Persistent socio-demographic inequalities in depression among older adults in China during and after COVID-19: a systematic review and meta-analysis

Abstract

Objective

To examine socio-demographic inequalities in depressive symptoms among older adults in China during and after the COVID-19 pandemic and assess whether these inequalities remained stable or changed over time.

Methodology

This systematic review and meta-analysis followed PRISMA guidelines. Nine English and Chinese databases were searched from 1 January 2020 to 15 June 2026. Eligible studies were conducted among community-dwelling adults aged ≥60 years in mainland China, used probability sampling and standardized measures of depressive symptoms, and reported cross-sectional data. Study quality was assessed using the 11-item Agency for Healthcare Research and Quality checklist. Potential publication bias was assessed using Begg’s and Egger’s tests and trim-and-fill analysis. Year-stratified prevalence and pooled odds ratios (ORs) were estimated for disparities by gender, marital status, urban–rural residence, living arrangement and educational attainment.

Results

Seventy-nine studies involving 199,116 older adults from 21 provincial-level regions were included; 52 studies were rated as high quality and 27 as moderate quality, with none rated as low quality. Pooled prevalence fluctuated across investigation years, from 16.6% (95% CI 12.1–22.4) in 2022 to 22.9% (95% CI 18.3–28.3) in 2024–2025, with substantial between-study heterogeneity (overall I² = 99.55%). Higher odds of depressive symptoms were observed among women, unmarried older adults, rural residents, those living alone and those with lower educational attainment. The largest disparities were related to education (OR 1.693, 95% CI 1.388–2.060), living arrangement (OR 1.671, 95% CI 1.437–1.945) and marital status (OR 1.652, 95% CI 1.467–1.860). Year-level analyses suggested limited temporal variation overall: marital-status-related inequality showed a small upward trend, whereas living-arrangement-related inequality showed a non-significant tendency toward attenuation. No broad systematic restructuring was observed.

Conclusion

Depressive symptoms among older adults in China fluctuated across the pandemic and post-pandemic periods, while socio-demographic inequalities remained broadly stable. Major public health disruption appeared more likely to expose and prolong pre-existing inequalities than fundamentally reshape them in the short term. These findings provide a useful comparative reference for understanding mental health vulnerability in aging societies facing public health emergencies.

Systematic review registration

https://www.crd.york.ac.uk/PROSPERO/view/CRD42025642200; identifier: CRD42025642200.

1. Introduction

Depression is a major contributor to the global burden of mental disorders and an important public health challenge in aging societies worldwide (1, 2). Reported prevalence of depression among older adults varies substantially across populations and studies. Recent global meta-analyses have reported pooled prevalence estimates of 31.74 and 35.1% (2, 3), with substantial variation across individual studies and population settings. In China, reported estimates have similarly varied widely, ranging from 3.8 to 50.32% (4, 5). Existing evidence consistently suggests that older adults who are female, live in rural areas, are unmarried, live alone, or have lower educational attainment are at higher risk of depression (6–10). These disparities are commonly linked to unequal access to social resources, social support and cumulative life-course disadvantage (11–13). Nevertheless, such patterns are not fully consistent across contexts. Cross-national evidence has shown substantial variation in socio-demographic inequalities in depression (6), and some studies have reported attenuation or even reversal of specific risk structures in particular regions or periods (14, 15). For example, a cross-sectional study of 1,173 community-dwelling older adults in Hunan, China, found no significant association between gender and depressive symptoms (16). Together, these findings suggest that the social patterning of late-life depression may be context-dependent rather than entirely static.

Major public health emergencies provide a distinctive context in which existing health inequalities may be exposed or amplified (17). Since the COVID-19 outbreak, a growing body of research has examined depression among older adults, but reported prevalence estimates remain highly heterogeneous, ranging from 2.79 to 37.34% (17–20). China, as the first country to experience the COVID-19 outbreak (18, 19), provides an informative setting for examining whether socio-demographic inequalities in depression among older adults persist or change under major public health disruption. Its very large and rapidly aging population is characterized by substantial geographic and socioeconomic disparities in access to health-related resources and public health services (21, 22). At the same time, China’s COVID-19 response underwent successive policy adjustments across the pandemic and post-pandemic periods, creating a changing social and public health environment (23). However, existing studies in China have focused mainly on overall prevalence estimates or pandemic-related temporal changes in depressive symptoms (18–20, 24, 25). Systematic evidence on whether socio-demographic inequalities in depression persisted or changed across these periods remains limited (26).

This study aimed to examine socio-demographic inequalities in depressive symptoms among older adults in China during and after the COVID-19 pandemic. Specifically, we assessed disparities by gender, marital status, residence, living arrangement and educational attainment, and explored whether these inequalities remained stable or changed across investigation years. By synthesizing evidence across regions and periods, this study aimed to inform the identification of high-risk groups in China and to provide comparative evidence for research on late-life depression inequalities in aging societies.

2. Methods

2.1. Study design and registration

This systematic review was registered with PROSPERO (CRD42025642200) and conducted in accordance with the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) 2020 statement (27). The research question and eligibility criteria were structured using the PICOS framework, as detailed below, and the PROSPERO record was updated to reflect the final review scope, eligibility criteria, search period and analysis plan.

2.2. Search strategy and data sources

To capture the most up-to-date evidence, we adopted a rolling search strategy and updated the database search approximately every 2 weeks until the final search on 15 June 2026. The search covered studies conducted from 1 January 2020, when WHO initiated its organizational emergency response to COVID-19 (19), to the same cutoff date. We searched PubMed, Embase, Web of Science, Scopus, PsycINFO, CNKI, Wanfang Data, CQVIP and CBMdisc. Search terms combined concepts related to older adults, depressive symptoms and China. The full search strategy for each database is provided in Supplementary Table 2. We also screened the reference lists of eligible studies as a supplementary search strategy; no additional eligible studies were identified.

2.3. Eligibility criteria

Eligibility criteria were developed using the PICOS framework, which encompasses Population, Intervention, Comparator, Outcome, and Study Design (26). We included studies conducted among community-dwelling older adults aged 60 years or older in mainland China, based on general-population samples and probability sampling methods, that assessed depressive symptoms using standardized measurement instruments and reported prevalence data collected after 1 January 2020. Studies focusing on specialized populations, including institutionalized older adults, inpatients, or populations defined by specific diseases or disaster exposures, were excluded. We also excluded studies without clear measurement tools or prevalence data, non-probability sampling studies, reviews, conference papers, case reports, editorials, comments, letters and unpublished manuscripts. Where multiple reports were based on the same dataset, only the most complete and recent study was retained.

2.4. Study selection

All retrieved records were imported into Microsoft Excel and screened in four stages by two reviewers (ZW and NW). After deduplication, titles and abstracts were screened for relevance to the target population and study design, followed by full-text review against the eligibility criteria. Disagreements were resolved through discussion, with consultation from additional reviewers (JZ, XF, YX and SD) when necessary. The study selection process is shown in the PRISMA flow diagram (Figure 1).

Figure 1.

Flow diagram of study selection for the meta-analysis.

2.5. Data extraction

Two reviewers (NW and XL) independently extracted data using a standardized data extraction form. Extracted information included study characteristics (author, publication year, survey year, region, sampling method, sample size and measurement instrument), overall prevalence of depressive symptoms, and subgroup data by gender, urban–rural residence, marital status, living arrangement and educational attainment where available. All extracted data were cross-checked twice to ensure accuracy.

2.6. Quality assessment

Study quality was assessed using the 11-item checklist recommended by the Agency for Healthcare Research and Quality (AHRQ) (28). Given that the evidence synthesized in this review was derived primarily from cross-sectional prevalence studies, the AHRQ checklist was selected because of its applicability to the methodological appraisal of cross-sectional studies (29). Each item was assigned one point if the criterion was met, yielding a total score ranging from 0 to 11; scores of 0–3, 4–7 and 8–11 were classified as low, moderate and high quality, respectively.

2.7. Statistical analysis

All quantitative syntheses were performed using Stata (version 18.0; StataCorp, College Station, TX, USA). Study-specific proportions of depressive symptoms were synthesized using random-effects models to account for substantial between-study heterogeneity and the expectation that true prevalence may vary across populations, settings and measurement approaches (30). Sensitivity analyses using the Freeman–Tukey double arcsine transformation for prevalence pooling were conducted to assess the robustness of year-stratified prevalence estimates. Heterogeneity was assessed using Cochran’s Q test and the I² statistic (31). High I² values are common in prevalence meta-analyses and were therefore interpreted alongside subgroup and meta-regression analyses rather than as a reason to preclude quantitative synthesis (32). Pooled prevalence estimates are presented with 95% confidence intervals (CIs) and were stratified by investigation year. For year-stratified analyses, adjacent later years were combined when the number of eligible studies within a single year was too small to support stable pooled estimates. Accordingly, 2024 and 2025 were combined in the prevalence analysis, and 2023–2024 estimates were combined for some year-stratified OR analyses. Univariable random-effects meta-regression was further conducted as an exploratory analysis to examine study-level sources of prevalence variation.

To describe subgroup patterns, subgroup analyses of pooled prevalence were conducted by investigation year, geographic region, gender, marital status, urban–rural residence, living arrangement, and educational attainment (junior secondary school or below vs. senior secondary school and above). Detailed results are provided in the Supplementary material.

To quantify socio-demographic inequalities in depressive symptoms, study-level 2 × 2 tables were constructed from the reported subgroup counts, and crude odds ratios (ORs) were pooled using random-effects models. Where applicable, ORs were further stratified by investigation year to examine temporal variations in inequality across the pandemic and post-pandemic phases. Differences between year strata were formally assessed using between-stratum heterogeneity statistics.

To explore temporal shifts in inequality, we conducted a secondary year-level trend analysis based on year-specific pooled ORs for each socio-demographic comparison. Because ratio effect measures are conventionally analyzed on the logarithmic scale, pooled ORs were transformed to ln(OR) prior to regression analysis (33). Similar applications of weighted regression to log-transformed odds ratios have been reported in previous meta-analytic research (34). Investigation year was then entered as a continuous predictor in an inverse-variance weighted linear regression, with weights derived from the variance of the year-specific pooled ln(OR). Standard errors for year-specific pooled ln(OR) were reconstructed from the corresponding 95% CIs on the log scale using standard methods (35, 36). Regression coefficients were exponentiated and expressed as exp.(β), representing the ratio of odds ratios (ROR) per year, so that the estimated slope could be interpreted on the original OR scale as the multiplicative change in the year-specific pooled OR associated with a 1-year increase in investigation year (37). Because this analysis was based on year-specific pooled estimates rather than study-level or individual-level observations, it was treated as an exploratory ecological analysis.

Potential small-study effects were assessed on the logit-transformed prevalence scale by visual inspection of funnel plot asymmetry, Begg’s rank correlation test and Egger’s regression test. Trim-and-fill analysis was further conducted to evaluate the potential influence of funnel plot asymmetry on the pooled prevalence estimate. Leave-one-out sensitivity analyses were subsequently conducted to evaluate the influence of individual studies on the pooled prevalence estimate and between-study heterogeneity.

3. Results

3.1. Literature search

A total of 24,259 records were identified across nine databases. After removal of 7,605 duplicates, 16,654 records were screened by title and abstract, and 2,037 full-text articles were assessed for eligibility. Ultimately, 79 studies met the inclusion criteria and were included in the final analysis (Figure 1).

3.2. Study characteristics and quality assessment

This review included 79 eligible studies conducted across 21 provincial-level regions in mainland China, including two mixed multi-province samples, encompassing a total of 199,116 participants aged 60 years and above. Survey years ranged from 2020 to 2025 (5, 16, 38–114), and individual study sample sizes ranged from 165 to 23,956 participants. All studies used probability sampling and standardized instruments to assess depressive symptoms. The most commonly used instruments were PHQ measures (n = 29), GDS measures (n = 28), and the CES-D (n = 14), while the remaining studies used other standardized instruments. Subgroup-specific data were available from 34 studies for gender, 21 for urban–rural residence, 26 for marital status, 18 for living arrangement and 19 for educational attainment. Study quality was assessed using the AHRQ tool; 27 studies were rated as moderate quality and 52 as high quality, with no studies rated as low quality (Supplementary Tables 3, 4). Additional subgroup-specific prevalence estimates are presented in Supplementary Table 5.

3.3. Year-stratified pooled prevalence of depressive symptoms among older adults

Between 2020 and 2025, the pooled prevalence of depressive symptoms among older adults in mainland China varied across investigation years. The pooled prevalence was 17.8% in 2020 (95% CI 12.9 to 24.1%, I² = 99.56%), 20.4% in 2021 (95% CI 13.6 to 29.3%, I² = 99.50%), 16.6% in 2022 (95% CI 12.1 to 22.4%, I² = 99.62%), and 21.9% in 2023 (95% CI 15.8 to 29.6%, I² = 99.60%). In studies conducted from 2024 to 2025, the pooled prevalence was 22.9% (95% CI 18.3 to 28.3%, I² = 97.20%) (Figures 2, 3; Supplementary Table 5). Given the substantial heterogeneity, univariable random-effects meta-regression was conducted to explore study-level sources of prevalence variation. Investigation year was not significantly associated with prevalence estimates, whereas region and measurement tool showed stronger associations, particularly the higher estimates observed in studies using CES-D compared with PHQ (Supplementary Table 6).

Figure 2.

Forest plot of depressive symptom prevalence among older adults in China.

Figure 3.

Year-stratified pooled prevalence of depressive symptoms among older adults in China. Points represent pooled prevalence estimates for each year stratum, vertical lines indicate 95% confidence intervals, and labels above the points indicate the number of included studies. The final category combined studies conducted in 2024 and 2025 because of the limited number of studies in each year.

3.4. Publication bias and sensitivity analysis

Visual inspection of the funnel plot suggested slight asymmetry. However, Begg’s rank correlation test (p = 0.059) and Egger’s regression test (p = 0.113) did not provide statistical evidence of small-study effects. Trim-and-fill analysis imputed no additional studies and produced the same pooled estimate as the observed analysis. Leave-one-out sensitivity analyses showed that omission of any single study did not materially alter the pooled prevalence estimate or the level of heterogeneity (Supplementary Table 9). Sensitivity analyses using the Freeman–Tukey double arcsine transformation produced slightly higher but broadly consistent year-stratified prevalence estimates, with no material change in the overall temporal pattern (Supplementary Table 10). These findings supported the overall robustness of the main results.

3.5. Socio-demographic disparities in depressive symptoms

Across all five socio-demographic dimensions, pooled prevalence estimates were consistently higher among disadvantaged groups than among their reference groups. Random-effects models further showed significant disparities in depression risk across all comparisons (Figure 4; Supplementary Table 7), with the largest pooled ORs observed for educational attainment (OR 1.693, 95% CI 1.388 to 2.060), living arrangement (OR 1.671, 95% CI 1.437 to 1.945) and marital status (OR 1.652, 95% CI 1.467 to 1.860), followed by gender (OR 1.476, 95% CI 1.319 to 1.662) and residence (OR 1.419, 95% CI 1.185 to 1.699), all p < 0.001. Detailed subgroup-specific prevalence and year-stratified OR estimates are provided in Supplementary Tables 5, 7 and Supplementary Figures 1–5.

Figure 4.

Pooled odds ratios for socio-demographic disparities in depressive symptoms among older adults in China. Points represent pooled odds ratios and horizontal lines indicate 95% confidence intervals. The vertical dashed line indicates OR = 1.

3.6. Year-level trend regression of year-specific pooled ORs

Year-level inverse-variance weighted regression of year-specific pooled ORs suggested that temporal changes in socio-demographic inequalities were generally small (Figures 5 and Supplementary Table 8). Marital-status-related inequality was the only comparison that reached conventional statistical significance, with the year-specific pooled OR increasing by an estimated 2.1% per year [exp(β) = 1.021, 95% CI: 1.016 to 1.026, p = 0.001]. Living-arrangement-related inequality showed a marginally non-significant tendency toward attenuation, with the year-specific pooled OR decreasing by an estimated 10.8% per year [exp(β) = 0.892, 95% CI: 0.788 to 1.009, p = 0.060]. No statistically significant temporal changes were observed for gender, residence or educational attainment.

Figure 5.

Year-stratified pooled odds ratios for socio-demographic disparities in depressive symptoms among older adults in China. Panels show disparities by (A) educational attainment, (B) living arrangement, (C) marital status, (D) gender, and (E) urban–rural residence. Dots indicate year-specific pooled odds ratios and vertical lines indicate 95% confidence intervals. The y-axis is displayed on a logarithmic scale to facilitate comparison across panels. For educational attainment, estimates for 2023 and 2024 were combined because year-specific data were sparse, and points are shown without a connecting line to avoid implying continuity across years with missing estimates. For urban–rural residence, estimates for 2023 and 2024 were also combined because year-specific data were sparse. Comparisons were lower versus higher educational attainment, living alone versus living with others, unmarried versus married, female versus male, and rural versus urban residence, respectively.

4. Discussion

4.1. Principal findings, interpretation and implications

One focus of the present study was to examine changes in the prevalence of depressive symptoms among older adults in China during and after the COVID-19 pandemic. Year-stratified pooled analyses showed an overall pattern of increase, decline and renewed increase between 2020 and 2025, with the lowest estimate observed in 2022 (16.6%) and higher estimates in the post-pandemic period. These fluctuations broadly coincide with changes in epidemic conditions and policy responses in China (115), while evidence from other settings also suggests that population mental health trajectories were not uniform across different phases of the pandemic and recovery period (116, 117). However, given the substantial between-study heterogeneity and the repeated cross-sectional nature of the evidence, these temporal patterns should be interpreted cautiously and should not be attributed to specific pandemic-related mechanisms.

More importantly, socio-demographic inequalities in depression remained broadly persistent across the pandemic and post-pandemic periods, consistent with evidence of social patterning in late-life mental health across different national settings (118, 119). The largest disparities were observed for educational attainment (OR 1.693, 95% CI 1.388–2.060), living arrangement (OR 1.671, 95% CI 1.437–1.945) and marital status (OR 1.652, 95% CI 1.467–1.860). These dimensions capture partly distinct but interconnected forms of social disadvantage. Lower educational attainment is associated with differences in socioeconomic position, health literacy and access to protective resources accumulated across the life course (6, 42, 43, 120, 121), while living alone and being unmarried may entail reduced access to day-to-day emotional, instrumental and caregiving support (9, 122–129). The persistence of these disparities despite substantial changes in the broader public health environment suggests that late-life mental health vulnerability remained strongly patterned by pre-existing social and material inequalities rather than being uniformly redistributed during the crisis.

Some variation across inequality dimensions was nevertheless observed. Marital-status-related inequality showed a small upward trend (exp(β) = 1.021), whereas the apparent attenuation in living-arrangement-related inequality was not statistically significant (exp(β) = 0.892). These divergent patterns caution against interpreting temporal variation through a single social-support mechanism. Changes in marital-status-related disparities may partly reflect loss-related vulnerability, including bereavement and widowhood (130–132), whereas living-arrangement-related differences may be more responsive to changes in social connection and community support (39, 133). Given the limited number of year-specific estimates and the exploratory nature of these analyses, however, these patterns require cautious interpretation.

Gender and urban–rural disparities further indicate that inequalities in late-life depression extend beyond educational and family domains. Older women consistently showed higher odds of depressive symptoms, in line with evidence linking gender disparities in later-life mental health to differences in caregiving burdens, adverse life experiences, social roles and cumulative resource disadvantage (134, 135). Rural older adults likewise showed higher odds than their urban counterparts, although international evidence on urban–rural disparities is less consistent (6, 136). In China, urban–rural differences in socioeconomic resources, healthcare and mental health service access, community support and opportunities for social participation may contribute to this pattern (137, 138). Evidence that such differences attenuate after adjustment for individual-, family- and community-level socioeconomic conditions further suggests that place of residence may operate partly as a spatial expression of broader structural inequalities rather than as an isolated risk factor (62).

A prominent feature of this study was the persistently high between-study heterogeneity, which remained substantial even after stratification by year, region and other subgroup dimensions. Univariable random-effects meta-regression showed that investigation year was not significantly associated with prevalence variation, whereas region and measurement tool were more closely related to between-study differences, particularly the higher estimates observed in studies using CES-D compared with PHQ. This suggests that the observed heterogeneity cannot be explained by temporal stage alone, but may reflect the combined influence of regional context, sampling frames and measurement-related variability, including differences between commonly used instruments such as the GDS-15, PHQ-9 and CES-D10, which is a common challenge in prevalence meta-analyses (33). Given the presence of genuine between-study variability, random-effects models were used to estimate mean effects under a distribution of true prevalence values (139), and emphasis was placed on the consistency and robustness of structural inequality patterns rather than on precise comparisons of absolute prevalence levels.

Taken together, the year-level fluctuation in prevalence and the relative stability of socio-demographic disparities suggest that major public health disruption may alter the overall burden of late-life depression without necessarily restructuring its underlying social distribution. This pattern is consistent with the broader literature on health inequalities (140) and echoes evidence from older adults in Hong Kong during the SARS epidemic, suggesting that the mental health consequences of major public health events may unfold within pre-existing social structures rather than simply overriding them (42). More recent international evidence during COVID-19 points in a similar direction. A scoping review of socioeconomic inequalities in depressive and anxiety symptoms found that inequalities persisted or increased in most analyses across high-income countries, despite considerable variation in their temporal trajectories (141). The present findings extend this evidence to older adults in China and across multiple dimensions of social stratification, suggesting that the persistence of mental health inequality under crisis conditions may represent a broader challenge for aging societies.

These findings also have implications for mental health preparedness during future public health emergencies. International evidence has emphasized the importance of monitoring socioeconomic inequalities over time and adapting preventive strategies to different phases of a crisis (141). For older populations, preparedness should therefore extend beyond preventing infection and monitoring average levels of psychological distress to identifying groups with persistent social vulnerability (142). Integrating routinely available socio-demographic and social-determinant information into mental health surveillance may facilitate risk stratification and more equitable allocation of preventive and supportive resources (143). In the present study, educational attainment, living arrangement, marital status, gender and place of residence emerged as readily identifiable markers of such vulnerability and may therefore help inform targeted screening, outreach and service planning. Maintaining social connection and access to community and mental health services may be particularly important for older adults with limited social or material resources, given longitudinal evidence linking social disconnectedness and perceived isolation to depressive and anxiety symptoms in later life (144). More broadly, declines in population-level prevalence should not be assumed to indicate corresponding reductions in inequality; monitoring both the overall mental health burden and its social distribution may therefore strengthen equity-oriented responses to future pandemics and other public health emergencies.

4.2. Strengths and limitations

This study has several strengths. First, it systematically synthesized evidence spanning both the pandemic and post-pandemic periods, allowing changes in the prevalence of depressive symptoms among older adults in China and their socio-demographic inequalities to be examined within a unified framework. Second, the included studies covered multiple provinces and both urban and rural settings across mainland China, and key dimensions of social stratification—including gender, marital status, residence, living arrangement and educational attainment—were compared using a consistent quantitative approach. Third, in addition to year-stratified pooled prevalence, we reported relative effect measures (ORs) and conducted exploratory year-level analyses based on year-specific pooled estimates, providing a more complete picture of the stability of inequality patterns under public health disruption. More broadly, evidence from China as a large and rapidly aging society offers useful contextual and comparative insight for the international literature on late-life depression inequalities.

Several limitations should also be acknowledged. First, despite acceptable overall study quality, substantial differences remained in sampling frames, survey timing, local context, and the instruments and cut-off values used to assess depressive symptoms, which probably contributed to the very high heterogeneity observed. Second, most included studies were cross-sectional, so causal inferences cannot be drawn regarding the observed associations between socio-demographic characteristics and depressive symptoms. In particular, the year-level regression of year-specific pooled ORs should be regarded as an exploratory ecological analysis, given that it was based on pooled estimates rather than study-level observations, involved a limited number of time points and may have been influenced by changes in study composition over time. Third, COVID-19 infection may be associated with depressive symptoms (145); however, information on individual infection history, frequency of infection and infection-related symptoms was not systematically available across the included studies. We were therefore unable to examine whether these infection-related characteristics were associated with depressive symptoms or contributed to the observed temporal patterns. Future studies incorporating individual-level infection and symptom data may help further distinguish infection-related mental health vulnerability from the broader social and public health consequences of the pandemic. Fourth, incomplete reporting of some stratified data may have reduced the precision and comparability of certain subgroup analyses. Fifth, although funnel plot-based tests, trim-and-fill analysis and leave-one-out sensitivity analyses did not indicate substantial instability in the main findings, selective publication or residual small-study effects cannot be fully excluded. Finally, this analysis was restricted to community-dwelling older adults in mainland China, and differences in healthcare systems, family support structures, cultural norms and public health responses may limit direct generalizability to other settings.

5. Conclusion

Depressive symptoms among older adults in China fluctuated across the pandemic and post-pandemic periods, while socio-demographic inequalities in depression risk remained broadly stable. Larger disparities were observed in relation to educational attainment, living arrangement and marital status, suggesting that major public health disruption did not fundamentally restructure existing inequalities in late-life depression but instead exposed and prolonged them in uneven ways. Evidence from China therefore provides a useful comparative reference for understanding how public health emergencies may shape mental health vulnerability in aging societies.

Funding Statement

The author(s) declared that financial support was received for this work and/or its publication. This work was supported by the Economic and Social Research Council (ESRC).

Data availability statement

The original contributions presented in the study are included in the article/Supplementary material, further inquiries can be directed to the corresponding author/s.

Author contributions

NW: Writing – original draft, Formal analysis, Writing – review & editing. XL: Writing – review & editing, Formal analysis, Writing – original draft. JZ: Supervision, Writing – review & editing. XF: Supervision, Writing – review & editing. YX: Supervision, Writing – review & editing. SD: Writing – review & editing, Supervision. ZW: Data curation, Supervision, Writing – review & editing, Writing – original draft, Conceptualization.

Conflict of interest

The author(s) declared that this work was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

Generative AI statement

The author(s) declared that Generative AI was used in the creation of this manuscript. ChatGPT-5.5 Thinking, developed by OpenAI, was used solely for English language editing and grammatical correction. No AI tool was used for literature screening, data extraction, data analysis, interpretation of findings, or drafting of scientific conclusions. The authors reviewed and approved all AI-assisted edits and take full responsibility for the content of the manuscript.

Any alternative text (alt text) provided alongside figures in this article has been generated by Frontiers with the support of artificial intelligence and reasonable efforts have been made to ensure accuracy, including review by the authors wherever possible. If you identify any issues, please contact us.

Publisher’s note

All claims expressed in this article are solely those of the authors and do not necessarily represent those of their affiliated organizations, or those of the publisher, the editors and the reviewers. Any product that may be evaluated in this article, or claim that may be made by its manufacturer, is not guaranteed or endorsed by the publisher.

Supplementary material

The Supplementary material for this article can be found online at: https://www.frontiersin.org/articles/10.3389/fpubh.2026.1929179/full#supplementary-material