Bidirectional longitudinal associations between social activity and depressive symptoms among Chinese older adults with multimorbidity: the moderating role of subjective life expectancy Authors: Xinru Li, Lei Shi, Shujuan Xiao, Benli Xue, Lingli Yang, Xiao Zheng, Xinyi Zhang, Mengjie Zhang, Yanming Liao, Chichen Zhang Published: 2026 ABSTRACT AbstractBackgroundSocial activity in older adults is often correlated with their depressive symptoms, and this correlation is stronger in older adults with multimorbidity than in those without. However, few studies have analyzed this longitudinal association over time. This study aimed to examine the bidirectional longitudinal associations between social activity and depressive symptoms in older adults with multimorbidity. Additionally, we examined subjective life expectancy’s (SLE) moderating association in this bidirectional longitudinal association.MethodsData were obtained from Waves 3 and 4 survey of 1435 older adults with multimorbidity (≥ 60 years) enrolled in the China Health and Retirement Longitudinal Study. We used cross-lagged panel models and grouped regression models to analyze the bidirectional longitudinal association between social activity and depressive symptoms, as well as the moderating association of subjective life expectancy, among Chinese older adults with multimorbidity.ResultsT1 social activity was significantly negatively associated with T2 depressive symptoms (β=-0.054, P < 0.050), and T1 depressive symptoms were significantly negatively associated with T2 social activity (β=-0.112, P < 0.010). SLE significantly moderated the association between T1 social activity and T2 depressive symptoms. Further, among participants with high SLE, the absolute value of the regression coefficient for the negative association of T1 social activity with T2 depressive symptoms (β=-0.134, P < 0.010) was larger than that for the association of T1 depressive symptoms with T2 social activity (β=-0.114, P < 0.050), indicating a more pronounced negative association for the former. All reported standardized coefficients were relatively modest; nevertheless, these statistically significant longitudinal associations are clinically and public health-relevant in this high-risk cohort of older adults with multimorbidity, who are disproportionately vulnerable to depressive symptom progression and social disengagement.ConclusionReciprocal longitudinal associations between social activity and depressive symptoms were observed among Chinese older adults with multimorbidity, with the negative longitudinal association from baseline depressive symptoms to subsequent social activity being larger in magnitude than the reverse association. Additionally, subjective life expectancy (SLE) showed a significant moderating effect on the association from baseline social activity to subsequent depressive symptoms.Supplementary InformationThe online version contains supplementary material available at 10.1186/s12888-026-08226-3.Keywords: Social activity, Depressive symptoms, Multimorbidity, Subjective life expectancy, Health management Introduction Depression is a common psychiatric disorder affecting an estimated 5% of adults globally [1], representing a leading cause of preventable suffering, a major contributor to worldwide disease burden [2], and a key driver of elevated suicide risk, premature mortality from physical illness, and impaired social and economic functioning [3]. For older adults—who face a uniquely high risk of depression closely tied to co-occurring chronic conditions [4]—depression is projected to become one of the leading causes of premature death. Against this backdrop, multimorbidity in later life is associated with compounded health impairments that exceed the additive effects of individual chronic diseases [5], including poorer quality of life, higher disability rates, and elevated mortality risk relative to peers with single chronic conditions [6]. Mounting evidence confirms that older adults with multimorbidity face disproportionately heightened vulnerability to depression [7–11], a challenge that remains critically under-addressed in China. Despite rising depression prevalence among Chinese older adults, widespread social stigma, disease-related shame, and limited health literacy drive extremely low consultation and treatment rates [6]. As multimorbidity prevalence continues to rise in this population, identifying safe, accessible, and effective strategies to prevent and alleviate depressive symptoms in this high-risk subgroup has become an urgent public health priority [1, 12]. Health-related behaviors are core social determinants of health [13], with evidence indicating that healthy lifestyles can mitigate chronic disease impacts and reduce mental disorder incidence [14, 15]. Social activity, a key modifiable health behavior, is closely tied to older adults’ mental health [16], with its protective effects grounded in two foundational theoretical frameworks. First, the Activity Theory of Aging posits that maintaining consistent social engagement in later life preserves role identity, reduces social disengagement, and buffers against depressive symptoms—an effect that may be especially critical for multimorbidity older adults, who face elevated risk of role loss and social isolation due to physical health declines. Second, the Active Aging Framework identifies social participation as a core pillar of healthy aging, providing a structural rationale for examining social activity as a scalable, low-barrier intervention target for vulnerable older subgroups. Consistent with these frameworks, multiple studies have documented a negative association between social activity and depressive symptoms: research using Chinese samples finds that social participation is linked to lower depression risk, with heterogeneity by gender and residence [17]; international studies similarly show that engagement in social, volunteer, or charitable activities correlates with reduced depressive symptoms, with effects moderated by emotional social support [18]; and higher frequency of social engagement is tied to lower depression incidence among middle-aged and older adults [19]. However, most existing longitudinal studies have focused on the general older adult population, and it remains unclear whether these mental health benefits extend to older adults with multimorbidity, who face unique, compounded physical and psychological stressors. Subjective life expectancy (SLE) is a malleable psychological construct within the biopsychosocial model of health [20], closely linked to depressive symptoms, self-rated health, and mortality risk [21]. The moderating role of SLE is directly grounded in socioemotional selectivity theory (SST) [22], which posits that an individual’s subjective perception of remaining time in life shapes their motivational priorities, resource allocation to social engagement, and emotional regulation strategies. Specifically, SST predicts that older adults with higher SLE will prioritize future-oriented, expansive social engagement, which in turn amplifies the mental health benefits of social activity; conversely, those with lower SLE will prioritize present-oriented emotional goals, reducing their motivation to engage in social activities and weakening the association between social activity and depressive symptoms. Prior research supports this framework, confirming that SLE moderates the association between psychosocial stressors (e.g., loneliness) and depressive symptoms [23], and that social activity can positively shape SLE [24]. For multimorbidity older adults, who often face constrained subjective time horizons due to chronic illness, SLE may act as a critical boundary condition that alters the strength of the longitudinal association between social activity and mental health. Integrating SLE into our analytic model thus clarifies for whom the social activity-depression association is strongest, directly informing the development of targeted, low-cost, scalable mental health interventions for this vulnerable population. Despite the above evidence, several gaps remain regarding older adults with multimorbidity. First, the bidirectional temporal association between social activity and depressive symptoms has been understudied in this high‑risk group. Second, few studies have explored modifiable protective factors. To address these gaps, we propose two hypotheses. Hypothesis 1 There is a significant reciprocal negative longitudinal association between social activity and depressive symptoms among Chinese older adults with multimorbidity; specifically, the negative association of baseline depressive symptoms with subsequent social activity will be stronger in magnitude than the reverse association. Hypothesis 2 SLE significantly moderates this bidirectional longitudinal association; specifically, the reciprocal negative association between social activity and depressive symptoms will be more pronounced among older adults with higher SLE. To test these hypotheses, this study aims to address two core research questions: (1) What is the direction and magnitude of the bidirectional temporal association between social activity and depressive symptoms among Chinese older adults with multimorbidity? (2) Does subjective life expectancy moderate this longitudinal association? We use two-wave balanced panel data from a nationally representative survey, and apply cross-lagged panel models and grouped regression models to systematically examine these questions. Theoretical framework and conceptual definition Activity Theory and the Active Aging Framework The core view of the Activity Theory holds that older adults’ active participation in social activities is a key path to cope with negative emotions caused by role loss (e.g., retirement, widowhood), maintain physical and mental health, and improve life satisfaction [25]. The theory advocates that older adults should integrate into society with a positive attitude. They can maintain the vitality of middle age and seek new social roles to alleviate negative emotions from role interruption or loss, and achieve social reintegration. From the reciprocal logic of Activity Theory, better mental health may in turn encourage sustained social engagement, laying the theoretical foundation for examining a bidirectional longitudinal association rather than a unidirectional relationship. Active aging is a consensus concept for the international community to address aging. Its policy exploration and practice originated in Western developed countries in the 1990s, and gradually formed a policy service system with three pillars: “Health”, “Participation” and “Security“ [26]. This framework emphasizes that old age is a process of continuously improving the quality of life by optimizing opportunities for health, participation and security [27]. Among them, “Participation” is the core element that distinguishes it from other aging theories, with the ultimate goal of improving older adults’ quality of life. Increasing active participation in life has a significant effect on alleviating depressive symptoms in older adults. “Participation” includes multidimensional meaningful roles and activities such as social participation, cognitive participation and productive participation. Such activities can promote behavioral activation, goal perception, social support and self-efficacy [28]. Older adults who actively participate in life and maintain social connections are more likely to preserve physical and mental health. In contrast, those who are socially isolated and lack activities face a higher risk of declining health and well-being [29]. This is particularly prominent among older adults with multimorbidity who have higher health vulnerability. Conversely, depressive symptoms may reduce an individual’s motivation and capacity to engage in social activities, reflecting the behavioral withdrawal commonly associated with depression. Together with the reciprocal logic of Activity Theory outlined above, this supports the theoretical plausibility of a bidirectional longitudinal association. This study constructs a theoretical model based on the Activity Theory and the Active Aging Framework (see Fig. 1). As depicted in Fig. 1, social activity (an important form of social participation) corresponds to the “Participation” pillar, depressive symptoms (the mental health indicator in this study) correspond to the “Health” pillar, and subjective life expectancy (SLE) corresponds to the “Security” pillar. The two solid black bidirectional arrows between social activity and depressive symptoms represent the hypothesized reciprocal longitudinal associations (baseline social activity with follow-up depressive symptoms, and baseline depressive symptoms with follow-up social activity), while the grey dashed line captures the cross-sectional correlation at each wave. The solid white vertical arrow from SLE to these bidirectional pathways denotes the hypothesized moderating effect of SLE; the two grey dashed lines linking SLE to both social activity and depressive symptoms further illustrate that SLE, as the core of the “Security” pillar, functions as a foundational psychological factor shaping both motivation for social engagement and mental health perceptions. Theoretical framework diagram Subjective life expectancy (SLE) is conceptualized as a dynamic cognitive appraisal of one’s remaining lifespan, reflecting an individual’s “mental model” of their future longevity constructed from personal experiences, health contexts, and psychosocial factors [20, 30]. SLE is anchored in the “Security” pillar of the Active Aging Framework for the following reasons. The “Security” pillar centers on providing a stable psychological foundation for healthy aging by supporting older adults’ ability to perceive and cope with future health and survival risks [26]. SLE, as a personalized time frame that quantifies perceived remaining lifespan, directly shapes individuals’ perceptions of future health risks, survival security, and long-term planning—thereby constituting the core cognitive basis for constructing psychological security in later life [30, 31]. Moreover, SLE is a malleable psychological construct characterized by dynamic revision in response to health or life events and susceptibility to external intervention [30], enabling it to serve as a modifiable psychological resource that enhances perceived health security. Consistent with this perspective, prior research has demonstrated that a longer SLE buffers the association between psychosocial stressors (e.g., loneliness) and depressive symptoms [23], directly confirming its role as a psychological security resource. In the context of multimorbidity older adults who face persistent health threats, SLE captures their subjective judgment of health security and survival prospects, and is therefore hypothesized to moderate the bidirectional longitudinal association between social activity and depressive symptoms. Specifically, older adults with higher SLE may perceive a longer future horizon, thereby increasing their motivation to engage in social activities and enhancing the mental health benefits derived from such engagement. Conversely, low SLE may weaken this protective association. Conceptual definition Social activity Social activity is defined as a behavioral mode where older adults achieve communication and sharing, meet needs and realize their own values by playing various social roles [32]. Given the health vulnerability of older adults with multimorbidity and the Chinese cultural background, this study adopts this definition and sets social activity as a continuous variable. Social activity, as a core component of the “Participation” pillar in active aging, provides older adults with social support, behavioral activation, and goal orientation, which are theoretically linked to fewer depressive symptoms. From Activity Theory, social engagement helps offset negative emotions from role loss and supports social reintegration. Depressive symptoms Depressive symptoms in this study refer to self‑reported depressive mood. The 10-item Center for Epidemiologic Studies Depression Scale (CESD-10) [33] is used for measurement, and depressive symptoms are set as a continuous variable to reflect gradient differences. Depressive symptoms represent impaired mental health within the “Health” pillar. Severe depressive symptoms may reduce motivation and functional capacity, thereby limiting social participation. This reciprocal dynamic supports the investigation of bidirectional longitudinal associations. Subjective life expectancy Subjective life expectancy (SLE) is an individual’s quantitative perception of their remaining lifespan, characterized by modifiability (plasticity) [30]. It is set as a binary moderating variable, measured using the probability estimation method from CHARLS. SLE acts as a malleable psychological resource under the “Security” pillar. It shapes health perceptions and behavioral choices, and is therefore expected to modify the strength of the longitudinal links between social activity and depressive symptoms. Method Study design and population The data of this study come from the China Health and Retirement Longitudinal Study (CHARLS). Peking University implements this project. It is a national microdata survey focusing on families and individuals aged 45 and above in China, and it provides important support for aging research and interdisciplinary analysis. To ensure the national representativeness of the sample, CHARLS uses a multistage sampling method. It selects samples step by step at the county (district), village (community), household, and individual levels. Specifically, it adopts probability-proportional-to-size (PPS) sampling for counties and villages. The survey covers 150 counties and 450 communities (villages) across 28 provinces (autonomous regions and municipalities directly under the Central Government) in China. The national baseline survey was officially launched in 2011, and it conducted multiple follow-up surveys in 2013, 2015, 2018, 2020, and 2021–2023. CHARLS provides a fully searchable data dictionary to query detailed information of all available data. Its official data query platform is http://charls.pku.edu.cn/. As of April 2026, the latest publicly available survey data wave of this project is from 2020. We used two cross-sectional waves of CHARLS, namely the 2015 and 2018 waves. We mainly chose these two years based on the following considerations. First, the 2020 wave coincided with the global COVID-19 pandemic and the associated public health containment measures, which substantially disrupted social activities and may have exerted transient but pronounced effects on mental health among older adults. Including 2020 data could introduce bias attributable to these extraordinary social circumstances, potentially confounding the longitudinal associations of interest. Therefore, the 2020 wave was excluded to maintain the interpretability of the findings within a more typical social environment. Second, this study adopts a longitudinal design requiring a two-wave balanced panel. The adjacent and relatively stable 2015 and 2018 waves provide sufficient sample size after matching and are less likely to be affected by major external shocks, making them a more suitable choice for examining the bidirectional associations between social activity and depressive symptoms over time. The study population was defined as older adults who (1) participated in both the 2015 and 2018 waves, (2) were aged 60 years or older at baseline (2015), and (3) reported two or more chronic diseases at baseline. Given the irreversible nature of most chronic conditions in later life, we followed the approach described by Zhang et al. [34] to handle missing chronic disease information at baseline. Specifically, when data on a specific chronic condition were missing in 2015 but the corresponding 2018 data indicated absence of that condition, the 2015 status was imputed as “no disease” using the 2018 information. This imputation was applied conservatively only to confirm the absence of a condition; it did not impute the presence of new chronic diseases retrospectively. This procedure helps reduce potential bias from item non‑response while preserving the validity of the multimorbidity classification. The sample selection process is illustrated in Fig. 2 and described in detail below. Among the 9,977 participants who completed both the 2015 and 2018 surveys, 4,780 individuals younger than 60 years were excluded, leaving 5,197 age‑eligible participants. Of these, 2,764 individuals with fewer than two chronic diseases were further excluded, resulting in 2,433 older adults with multimorbidity. After excluding 998 cases with missing values on key analytic variables, the final analytical sample comprised 1,435 participants (815 men and 620 women), with a mean age of 67.71 years (SD = 5.65). Little’s test of missing completely at random (MCAR) was applied to evaluate the pattern of missing data; results indicated that missingness was consistent with the MCAR assumption for both waves (2015: χ2(1) = 0.770, P = 0.380; 2018: χ2(1) = 0.700, P = 0.403). A summary of missing data proportions across all variables is provided in Supplementary Table 1. Baseline characteristics of participants retained versus lost to follow-up are compared in Supplementary Tables 2, and comparisons between respondents and non-respondents on key items are shown in Supplementary Table 3 using Chi-square tests and Mann-Whitney U tests. Given that the MCAR assumption was not rejected, listwise deletion is an acceptable approach under this condition and is not expected to introduce systematic bias in parameter estimation [35]. Accordingly, complete-case analysis was employed, yielding a final sample of 1,435 older adults with multimorbidity. Flow chart of sample selection Measures and measurements Independent variable Social activity was the independent variable. The questions were as follows: DA056 Have you performed any of these activities in the past month? (Check all that apply; multiple choices are allowed) and DA057 How frequently did you perform these activities in the past month? (From the CHARLS questionnaires administered in 2015 and 2018). DA056 includes (1) Interacted with friends; (2) Played Mahjong; played chess, played cards, or went to community club; (3) Provided help to family, friends, or neighbors who do not live with you; (4) Went to a sport, social, or other kind of club; (5) Took part in a community-related organization; (6) Done voluntary or charity work; (7) Cared for a sick or disabled adult who does not live with you; (8) Attended an educational or training course; (9) Stock investment; (10) Used the Internet; (11) Other activities. The specific categories are detailed in Supplementary Table 4. DA057 includes the following three options: almost daily (assigned as 3), almost weekly (assigned as 2), and infrequently (assigned as 1). Although the listed activities are diverse in type, the core interest of this study lies in the overall level of social engagement rather than the independent effects of specific activities. Accordingly, these items were combined into a single continuous index that integrates both the number of activity types and the frequency of participation, providing a more comprehensive and sensitive measure of overall social activity than simple counts or binary indicators. We constructed the social activity index C for older adults by referring to the practices of Yang and Wang to measure the social activity of older adults with multimorbidity [32], calculated as follows: In the formula, C represents social activity, and represents the 11 social activities, assigned a score of 1 or 0 depending on the activity’s presence or absence, and represents the social activities’ frequency. Social activity is the sum of the product of each social activity and its frequency, with a total score of 0–33. Dependent variable The dependent variable was depressive symptoms, measured using 10 items of the Center for Epidemiological Studies Depression Scale-10 (CESD-10) profile in the CHARLS questionnaire, including the following items: (1) I was affected by things that do not usually affect me. (2) I had trouble staying focused on what I was doing. (3) I felt depressed. (4) I felt that everything I did was an effort. (5) I felt hopeful about the future. (6) I felt afraid. (7) My sleep was restless. (8) I was happy. (9) I felt lonely. (10) I could not be active. Questions 5 and 8 are reverse-scored. The total CESD-10 score ranges from 0 to 30 points. The CESD-10 exhibits high reliability and validity for older adults in China [36, 37] and has previously been demonstrated to exhibit good reliability and validity for the general population [38, 39]. Cronbach’s alpha coefficients for depressive symptoms scores were 0.805 and 0.804 in 2015 and 2018, respectively, indicating good internal consistency of the scale. Moderating variable SLE was measured by asking respondents the DA081 question regarding their SLE in the CHARLS 2015 questionnaire. In other words, people of different ages were asked regarding their likelihood of living until a certain age. Respondents aged younger than 65, 65–69, 70–74, 75–79, 80–84, 85–89, and 90–94, respectively, were asked regarding their perceived possibility of living until 75, 80, 85, 90, 95, 100, and 105 years. Responses were recorded on a 5-point ordinal scale, with the options: Almost impossible, Not very likely, Maybe, Very likely, Almost certain. Following the approach of previous studies [40], we divided SLE into two categories. While the first three answers represented low SLE, the last two answers represented high SLE. This binary categorization was adopted because it directly aligns with the core theoretical distinction of socioemotional selectivity theory between “limited future time horizon” and “open-ended future time horizon”, which is the theoretical foundation for SLE’s moderating role. It also simplifies the interpretation of moderating effects by clearly comparing the association patterns between two distinct groups, and this operationalization ensures consistency with existing literature and enhances the comparability of our findings. Control variables Comprehensively referring to previous relevant studies, we included four core control variables based on their well-documented independent associations with both social activity and depressive symptoms: Marital Status, which shapes social support availability and is a known protective factor against depression in later life; Disability, which directly limits physical capacity for social engagement and increases vulnerability to depressive symptoms; Life satisfaction, a global indicator of psychological well-being that correlates with both social participation motivation and mental health status; and Self-reported health status, which reflects overall physical health and may confound the bidirectional association between social activity and depression. Moderating variables were included only for the 2015 status. We acknowledge that additional variables such as socioeconomic status and social support may also influence the associations under investigation. However, these four variables represent the most consistently identified confounders in the existing literature on social activity and depression among older adults, and their inclusion helps minimize potential confounding bias in our cross-lagged and grouped regression analyses. Table 1 presents the specific assignments of the variables using descriptive statistics. Table 1Descriptive statistics of variablesVariablesVariable Assignment2015 (%)2018 (%)MaleFemaleMaleFemale Dependent variable Depressive symptoms(Mean ± S.D.)continuous variable16.47(6.48)18.92(7.30)16.76(6.51)19.31(7.14) Independent variable Social Activity(Median)continuous variable1.001.000.001.00 Moderating variables SLE(%)1 = low78.0480.16——2 = high21.9619.84—— Control variable Marital Status(%)1 = Spouse86.2661.4583.4456.132 = Spouseless13.7438.5516.5643.87Disability (%)0 = no53.1354.0345.4045.161 = yes46.8745.9754.6054.84Life Satisfaction (%)1 = Completely Satisfied6.874.683.564.842 = Very Satisfied30.4335.6527.9831.453 = Somewhat Satisfied55.2147.9059.7548.394 = Not Very Satisfied6.387.426.3810.325 = Not at All Satisfied1.114.352.335.00Self-Reported Health Status (%)1 = Very Good5.655.006.015.652 = Good9.697.107.986.453 = Fair50.9250.8146.2644.844 = Poor26.0127.4230.0632.905 = Very Poor7.739.679.6910.16 Descriptive statistics of variables Depressive symptoms (Mean ± S.D.) 16.47 (6.48) 18.92 (7.30) 16.76 (6.51) 19.31 (7.14) Social Activity (Median) Statistical analysis First, Harman’s single-factor test was used to test for common method deviation. Notably, Harman’s single-factor test serves only as a preliminary screening tool for common method bias and has inherent limitations, as it cannot fully rule out all potential sources of method variance. The variables’ means, medians, or percentages were calculated for descriptive statistics. The correlation between social activity and depressive symptoms was explored using Spearman’s correlation coefficient. All descriptive statistical analyses were conducted using SPSS 27.0 software. Second, a Structural Equation Model (SEM) was developed using Mplus 8.3 statistical software to conduct a cross-lag analysis of social activity and depressive symptoms measured at stages T1 and T2. All variables were retained in their original metric for model estimation, and standardized coefficients are reported throughout the results to facilitate comparison of effect magnitudes. The model included correlated residuals between social activity and depressive symptoms at each time point, a standard assumption in cross-lagged panel designs to account for contemporaneous associations between the constructs. The model was estimated using maximum likelihood (ML) estimation with 1,000 bootstrap resamples to obtain robust standard errors and bias-corrected confidence intervals. Control variables—including Marital Status, Disability, Life Satisfaction, and Self-Reported Health Status—were included in the model by specifying direct paths to the endogenous variables at both time points, consistent with the cross-lagged panel design. We acknowledge that cross-lagged panel models cannot fully separate within-person changes over time from stable between-individual differences, and thus the results should be interpreted as population-level associations rather than individual-level causal effects. Finally, a multigroup SEM analysis was conducted to examine the moderating role of subjective life expectancy (SLE). Baseline SLE was dichotomized into low and high groups as the grouping variable. In the low SLE group, the cross‑lagged paths from T1 social activity to T2 depressive symptoms (S1→C2) and from T1 depressive symptoms to T2 social activity (C1→S2) were labeled A and a, respectively; in the high SLE group, the corresponding paths were labeled B and b. Group differences were tested by defining difference parameters (DIFF1 = A-B; DIFF2 = a-b), with significance evaluated using bias-corrected bootstrap confidence intervals (1,000 resamples). Standardized coefficients are reported throughout the results to facilitate comparison of effect magnitudes. We considered a P value < 0.05 as statistically significant. Results Harman’s single-factor test The results of Harman’s single-factor test revealed that in 2015 and 2018, the number of factors whose eigenvalues were greater than 1 was two, and the variance explained by the first factor was 38.881% and 38.945%, respectively, which was less than the critical standard of 40% [41]. These results suggest that common method bias is unlikely to be a serious concern in this study, although it should be noted that Harman’s single-factor test is a diagnostic tool rather than a definitive test for the absence of such bias. Social activity’s association with depressive symptoms Using the Kolmogorov-Smirnov test to assess the continuous variables’ normality, we found that neither social activity nor depressive symptoms scores followed a normal distribution. Therefore, Spearman’s test was used to analyze the correlation between the independent and dependent variables. Table 2 presents the correlation tests’ results, which show statistically significant albeit modest negative correlations between social activity and depressive symptoms at both time points, with absolute coefficients ranging from 0.163 to 0.193. Additionally, the correlation coefficients of T1 and T2 social activity, T1 depressive symptoms, and T2 depressive symptoms were greater than 0.300, indicating that social activity and depressive symptoms in older adults with multimorbidity exhibit modest temporal stability over the follow-up period. Table 2Descriptive statistics and correlation analysis of social activity and depressive symptomsMean ± SDS1S2C1C2S11.905 ± 2.3951.000S21.716 ± 2.4070.414**1.000C117.529 ± 6.950-0.187**-0.193**1.000C217.866 ± 6.904-0.163**-0.180**0.539**1.000**Significant at P < 0.010S1 = T1 social activity, S2 = T2 social activity, C1 = T1 depressive symptoms, C2 = T2 depressive symptoms Descriptive statistics and correlation analysis of social activity and depressive symptoms **Significant at P < 0.010 S1 = T1 social activity, S2 = T2 social activity, C1 = T1 depressive symptoms, C2 = T2 depressive symptoms Cross-lag analysis results We constructed structural equation models (SEMs) to examine bidirectional longitudinal associations between social activity and depressive symptoms at T1 and T2. The model demonstrated good fit to the data (Table 3, Model 1), with standardized path coefficients presented in Fig. 3. After adjusting for control variables (paths not shown in the figure), both constructs showed statistically significant autoregressive effects, reflecting their temporal stability among multimorbidity older adults. Cross-lagged analyses revealed statistically significant reciprocal negative associations between the two constructs. Higher baseline social activity was negatively associated with subsequent depressive symptoms (β=-0.054, P < 0.050), and higher baseline depressive symptoms were negatively associated with subsequent social activity (β=-0.112, P < 0.010). Both cross-lagged coefficients were small in magnitude. The absolute standardized coefficient from baseline depressive symptoms to follow-up social activity was larger than that of the reverse direction. While these effects are small, they represent meaningful associations at the population level. Table 3Model fit testEvaluation IndexFit CriteriaModel ResultsAdaptation JudgmentModel 1Model 2CFIa> 0.900.9470.954YesTLIb> 0.900.9070.920YesRMSEAc< 0.080.0530.048Yesχ2/df< 5.004.9732.625YesaCFI=Comparative Fit IndexbTLI=Tucker-Lewis IndexcRMSEA=Root Mean Square Error of Approximation Model fit test aCFI=Comparative Fit Index bTLI=Tucker-Lewis Index cRMSEA=Root Mean Square Error of Approximation Cross-lag analysis of social activity and depressive symptoms. Note: **Significant at P < 0.010, *Significant at P < 0.050 SLE’s moderating effect Multigroup SEM was conducted to examine the moderating role of SLE. The model demonstrated good fit (Table 3, Model 2), with standardized results presented in Table 4; Figs. 4 and 5. Among older adults with low SLE, a significant negative association was observed from baseline depressive symptoms to subsequent social activity (β=-0.105, P < 0.01), but no significant association was found from baseline social activity to subsequent depressive symptoms (β=-0.030, P = 0.176). In contrast, among those with high SLE, significant reciprocal negative associations were present between the two constructs (β1=-0.114, P < 0.050; β2=-0.134, P < 0.010). For this subgroup, higher social activity is associated with less severe depressive symptoms, suggesting that prioritizing social activity for older adults with high SLE may represent a promising preliminary direction for targeted intervention. Table 4Test for the moderating effect of SLE using grouped regressionpathEstimateS.E.Est./S.E.Lower 2.5%Upper 2.5% P SLE = 1C1→S2(a)-0.1050.025-4.258-0.150-0.058< 0.001**S1→C2(A)-0.0300.022-1.354-0.0780.0140.176SLE = 2C1→S2(b)-0.1140.050-2.291-0.207-0.0100.022*S1→C2(B)-0.1340.040-3.389-0.222-0.0580.001*** Significant at P < 0.050, ** Significant at P < 0.010 Test for the moderating effect of SLE using grouped regression * Significant at P < 0.050, ** Significant at P < 0.010 Analysis of cross-lag models of social activity and depressive symptoms (SLE = 1). Note: **Significant at P < 0.010, *Significant at P < 0.050 Analysis of cross-lag models of social activity and depressive symptoms (SLE = 2). Note: **Significant at P < 0.010, *Significant at P < 0.050 Group differences in the cross-lagged pathways were formally tested (see Table 5). Only the social activity→depressive symptoms pathway differed significantly between the low and high SLE groups (P < 0.050), while the depressive symptoms→social activity pathway showed no significant group difference (P > 0.050). This indicates that the negative association of social activity with subsequent depressive symptoms was observed only among individuals with high SLE. This pattern is consistent with the interpretation that the potential mental health benefits of social activity may be contingent upon individuals’ future time perspective. Table 5Social activity and depressive symptoms pathway coefficient difference testEstimateS.E.Est./S.E.Lower 2.5%Upper 2.5% P Diff(a-b)0.0170.0240.704-0.0300.0590.481Diff(A-B)0.2230.1131.9730.0070.4520.049** Significant at P < 0.050, ** Significant at P < 0.010, Diff = unstandardized difference. Standardized bootstrap CIs are not directly available due to Mplus 8.3 limitations Social activity and depressive symptoms pathway coefficient difference test * Significant at P < 0.050, ** Significant at P < 0.010, Diff = unstandardized difference. Standardized bootstrap CIs are not directly available due to Mplus 8.3 limitations Discussion This is a large-scale representative study to examine the longitudinal associations between social activity and depressive symptoms in older adults with multimorbidity. Using national representative data from the CHARLS, we corroborate some associations frequently reported in the literature and offer preliminary novel insights. We found that T1 social activity, T1 depressive symptoms, T2 social activity, and T2 depressive symptoms all exhibit pairwise connections, and that social activity and depressive symptoms are negatively correlated. In line with our hypothesis, we found an asymmetric bidirectional association between social activity and depressive symptoms in older adults with multimorbidity. Additionally, SLE appeared to moderate this association. These findings are discussed below. It is important to note that, due to the observational nature of the study design, these findings represent associations rather than causal relationships, and alternative explanations such as residual confounding or unmeasured variables cannot be entirely ruled out. We found a bidirectional relationship between social activity and depressive symptoms in older adults with multimorbidity. First, T1 social activity was significantly and negatively associated with T2 depressive symptoms, which extends the existing literature [42]. From the perspective of the Activity Theory, social activity may be associated with helping older adults with multimorbidity to cope with the negative emotions brought by role loss (e.g., retirement and illness-related functional decline) and to re-establish social identity by undertaking diverse social roles [25]. In the cultural context of China, the social activities of older adults are deeply rooted in collective life and family-oriented values—for example, activities like playing mahjong, participating in community square dancing, and providing mutual help to neighbors are not only social interactions but also reflect the traditional Chinese concept of “interdependence among community members” [32]. Social activity may be linked to older adults taking on various social roles. Participation in social activities is associated with a greater sense of value realization, participation, and satisfaction, as well as lower levels of anxiety and pressure related to retirement and aging [43], and reduced loneliness—factors that may in turn be associated with a lower likelihood of depressive symptoms. Taken together, these potential pathways suggest that social activities may be associated with reduced health vulnerability among older adults with multimorbidity, which could potentially correlate with a lower likelihood of depressive symptoms. Long-term social isolation, particularly the scarcity of formal social activities, may pose a significant threat to the cognitive function of older adults [44]. Furthermore, social isolation or loneliness resulting from the lack of informal social contact (rather than insufficient formal social activities) appears to exert a detrimental effect on cognitive function that is significantly greater than the protective effect derived from alleviating such isolation [45]. This pattern aligns with the present findings: limited social activity is directly associated with elevated depressive symptom risk and may also be indirectly associated with poorer mental health via lower cognitive function. However, we cannot rule out the possibility that pre-existing cognitive impairment may also be associated with both reduced social activity and increased depressive symptoms. The current evidence regarding the relationship between social activity and depressive symptoms is inconsistent. Coyne and colleagues assumed that the interpersonal behaviors of patients with depressive symptoms may be associated with exclusion by others, which could potentially lead to alienation of potential supporters and reduction of the social support network [46, 47]. Further, Mohr et al. investigated how depressive symptoms medication impacts the social support system of patients with multiple sclerosis. They discovered that a reduction in depressive symptoms completely mediated improvements in social support utilization and satisfaction during the treatment phase [48]. This indicates that improvement in depressive symptoms following treatment is associated with increased utilization of—and satisfaction with—support services. Son and Sung examined the relationship between social participation and depressive symptoms in older adult South Koreans. They discovered that formal social engagement exhibited a cross-lagged effect in reducing depressive symptoms, whereas depressive symptoms did not exhibit a cross-lagged effect on either type of social engagement [49]. Our study demonstrates that T1 depressive symptoms were significantly and negatively associated with T2 social activity, which is consistent with the findings of Coyne and colleagues [46, 47]. In the Chinese cultural context, older adults with multimorbidity and depressive symptoms often face greater social and family expectations of “being healthy and contributing to the family”, which may intensify their sense of powerlessness and withdrawal when they experience depressive emotions [32]. This is consistent with the Activity Theory’s view that negative emotions are linked to impaired maintenance of social roles, and further may help explain why depressive symptoms are associated with reduced social activity participation among Chinese older adults with multimorbidity [25]. Depressive symptoms are characterized by low mood, anhedonia, apathy, and reduced autonomy, which are associated with diminished motivation for social and family life, lower social function, and increased social isolation. Collectively, these features suggest that depressive symptoms in this population are associated with lower social activity participation, reflecting a longitudinal associational link. Third, the absolute coefficient of T1 depressive symptoms → T2 social activity (0.112) was larger than that of T1 social activity → T2 depressive symptoms (0.054), indicating an asymmetric bidirectional longitudinal association: the negative association of baseline depressive symptoms with subsequent social activity was stronger than the reverse. This asymmetry suggests that depressive symptoms may be more strongly associated with reduced social engagement in vulnerable multimorbid groups. Although the observed effect sizes are relatively modest and warrant cautious interpretation, the pattern is consistent with the theoretical expectation that depressive symptoms may precipitate social withdrawal in this population. Within China’s community-based elderly care system, organized social activities are key carriers of active aging strategies [6, 32]. The stronger association between baseline depressive symptoms and later social activity reduction highlights the preliminary potential value of early mental health screening for this population. Depressive symptoms may be associated with progressive social withdrawal, which may further relate to increased health vulnerability. However, it is also possible that unmeasured factors such as baseline functional status or disease severity may be associated with both higher depressive symptoms and subsequent social withdrawal. SLE showed a significant moderating association with the link between T1 social activity and T2 depressive symptoms, and high SLE was associated with a stronger negative longitudinal association between social activity and depressive symptoms. This pattern aligns with Bodner and Bergman’s findings [23]. In the Active Aging framework, SLE is positioned as a core component of the “Security” pillar [30]. High SLE reflects positive future life expectations, which may be associated with greater willingness to invest in social activities for better quality of life [24], consistent with the cultural value of “longevity and well-being”.From a life-course perspective, subjective mental health is associated with reducing health inequalities in older adults [50], as social interaction and support are linked to positive emotional states [30]. SLE is a malleable construct [20], and older adults with higher SLE may be more likely to engage in social activities for emotional well-being. Given that SLE and depressive symptoms were measured concurrently at baseline, the temporal order of this moderating pattern cannot be confirmed, and reverse associational pathways or unmeasured confounding cannot be excluded. For example, individuals with better overall physical health may have both higher SLE and stronger associations between social activity and mental health outcomes. This study has several noteworthy advantages. First, we modeled the association between social activity and depressive symptoms using balanced panel data, which allowed for examination of longitudinal associations while accounting for temporal ordering, an improvement over cross-sectional designs that cannot assess directionality. Second, this study included a group of older adults with multimorbidity who were prone to depressive symptoms, thereby substantially increasing academic attention to this population. Finally, we incorporated SLE into the research paradigm, thereby providing preliminary insights into accessible therapies and novel theoretical approaches for understanding depressive symptoms in older adults with multimorbidity. Despite these strengths, the observational design precludes causal inference, and residual confounding from unmeasured variables such as genetic factors, childhood adversity, or detailed medication history remains a possibility. Additionally, reverse causality cannot be ruled out for any of the observed associations. Practically speaking, this study offers preliminary insights with potential clinical relevance for caring for older adults with multimorbidity and depressive symptoms. Depressive symptoms are a mental disorder causing inexplicable suffering, which is difficult to describe. Its extreme manifestations are nearly impossible to understand without experiencing them firsthand. Compared with older adults with a single chronic disease, those with multimorbidity are more likely to experience mental health issues, including anxiety and depressive symptoms [51]. As the number and severity of their chronic illnesses increase, so does the severity of their mental health problems [52]. Our study’s findings provide preliminary perspectives and ideas for treating depressive symptoms in older adults with multimorbidity as well as an important reference value for older adults with varied characteristics. On the one hand, given the observed association between social activity and fewer depressive symptoms, it may be worthwhile to preliminarily explore encouraging participation in social activities—particularly those combining social interaction and physical exercise (e.g., square dancing, tai chi, qigong)—among older adults with multimorbidity and depressive tendencies who have adequate physical health. These activities may be particularly suitable for older adults with multimorbidity in China, as they align with the Activity Theory’s call for active social participation and adapt to their physical condition, and are easy to promote in community-based elderly care services [25, 32]. Such approaches may help older adults with multimorbidity to expand their social networks while exercising, thereby potentially contributing to reduced risk of mental illness and, simultaneously, possibly enhancing their physical condition. Notably, professional health management personnel should be trained to provide advice on physical exercise to older adults to prevent the aggravation of chronic diseases owing to inappropriate exercise techniques. On the other hand, considering the observed moderating role of SLE in the association between social activities and depressive symptoms, interventions aimed at improving the SLE of older adults with multimorbidity may be worth exploring as a complementary strategy. Based on the modifiability of SLE [30], we could potentially leverage China’s community health education system to disseminate knowledge about chronic disease management and mental health, which not only reduces the stigma of mental illness among older adults but also aligns with the “Security” connotation in the Active Aging framework, potentially helping older adults establish a more positive life expectancy perception [26, 27]. Leveraging digital media to disseminate chronic disease prevention and mental health knowledge, and organizing mental health education sessions, may help older adults with multimorbidity better understand health-related concepts, reduce stigma associated with mental health issues, and potentially foster more positive perceptions of life—factors that are linked to higher SLE scores in this population. It is important to emphasize that these practical implications are preliminary suggestions derived from the observed associations and should be tested in future intervention studies to establish effectiveness. Finally, noteworthily, this study’s findings do not reject conventional medical treatment but, instead, provide a preliminary foundation for exploring complementary approaches to improve the psychological well-being of older adults with both depressive disorders and multimorbidity. Conclusion This observational study explored the bidirectional longitudinal association between social activity and depressive symptoms among Chinese older adults with multimorbidity, as well as the potential moderating role of subjective life expectancy (SLE) in this link. The key findings revealed an asymmetric bidirectional longitudinal correlation between social activity and depressive symptoms: higher T1 social activity was associated with lower T2 depressive symptoms, and higher T1 depressive symptoms were also associated with lower T2 social activity, with the latter association showing a larger longitudinal association. Additionally, SLE was observed to moderate the longitudinal relationship between social activity and depressive symptoms; specifically, among participants with high SLE, T1 social activity demonstrated a significantly stronger negative longitudinal association with T2 depressive symptoms, whereas this association was not statistically significant among those with low SLE. Limitations This study has several limitations. First, the social activity index constructed from CHARLS data encompasses a heterogeneous set of activities, including recreational pursuits (e.g., playing Mahjong, attending clubs), digital engagement (e.g., using the Internet), and financial behaviors (e.g., stock investment). These activities may reflect distinct underlying dimensions of social participation rather than a unidimensional construct, and this heterogeneity should be considered when interpreting the observed associations. Second, this study employed only two waves of longitudinal data (2015 and 2018) to examine the bidirectional associations between social activity and depressive symptoms. The use of two time points limits the ability to model more complex temporal dynamics, such as lagged effects extending beyond a single wave interval or nonlinear trajectories over longer periods. Future research incorporating three or more waves would permit a more nuanced examination of the temporal patterning of these associations. Third, a first-order cross-lag model was used to assess the temporal association between social activity and depressive symptoms, and as an observational study, it cannot establish causal relationships between these variables. Additional testing is required to ascertain whether the two continue exhibiting a bidirectional relationship over a longer period. Fourth, owing to the constraints of the data source questionnaire, the SLE measurement was a probability estimate. Although probabilistic estimates are preferred over point estimates [30], participants’ instant SLE assessments are susceptible to subjective and objective influences, ultimately precipitating assessment bias. Furthermore, this study adopted a dichotomous treatment for SLE by collapsing the original five-point ordinal scale into low versus high categories. While this approach facilitates interpretation, dichotomization inevitably results in loss of information and statistical power, may obscure potential non-linear relationships, and relies on an arbitrary cut-point not subjected to sensitivity analysis. Consequently, the observed moderating effect of SLE should be interpreted with caution. Fifth, this study relied on self-reported data, which may lead to the occurrence of subjective bias; in addition, some unmeasured confounding factors may still exert certain impacts on the research results. Additionally, although Little’s test confirmed that the missing data followed a missing completely at random (MCAR) pattern, the study employed listwise deletion to handle missing values. Despite MCAR conditions theoretically yielding unbiased parameter estimates [35], this approach reduces sample size and may affect statistical power. Furthermore, sample attrition between waves is an inherent concern in longitudinal research. Comparisons between retained and lost participants (Supplementary Table 2) as well as between respondents and non-respondents on key items (Supplementary Table 3) were conducted to assess potential attrition bias; however, the possibility of non-random attrition on unmeasured characteristics cannot be fully ruled out. Additionally, the imputation of missing baseline chronic disease status using subsequent wave data, while theoretically grounded [34], may introduce concerns regarding temporal validity. Future research could further optimize the analysis: on the one hand, multiple imputation could be used to handle missing values to expand the sample size and improve the stability of results; on the other hand, a comprehensive judgment can be made by combining multiple effect size indicators such as the standardized mean difference (SMD), so as to provide more abundant evidence for bias assessment. We will address these problems in future studies to enhance the robustness and generalizability of the research findings. Electronic Supplementary Material Below is the link to the electronic supplementary material. Supplementary Material 1 Acknowledgements We are very grateful to all the members who participated in this study, to the older adults for their help in recruiting others, and to the investigators for their help in collecting data. Author contributions CZ conceived the idea and conceptualized the study questions. XL, LS and SX participated in data cleaning and statistical analysis. XL and LS wrote the manuscript, and all supplementary analyses and final manuscript preparation were completed independently by XL. BX, LY, XZ (Xiao Zheng), XZ (Xinyi Zhang), MZ, YL and CZ edited the paper. CZ gave many valuable comments on the draft and polished it. All authors approved the final version. Funding This work was supported by the National Natural Science Foundation of China under Grant [grant numbers 72274091], the General Program of the Planning Project of Guangdong Philosophy and Social Science [grant number GD23CGL06], the Basic and Applied Basic Research Foundation of Guangdong Province under Grant [grant number 2022A1515011591], Key Laboratory of Philosophy and Social Sciences of Colleges and Universities in Guangdong Province (2023WSYS005; 2015WSY0010). Data availability The datasets generated during and analyzed during the current study are not publicly available due to other related undergoing study but are available from the corresponding author on reasonable request. Declarations Ethics approval and consent to participate Ethical approval for all the CHARLS waves was granted by the Institutional Review Board at Peking University. This study was conducted in accordance with the Declaration of Helsinki. The IRB approval number for the main household survey, including anthropometrics, is IRB00001052-11015. Consent for publication During the fieldwork, each respondent who agreed to participate in the survey was asked to sign two informed consent forms. One copy was kept by the respondent and the other was kept in the CHARLS office and scanned in PDF format. Competing interests The authors declare no competing interests. Footnotes Publisher’s note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations. References Associated Data Supplementary Materials Supplementary Material 1 Data Availability Statement The datasets generated during and analyzed during the current study are not publicly available due to other related undergoing study but are available from the corresponding author on reasonable request.