationalized as a composite measure capturing both structural and relational dimensions of parent–child relationships. Structural proximity was assessed based on co-residence with children, while relational closeness was measured by the frequency of contact with children, including in-person interactions as well as communication via telephone, mail, or email. Based on these two dimensions, participants were categorized into three mutually exclusive groups: co-residence, living apart with frequent (at least weekly) contact, and living apart with infrequent contact (Supplementary Table S1). Co-residence referred to participants who lived with at least one child. For participants not living with any child, weekly contact was assessed using cohort-specific definitions. CHARLS, HRS, SHARE, and MHAS measured contact at the family level (respondent and/or spouse), whereas ELSA, and KLoSA measured contact at the individual level (respondent only). Participants reporting at least weekly contact with any child were classified as living apart with frequent contact, whereas those reporting less frequent contact were classified as living apart with infrequent contact.

Depressive symptoms

Depressive symptoms were assessed using validated self-administered instruments, including the Centre for Epidemiologic Studies Depression Scale (CES-D) and the Euro-Depression Scale (Euro-D), depending on the cohort. HRS and ELSA employed an 8-item CES-D (CESD-8); CHARLS and KLoSA used a 10-item version (CESD-10); MHAS used a 9-item version (CESD-9); and SHARE used the Euro-D scale. The maximum possible scores were 8 for CESD-8, 30 for CESD-10, 9 for CESD-9, and 12 for Euro-D. Clinically significant depressive symptoms were defined using established cut-off points: scores of 3 or higher for HRS and ELSA, 10 or higher for CHARLS and KLoSA, 5 or higher for MHAS, and 4 or higher for SHARE.33, 34, 35, 36 Both CES-D and Euro-D items were dichotomized, with responses coded as “yes” to indicate endorsement of a depressive symptom and “no” to indicate its absence.

Covariates

Covariates were selected a priori based on prior literature and included demographic characteristics (age and gender), socioeconomic factors (educational attainment, employment status, marital status, urbanicity and household income), health status indicators (limitations in activities of daily living [ADL], limitations in instrumental activities of daily living [IADL], and number of chronic diseases) and health behaviors (smoking status, alcohol consumption, participation in social activities, physical activity).^13,37,38 Detailed definitions and coding schemes for all covariates are provided in Supplementary Table S1.

Ethics statement

This study used de-identified, publicly available secondary data from CHARLS, HRS, ELSA, SHARE, KLoSA, and MHAS. Ethical approval for each original survey was obtained by the respective study teams from their relevant institutional review boards or ethics committees, and informed consent was obtained from all participants at the time of data collection. Because the present study involved secondary analyses of de-identified data, no additional ethical approval or participant consent was required.

Statistical analysis

Statistical analyses were conducted separately for each of the six cohorts. The assumption that data were missing at random was supported by results from Little's MCAR test.³⁹ Accordingly, multiple imputation by chained equations was used to handle missing data (Supplementary Tables S2–S10). Descriptive statistics were based on the observed data. All other analyses used the imputed datasets, with results pooled via Rubin's rules.

The model expression of generalized estimating equations (GEE) is as follows:

$$logit\left(\mathrm{Pr}\left(Y_{it}=1\right)\right)={\beta }_0+{\beta }_1{X}_{it}+{\gamma }^T{C}_{it}\text{,}$$

where $Y_{it}$ referred to the depressive symptoms (1 for Yes, 0 for No) for participant $i$ at survey wave $t$ (t indexed survey waves within each cohort). $X_{it}$ was the intergenerational contact measured in the same wave. This model estimates the concurrent association—i.e., the effect of intergenerational contact in wave $t$ on mental health in wave $t$.

To account for the correlation among repeated observations within individuals, we specified an exchangeable working correlation structure, which assumes a constant correlation between any two observations contributed by the same participant across survey waves:

$$Corr\left(Y_{it},Y_{is}\right)=\alpha ,\text{for}\text{all}t\ne s\text{.}$$

Longitudinal analyses were performed using GEE with logistic regression and an exchangeable correlation structure to estimate population–average associations while accounting for repeated observations within individuals.

Effect estimates are reported as odds ratios (ORs) with corresponding 95% confidence intervals (CIs). To assess the robustness of the associations, a sequence of five models with progressively adjusted covariates were fitted. Model 1 was an unadjusted model. Model 2 adjusted for demographic characteristics, including age and gender. Model 3 additionally adjusted for socioeconomic status, including educational attainment, employment status, marital status, residential location, and household income. Model 4 further incorporated health status indicators, including limitations in ADL, limitations in IADL, and the number of chronic diseases. Model 5 represented the fully adjusted model and additionally controlled for health-related behaviors, including smoking status, alcohol consumption, participation in social activity, and physical activity. Because our focus was on cohort-specific associations and contextual heterogeneity rather than estimation of an overall pooled effect, we did not conduct a formal meta-analysis.

To assess the robustness of the findings and reduce concerns regarding reverse causality, the GEE analyses were repeated in a restricted sample of participants who were free of depressive symptoms at baseline. We performed stratified analyses by gender, age groups (50–59, 60–69, 70–79, and ≥80 years), educational attainment, employment status, marital status, urbanicity and household income, limitations in ADL, limitations in IADL, and chronic diseases (yes and no), smoking status, alcohol consumption, participation in social activities, and physical activity. To account for multiple comparisons, all subgroup analyses were adjusted using the Benjamini–Hochberg false discovery rate (FDR) procedure as a post-hoc analysis.

Because the primary objective was to estimate within-cohort associations between intergenerational contact and depressive symptoms rather than population-level prevalence, regression analyses were conducted without survey weights and adjusted for demographic, socioeconomic, health status, and health behavior covariates. This approach is consistent with prior cross-national studies using harmonized ageing cohorts.^37,38

All statistical analyses were performed using Stata (version 18·0). A two-sided p value < 0·05 was considered statistically significant. Statistical significance for subgroup analyses was determined based on FDR-adjusted p values <0·05. Data visualization was conducted using R (version 4·5·1).

Role of the funding source

The funder of the study had no role in study design, data collection, data analysis, data interpretation, or writing of the report.

Results

The baseline characteristics of included observations across the six cohort studies are summarized in Table 1. Mean participant age ranged from 62·3 years in China to 68·0 years in England, with broadly similar age distributions across cohorts. Women comprised the majority of participants in all studies, with the proportion of women ranging from 50·9% (27,622 of 54,301) in China to 58·9% (49,453 of 84,004) in the US. The prevalence of depressive symptoms varied substantially across cohorts, from 20·3% (7564 of 37,187) in England to 33·3% (18,108 of 54,301) in China, underscoring marked cross-national differences in mental health burden among older adults. Marked cross-national variation in intergenerational contact patterns was observed (Table 1). Co-residence was the predominant pattern in Mexico (73·7%, 29,247 of 39,707), and China (48·5%, 26,363 of 54,301). In contrast, living apart with frequent contact was more common in England (57·8%, 21,486 of 37,187), continental Europe (70·2%, 162,925 of 232,006), and in the US (37·1%, 31,179 of 84,004). The dominant intergenerational contact patterns remained stable across most cohorts except South Korea, where the predominant pattern transitioned from co-residence in waves 3–5 to living apart with frequent contact in waves 6–7 (Fig. 2).

Table 1.

Descriptive characteristics of included observations across six cohort studies.

Variables	CHARLS N = 54,301	HRS N = 84,004	ELSA N = 37,187	SHARE N = 232,006	KLoSA N = 36,480	MHAS N = 39,707
Age, mean (SD)	62·3 (8·5)	67·1 (10·8)	68·0 (9·6)	67·6 (10·0)	67·4 (10·4)	65·4 (9·8)
Gender
Man	26,679 (49·1%)	34,551 (41·1%)	16,208 (43·6%)	101,007 (43·5%)	15,601 (42·8%)	16,949 (42·7%)
Woman	27,622 (50·9%)	49,453 (58·9%)	20,979 (56·4%)	130,999 (56·5%)	20,879 (57·2%)	22,758 (57·3%)
Educational attainment
Less than lower secondary	47,715 (87·9%)	15,196 (18·1%)	9768 (26·3%)	92,210 (39·7%)	21,310 (58·4%)	33,463 (84·3%)
Upper secondary & vocational training	5593 (10·3%)	49,852 (59·3%)	17,733 (47·7%)	89,179 (38·4%)	11,351 (31·1%)	1277 (3·2%)
Tertiary	986 (1·8%)	18,938 (22·5%)	6426 (17·3%)	50,617 (21·8%)	3819 (10·5%)	4572 (11·5%)
Missing	7 (<0·1%)	18 (<0·1%)	3260 (8·8%)	–	–	395 (1·0%)
Household income
Q1 (deprived)	8234 (15·2%)	20,313 (24·2%)	8912 (24·0%)	51,478 (22·2%)	6255 (17·1%)	10,057 (25·3%)
Q2	8469 (15·6%)	21,234 (25·3%)	9284 (25·0%)	52,362 (22·6%)	6092 (16·7%)	9381 (23·6%)
Q3	8153 (15·0%)	21,372 (25·4%)	9341 (25·1%)	52,227 (22·5%)	6140 (16·8%)	10,033 (25·3%)
Q4 (affluent)	7971 (14·7%)	21,085 (25·1%)	9058 (24·4%)	52,169 (22·5%)	6104 (16·7%)	9774 (24·6%)
Missing	21,474 (39·5%)	–	592 (1·6%)	23,770 (10·2%)	11,889 (32·6%)	462 (1·2%)
Marital status
Married or partnered	47,039 (86·6%)	51,860 (61·7%)	26,117 (70·2%)	168,599 (72·7%)	28,131 (77·1%)	30,394 (76·5%)
Single	7258 (13·4%)	32,130 (38·2%)	11,061 (29·7%)	60,887 (26·2%)	8146 (22·3%)	9313 (23·5%)
Missing	4 (<0·1%)	14 (<0·1%)	9 (<0·1%)	2520 (1·1%)	203 (0·6%)	–
Employment status
Not working	927 (1·7%)	7997 (9·5%)	4264 (11·5%)	31,978 (13·8%)	9964 (27·3%)	18,921 (47·7%)
Working	33,195 (61·1%)	26,582 (31·6%)	9904 (26·6%)	54,855 (23·6%)	14,499 (39·7%)	15,280 (38·5%)
Retired	19,485 (35·9%)	49,425 (58·8%)	22,591 (60·7%)	139,094 (60·0%)	11,080 (30·4%)	5415 (13·6%)
Missing	694 (1·3%)	–	428 (1·2%)	6079 (2·6%)	937 (2·6%)	91 (0·2%)
Urbanicity
Urban	21,684 (39·9%)	58,872 (70·1%)	–	146,500 (63·1%)	27,305 (74·8%)	27,862 (70·2%)
Rural	32,617 (60·1%)	21,197 (25·2%)	–	74,381 (32·17%)	9175 (25·2%)	11,845 (29·8%)
Missing	–	3935 (4·7%)	–	11,125 (4·8%)	–	–
Limitations in ADL
No	44,699 (82·3%)	69,766 (83·1%)	30,914 (83·1%)	203,713 (87·8%)	34,826 (95·5%)	33,199 (83·6%)
Yes	9429 (17·4%)	13,970 (16·6%)	6266 (16·8%)	27,551 (11·9%)	1654 (4·5%)	6352 (16·0%)
Missing	173 (0·3%)	268 (0·3%)	7 (<0·1%)	742 (0·3%)	–	156 (0·4%)
Limitations in IADL
No	43,638 (80·4%)	68,405 (81·4%)	33,411 (89·8%)	208,813 (90·0%)	32,586 (89·3%)	33,309 (83·9%)
Yes	10,355 (19·1%)	12,212 (14·5%)	3769 (10·1%)	22,451 (9·7%)	3893 (10·7%)	4494 (11·3%)
Missing	308 (0·6%)	3387 (4·0%)	7 (<0·1%)	742 (0·3%)	1 (<0·1%)	1904 (4·8%)
Number of chronic diseases, mean (SD)	0·7 (0·9)	1·2 (1·0)	0·8 (0·9)	0·9 (0·9)	0·7 (0·9)	0·9 (0·8)
Participation in social activities
No	28,173 (51·9%)	15,500 (18·5%)	20,697 (55·7%)	129,218 (55·7%)	25,230 (69·2%)	19,409 (48·9%)
Yes	26,106 (48·1%)	15,651 (18·6%)	10,037 (27·0%)	96,461 (41·6%)	11,249 (30·8%)	20,278 (51·1%)
Missing	22 (<0·1%)	52,853 (62·9%)	6453 (17·4%)	6327 (2·7%)	1 (<0·1%)	20 (0·1%)
Alcohol consumption
No	29,615 (54·5%)	36,917 (43·9%)	4437 (11·9%)	119,030 (51·3%)	2443 (6·7%)	29,793 (75·0%)
Yes	24,637 (45·4%)	47,059 (56·0%)	29,086 (78·2%)	111,922 (48·2%)	34,037 (93·3%)	9905 (24·9%)
Missing	49 (0·1%)	28 (<0·1%)	3664 (9·9%)	1054 (0·5%)	–	9 (<0·1%)
Smoking status
Non-smoker	30,516 (56·2%)	37,091 (44·2%)	14,032 (37·7%)	125,128 (53·9%)	25,367 (69·5%)	24,332 (61·3%)
Past smoker	7072 (13·0%)	34,827 (41·5%)	19,023 (51·2%)	47,872 (20·6%)	6312 (17·3%)	10,700 (26·9%)
Current smoker	13,939 (25·7%)	11,663 (13·9%)	4101 (11·0%)	30,052 (13·0%)	4801 (13·2%)	4657 (11·7%)
Missing	2774 (5·1%)	423 (0·5%)	31 (0·1%)	28,954 (12·5%)	–	18 (<0·1%)
Physical activity
Yes	19,947 (36·7%)	57,735 (68·7%)	28,281 (76·1%)	190,123 (81·9%)	12,446 (34·1%)	14,679 (37·0%)
No	11,984 (22·1%)	26,068 (31·0%)	8903 (23·9%)	40,852 (17·6%)	24,034 (65·9%)	25,009 (63·0%)
Missing	22,370 (41·2%)	201 (0·2%)	3 (<0·1%)	1031 (0·4%)	–	19 (0·0%)
Intergenerational contact
Co-residence	26,363 (48·5%)	18,530 (22·1%)	9472 (25·5%)	50,641 (21·8%)	15,095 (41·4%)	29,247 (73·7%)
Living apart with frequent contact	22,560 (41·5%)	31,179 (37·1%)	21,486 (57·8%)	162,925 (70·2%)	15,940 (43·7%)	9538 (24·0%)
Living apart with infrequent contact	5042 (9·3%)	3803 (4·5%)	3027 (8·1%)	13,755 (5·9%)	5157 (14·1%)	922 (2·3%)
Missing	336 (0·6%)	30,492 (36·3%)	3202 (8·6%)	4685 (2·0%)	288 (0·8%)	–
Depressive symptoms
No	36,193 (66·7%)	65,479 (77·9%)	29,623 (79·7%)	168,021 (72·4%)	26,689 (73·2%)	27,453 (69·1%)
Yes	18,108 (33·3%)	18,525 (22·1%)	7564 (20·3%)	63,985 (27·6%)	9791 (26·8%)	12,254 (30·9%)

Note: Urbanicity was not available in ELSA.

CHARLS, the China Health and Retirement Longitudinal Study; HRS, the Health and Retirement Study; ELSA, the English Longitudinal Study of Ageing; SHARE, the Survey of Health, Ageing and Retirement in Europe; KLoSA, the Korea Longitudinal Study of Aging; MHAS, the Mexican Health and Ageing Study; ADL, activities of daily living; IADL, instrumental activities of daily living; SD, standard deviation; Q1, first quartile; Q2, second quartile; Q3, third quartile; Q4, fourth quartile.

Fig. 2.

Patterns of intergenerational contacts across different waves of the six cohort studies. CHARLS, the China Health and Retirement Longitudinal Study; HRS, the Health and Retirement Study; ELSA, the English Longitudinal Study of Ageing; SHARE, the Survey of Health, Ageing and Retirement in Europe; KLoSA, the Korea Longitudinal Study of Aging; MHAS, the Mexican Health and Ageing Study.

Fig. 3 and Supplementary Table S11 illustrate the associations between intergenerational contact and depressive symptoms. In five of the six cohorts—China, the US, continental Europe, South Korea, and Mexico—older adults who lived apart from their children with infrequent contact had higher odds of depressive symptoms than those who co-resided with their children. Specifically, compared to co-residence, living apart with infrequent contact was consistently associated with increased odds of depressive symptoms in China (OR 1·19, 95% CI 1·11–1·27), the US (OR 1·18, 95% CI 1·10–1·28), continental Europe (OR 1·10, 95% CI 1·05–1·16), South Korea (OR 1·24, 95% CI 1·14–1·35), and Mexico (OR 1·43, 95% CI 1·23–1·66). Interestingly, compared to co-residence, living apart with frequent contact was associated with a lower risk of depressive symptoms in European cohorts, including England (OR 0·88, 95% CI 0·81–0·96) and continental Europe (OR 0·93, 95% CI 0·91–0·96). The figure revealed both consistency and heterogeneity across cohorts.

Fig. 3.

Association between intergenerational contact and depressive symptoms. Odds ratios (OR) with 95% confidence intervals (CI) are shown for each dataset. The reference group is co-residence. The model was fully adjusted for demographic characteristics (age and gender), socioeconomic factors (educational attainment, employment status, marital status, urbanicity and household income), health status indicators (limitations in activities of daily living [ADL], limitations in instrumental activities of daily living [IADL], and number of chronic diseases) and health behaviors (smoking status, alcohol consumption, participation in social activities, physical activity). ∗p < 0·05. CHARLS, the China Health and Retirement Longitudinal Study; HRS, the Health and Retirement Study; ELSA, the English Longitudinal Study of Ageing; SHARE, the Survey of Health, Ageing and Retirement in Europe; KLoSA, the Korea Longitudinal Study of Aging; MHAS, the Mexican Health and Ageing Study.

Fig. 4 and Supplementary Table S12 present the results for incident depressive symptoms, revealing patterns similar to those observed for depressive symptom prevalence. Specifically, compared to co-residence, living apart with infrequent contact was associated with a higher risk of depressive symptoms in the US (OR 1·20, 95% CI 1·05–1·36), South Korea (OR 1·19, 95% CI 1·04–1·36), and Mexico (OR 1·55, 95% CI 1·18–2·03). In contrast, compared to co-residence, living apart with frequent contact was associated with a reduced risk of incident depressive symptoms in continental Europe (OR 0·92, 95% CI 0·86–0·97).

Fig. 4.

Association between intergenerational contact and incident depressive symptoms. Odds ratios (OR) with 95% confidence intervals (CI) are shown for each dataset. The reference group is co-residence. The model was fully adjusted for demographic characteristics (age and gender), socioeconomic factors (educational attainment, employment status, marital status, urbanicity and household income), health status indicators (limitations in activities of daily living [ADL], limitations in instrumental activities of daily living [IADL], and number of chronic diseases) and health behaviors (smoking status, alcohol consumption, participation in social activities, physical activity). ∗p < 0·05. CHARLS, the China Health and Retirement Longitudinal Study; HRS, the Health and Retirement Study; ELSA, the English Longitudinal Study of Ageing; SHARE, the Survey of Health, Ageing and Retirement in Europe; KLoSA, the Korea Longitudinal Study of Aging; MHAS, the Mexican Health and Ageing Study.

Subgroup analyses revealed cross-national heterogeneity in these associations (Fig. 5; Supplementary Table S13). After correction for multiple comparisons using the Benjamini–Hochberg false discovery rate procedure, several subgroup associations remained statistically significant. Compared with co-residence, living apart with infrequent contact was associated with a higher risk of depressive symptoms among adults aged 50–59 years in the US, continental Europe, South Korea, and Mexico, and among adults aged 60–69 years in China, the US, continental Europe, and Mexico. Among the oldest-old (≥80 years), living apart with infrequent contact was also associated with a higher risk of depressive symptoms in South Korea and Mexico. In contrast, living apart with frequent contact was associated with a lower risk of depressive symptoms among adults aged 60–69 years in China and 60–70 years in continental Europe.

Fig. 5.

Subgroup analyses on the association between intergenerational contact and depressive symptoms. Odds ratios (OR) with 95% confidence intervals (CI) are shown for each dataset. The reference group is co-residence. The age-group-stratified model was adjusted for demographic characteristics (age[as a continuous variable] and gender), socioeconomic factors (educational attainment, employment status, marital status, urbanicity and household income), health status indicators (limitations in activities of daily living [ADL], limitations in instrumental activities of daily living [IADL], and number of chronic diseases) and health behaviors (smoking status, alcohol consumption, participation in social activities, physical activity). The gender-stratified model was adjusted for demographic characteristics (age), socioeconomic factors (educational attainment, employment status, marital status, urbanicity and household income), health status indicators (limitations in ADL, limitations in IADL, and number of chronic diseases) and health behaviors (smoking status, alcohol consumption, participation in social activities, physical activity). ∗p < 0·05 after Benjamini-Hochberg false discovery rate (FDR) correction. CHARLS, the China Health and Retirement Longitudinal Study; HRS, the Health and Retirement Study; ELSA, the English Longitudinal Study of Ageing; SHARE, the Survey of Health, Ageing and Retirement in Europe; KLoSA, the Korea Longitudinal Study of Aging; MHAS, the Mexican Health and Ageing Study.

Gender-stratified analyses were largely consistent with the overall findings but revealed several cohort-specific differences. For instance, living apart with frequent contact was associated with a reduced risk of depressive symptoms compared to co-residence among men in China. However, this protective association was not observed in the full China sample. Similarly, in continental Europe, living apart with infrequent contact did not show the increased risk observed in overall sample among men.

Discussion

To our knowledge, this is the first cross-national study using longitudinal panel data to examine the association between intergenerational contact and depressive symptoms among midlife and older adults across 33 countries. In most cohorts, living apart from children without frequent contact was associated with a higher risk of depressive symptoms compared to both co-residence and living apart with regular contact. Additionally, several European cohorts showed that living apart while maintaining frequent contact with children was associated with a lower risk of depressive symptoms compared to co-residence. These results underscore that the relationship between intergenerational contact and mental health is highly context-dependent. In many settings, residential separation combined with infrequent contact may signify reduced emotional support and reduced social engagement, which can foster loneliness and contribute to both the prevalence and onset of depressive symptoms.

Patterns of intergenerational contact varied substantially across cohorts, likely reflecting sociocultural norms and structural conditions.⁴⁰ Mexico had the highest co-residence rates, shaped not only by economic constraints but also by cultural values such as familismo, which emphasize familial obligation and multigenerational living.^41,42 By contrast, South Korea showed a relatively high prevalence of living apart with infrequent contact, even compared with many high-income Western countries. This finding aligns with Seoul Metropolitan Government surveys showing a growing reluctance among older adults to co-reside with children.⁴³ Driven by urbanization and shrinking household sizes, this shift reflects evolving intergenerational expectations but reveals a critical structural mismatch: while residential independence is rising, social support infrastructure remains underdeveloped. Without robust state or community safeguards, infrequent intergenerational contact leaves older adults vulnerable to isolation and significantly elevates the risk of depressive symptoms.⁴⁴

Our results show that limited intergenerational contact may be associated with poorer mental health in a wide range of contexts. At the same time, the magnitude and statistical significance of associations varied across cohorts, and associations involving living apart with frequent contact were less consistent. These differences suggest that the implications of intergenerational contact may depend on broader contextual factors, including cultural expectations regarding family involvement, living arrangements, welfare systems, and the availability of alternative sources of social support. In England and other European countries, midlife and older adults who lived apart from their children but maintained frequent contact reported fewer depressive symptoms than those who co-resided. These settings provide social environments where independent living is the norm and is supported by robust public welfare systems.^12,45 In such contexts, intergenerational co-residence often stems from necessity rather than personal preference, potentially leading to intergenerational strain or diminished autonomy.⁴³ Regular contact without shared residence may provide emotional support while preserving independence and autonomy, thereby promoting psychological well-being.^46,47

Age-stratified analyses further illustrate the complexity of the relationship between intergenerational contact and depressive symptoms. In several cohorts, living apart with infrequent contact was associated with higher depressive symptom risk among younger-old adults. One possible explanation is that this period often coincides with the transition to an “empty nest,” during which reduced contact with children may be experienced as a loss of emotional support or family involvement.⁷ However, these patterns were not observed consistently across all countries or age groups, suggesting that the meaning and consequences of intergenerational contact likely depend on broader social, cultural, and family contexts.

Among the oldest-old, living apart with infrequent contact remained associated with higher depressive symptom risk in South Korea and Mexico. In these settings, where family-based support may play a particularly important role in later life, limited contact with children may reflect reduced access to emotional and practical support. By contrast, age-specific associations were less evident in some Western cohorts, potentially reflecting differences in social support systems, living arrangements, or cohort-specific expectations regarding family involvement.^45,48 However, these interpretations remain speculative and should be viewed cautiously.

Notably, several associations observed in the overall analyses were not consistently replicated in age-stratified analyses. For example, the protective association of frequent contact observed in the overall England sample was not evident within individual age groups. This discrepancy may reflect reduced statistical power after stratification, as well as genuine heterogeneity across age groups. Future studies are needed to better understand how the role of intergenerational contact may change across different stages of later life. For example, qualitative or mixed-methods data on the quality and context of intergenerational relationships could provide greater insight into the mechanisms underlying age-related differences, while longitudinal studies with repeated assessments of family relationships, social support, and mental health would help determine whether these age-related patterns reflect developmental changes or cohort effects.

Gender differences were also evident. Men in China, England, and continental Europe who maintained frequent contact with their children reported fewer depressive symptoms compared to co-residence. While men generally maintain less frequent contact with children than women throughout adulthood, our multi-study analysis reveals a significant mental health advantage for older men who sustain frequent child contact. Infrequent contact was associated with increased risk in most cohorts, particularly among women in China, South Korea, and Mexico. In many cultural contexts, women play central roles in maintaining family relationships and caregiving networks.⁴⁹ When intergenerational contact declines, women may experience greater emotional strain and loss of social support. Their mental health may therefore be influenced not only by contact itself but also by the responsibilities associated with sustaining family connections.

The association between living apart with infrequent contact and higher depressive symptom risk was observed in both the full-sample and incident-case analyses. This consistency across analyses strengthens confidence in the robustness of the finding and suggests that it is less likely to be explained solely by baseline depression or reverse causality. In contrast, associations involving living apart with frequent contact, as well as several subgroup findings, were not consistently replicated in the incident-case analysis. These findings should therefore be interpreted more cautiously, as they may be more sensitive to sample composition, baseline depressive symptoms, or other unmeasured factors.

Our findings suggest several possible implications. First, the full-sample analysis identifies intergenerational contact patterns that are associated with the current burden of depressive symptoms among midlife and older adults, whereas the incident-case analysis provides additional evidence regarding the temporal ordering of these associations. Taken together, the findings highlight the potential importance of intergenerational contact for mental health in later life while underscoring the need for further research to clarify the underlying mechanisms and causal pathways. Second, the mental health effects of intergenerational contact depend on both cultural norms and the structure of social welfare systems. Third, gender inequalities in caregiving and emotional labor shape how family relationships influence mental health in later life. Policies aimed at promoting mental well-being among aging populations should therefore extend beyond encouraging family involvement alone and instead strengthen formal care systems, community support networks, and social welfare protections—particularly for older women.

This study has several limitations. First, although we applied a harmonized three-category classification of intergenerational contact, the underlying survey items differed across cohorts. Some cohorts assessed contact at the respondent level, whereas others measured contact at the family or household level. These differences may have introduced measurement heterogeneity and limit direct comparisons of effect sizes across cohorts. Accordingly, our findings should be interpreted primarily as cohort-specific associations, with greater emphasis placed on the consistency of overall patterns and directions of association than on precise differences in effect magnitude between cohorts. Second, as with all observational studies, our findings reflect associations rather than causal relationships, due to the potential for residual confounding and reverse causality. However, our analyses of incident depressive symptoms strengthened the temporal ordering and supported the plausibility of the observed associations. Additionally, variations in response rates across cohorts and waves may have introduced selection and attrition biases, as non-respondents and participants lost to follow-up could differ systematically from those retained. Third, we were not able to differentiate between face-to-face and digital contact, assess relationship quality (e.g., emotional closeness, conflict, or caregiving burden), or capture variation in contact across individual children. Consequently, the findings should be interpreted as reflecting broad patterns of parent–child living and contact arrangements rather than the full complexity of intergenerational relationships. In addition, we did not include measures of non-family social support, such as relationships with friends, peers, neighbors, or community organizations. These forms of social connection may independently influence mental health and may also buffer or amplify the associations between intergenerational contact and depressive symptoms. Although we adjusted for participation in social activities, this variable represents only a crude proxy for broader social support networks. Future studies should incorporate more comprehensive assessments of both family and non-family social relationships, distinguish contact modalities, and evaluate relationship quality to better understand the social determinants of mental health in later life. Fourth, we included as many relevant and available covariates as possible in our models, but cognitive functioning was not adjusted for because it may either lie on the causal pathway or function as a collider between intergenerational contact and depressive symptoms; adjusting for it could risk overadjustment bias. Another limitation is that child-level age information was not consistently available across the harmonized cohorts. As a result, we could not fully distinguish intergenerational contact with adult children from contact with minor children or restrict the analysis to respondents with adult children only. Because the meaning and implications of parent–child contact may differ according to the child's developmental stage, this limitation may have introduced some heterogeneity into the exposure measure and should be considered when interpreting the findings. Fifth, our measure of intergenerational contact was based on whether respondents had at least weekly contact with any child and therefore did not capture variation in contact frequency across individual children within the same family. As a result, respondents with frequent contact with one child but limited contact with others were classified as having frequent contact. Future studies with child-level data are needed to examine how heterogeneity in contact patterns across multiple children may influence mental health outcomes. Finally, the wider confidence intervals observed for participants with infrequent intergenerational contact likely reflect the smaller size of this subgroup, particularly in stratified analyses, and indicate greater statistical uncertainty around these estimates. Accordingly, subgroup-specific findings with wide confidence intervals should be interpreted cautiously.

In conclusion, our findings suggest that the association between intergenerational contact and depressive symptoms among midlife and older adults is shaped by both family relationships and broader sociocultural contexts. While causal interpretations should be avoided, the findings highlight the potential value of considering both family-based and community-based sources of support when promoting mental well-being in ageing populations. Future research is needed to clarify causal pathways and to determine whether interventions that strengthen intergenerational and social connections can improve mental health outcomes across different cultural settings.

Contributors

LZ, BW, and YL conceptualized the study. RG and XL conducted data curation and formal analysis. RG, XL, YL and BW led the data interpretation. RG and XL wrote the initial draft, and YZ, MW, XF, LZ, BW, and YL provided critical revision of the manuscript. YL supervised the study. RG and XL accessed and verified the underlying data. All authors have read and approved the final version of the manuscript.

Data sharing statement

The original survey datasets from CHARLS, HRS, ELSA, SHARE, KLoSA, and MHAS are freely available to all bona fide researchers.

Editor note

The Lancet Group takes a neutral position with respect to territorial claims in published maps and institutional affiliations.

Declaration of interests

We declare no competing interests.