This study utilized data from CHARLS (11) waves conducted in 2011, 2013, 2015, and 2018. CHARLS is a longitudinal survey that is nationally representative and gathers extensive data on the lifestyle, health, and socioeconomic situation of middle-aged and older Chinese citizens. Its comprehensive design and longitudinal nature, combined with its rich dataset, make it particularly suitable for studying aging-related changes and the complex interplay between social, health, and economic factors. Although 2020 is the latest wave of CHARLS data, the COVID-19 pandemic led to significant data missingness in this wave. Therefore, we chose to use the previous four waves for this study. The Peking University Biomedical Ethics Review Committee authorized the study protocol (IRB00001052-11015), and all subjects gave their informed permission.

The 7,019 participants had a mean age of 57.71 ± 8.39 years, with 55.27% of them being female. The baseline covariate distribution, dependent variable distribution, and mediating variable distribution (Table 1).

The sensitivity analysis revealed significant differences between participants included in the study (n = 6,802) and those excluded (n = 5,701), providing crucial insights into the generalizability of our findings (Table 2). The included group demonstrated significantly higher participation rates across most social activity types, particularly in activities social01 through social06. This may indicate that the older population in our study sample is more socially active than the general population, which might cause the influence of social activities to be overvalued.

Demographically, the included group was slightly younger (57.71 ± 8.39 vs. 57.93 ± 9.16 years), and had a slightly higher proportion of females (55.27% vs. 44.73%). Interestingly, education levels did not differ significantly between the groups (p = 0.33). These disparities indicate that our findings may be more applicable to slightly older and more female older adult populations. Notably, the included group had significantly higher rates of diabetes, possibly reflecting a greater tendency among those with health issues to participate in surveys or social activities, or a heightened awareness of their health status.

These differences unveil potential selection bias, indicating that study participants may be healthier, more active, and cognitively better functioning. This “healthy participant effect” might lead to an overestimation of the protective effect of social activities on cognitive function. Furthermore, the significant differences between included and excluded groups may limit the generalizability of our results, potentially making them more applicable to healthier, more socially engaged older adult populations.

These findings resonate with our main analysis results. For instance, the positive correlation we found between total social activity scores and cognitive function is consistent with the higher social engagement and better cognitive function observed in the included group. Indeed, the included group showed slightly higher MMSE scores (14.86 ± 6.47 vs. 13.99 ± 6.47, p < 0.0001).

It’s worth noting that the included group showed slightly lower depression scores as measured by CES-D 10 (8.74 ± 6.48 vs. 8.88 ± 6.67, p = 0.29), although this difference is not statistically significant and relatively small.

Given these limitations, caution is warranted in interpreting our results. While our study provides valuable insights into the relationships between social activities, depression, and cognitive function, these findings may primarily apply to older adult populations with higher social engagement and better health status. Future research should focus on older adult groups with lower social participation and poorer health conditions to gain a more comprehensive understanding.

Methodologically, these findings offer important implications. In our main analysis, we need to adequately control for potential confounding factors such as baseline cognitive function and chronic health conditions. Additionally, considering weighted analyses or propensity score matching might help reduce the impact of selection bias and enhance the reliability and generalizability of our results.

In conclusion, this sensitivity analysis not only reveals important sample differences but also provides valuable context for understanding and interpreting our main research findings. It underscores the complexity of conducting large-scale longitudinal studies in older adult populations and highlights the numerous factors that need to be considered when interpreting and generalizing research results in this field.

The comprehensive analysis of social activity frequency and cognitive function across diverse subgroups of older Chinese adults consistently revealed protective effects (Table 3), reinforcing our study’s main findings. Higher social activity participation was associated with lower cognitive impairment risk across various demographic and health-related categories, including education levels, genders, residence types, and chronic condition statuses.

Notably, the protective effects were observed across all subgroups, with only slight variations. Among education levels, all categories showed similar protective effects, with odds ratios ranging from 0.933 to 0.967. Males (OR: 0.960, 95% CI: 0.947–0.974) showed a marginally stronger protective effect compared to females (OR: 0.945, 95% CI: 0.932–0.958). Interestingly, smokers (OR: 0.967, 95% CI: 0.949–0.984) and drinkers (OR: 0.970, 95% CI: 0.954–0.987) demonstrated slightly stronger protective effects compared to non-smokers and non-drinkers, respectively. Rural residents (OR: 0.967, 95% CI: 0.953–0.981) exhibited a marginally stronger effect compared to urban residents (OR: 0.955, 95% CI: 0.942–0.968).

The persistence of these effects, regardless of chronic condition status, highlights the potential of social activities as a universal strategy for cognitive health maintenance in the older adult. This is evidenced by the similar odds ratios for those with and without hypertension, diabetes (DM_test), and dyslipidemia.

These findings align with our primary analysis, demonstrating a significant inverse relationship between social activity participation and cognitive decline risk. They provide a nuanced understanding of how this relationship manifests across different population segments, suggesting the robustness of social activity benefits across various demographic and health profiles.

Importantly, the p for trend values indicate that the protective effect of social activities does not significantly differ across most subgroups (education: p = 0.821, residence: p = 0.363, dyslipidemia: p = 0.405, DM_test: p = 0.877, hypertension: p = 0.99). However, there’s a marginally significant trend for gender (p = 0.053) and a significant trend for drinking status (p = 0.004), suggesting that the effect of social activities might vary more substantially across these categories. While these results offer robust support for our conclusions, it’s important to consider limitations such as varying sample sizes across subgroups. Nevertheless, in the context of China’s rapidly aging population, these findings have significant implications for public health strategies, suggesting that promoting social activities could be a widely applicable approach to maintaining cognitive health in diverse older adult populations. The consistency of protective effects across various subgroups underscores the potential of social engagement as a universal intervention for cognitive health preservation in older adults.

A significant negative linear association between social activity and cognitive impairment risk was consistently demonstrated by the results (Figure 2). This relationship remained robust across all models (P-overall <0.001), with cognitive impairment probability decreasing from approximately 0.8 to 0.4 as social activity scores increased from 0 to 40. The relationship showed no significant non-linear pattern (P-non-linear >0.05 in all models, ranging from 0.1291 to 0.2085), indicating a steady decline in cognitive impairment risk with increasing social activity participation.

These findings suggest that increasing social activity participation could be an effective strategy for cognitive impairment prevention, with consistent benefits across different levels of social engagement. The widening confidence intervals at higher activity scores (>30) indicate a need for cautious interpretation of effects at these extremes, possibly due to fewer participants with very high social activity levels.

According to the results (Figure 3), depression scores showed a positive association with cognitive impairment risk across all models, indicating a significant link between increased depression and higher cognitive impairment probability. This relationship was significant across all models (P-overall <0.001).

The graphs displayed non-linear features (P-non-linear <0.001), with a steeper increase in cognitive impairment risk at lower depression scores (0–10 points) and a more gradual increase at higher scores (10–30 points). As depression scores increased from 0 to 30, the probability of cognitive impairment increased from approximately 0.60 to 0.90, with the relationship remaining consistent across all adjusted models.

According to the results (Figure 4), depression scores consistently decreased as social activity total scores increased across all models, suggesting a negative association between social activity participation and depressive symptoms. This relationship remained robust across all models (P-overall <0.001).

The relationship exhibited non-linear features in the crude model, with a more pronounced decrease in depression scores at higher social activity levels (>20 points). However, after adjusting for covariates (Models 1 and 2), the relationship became more gradual. As social activity scores increased from 0 to 40, depression scores decreased from approximately 0.95 to 0.85, indicating a modest but significant effect of social activity on depressive symptoms.

The confidence intervals remained relatively narrow for social activity scores between 0–30, but widened at higher scores (>30), suggesting less precise estimates at the upper range of social activity participation. This pattern was consistent across all models, indicating the robustness of the relationship despite adjustment for potential confounders.

Analysis reveals distinctive patterns in how different types of social activities influence cognitive function among older adults. Specifically, activities incorporating structured community engagement demonstrated superior protective effects. Community organization participation and training programs showed the strongest associations with cognitive preservation (β = 1.110 and β = 1.020 respectively), significantly outperforming casual social interactions and individual assistance activities. This finding aligns with previous research showing that organized social networks combining structured activities with cognitive engagement were more effective in preventing cognitive decline (29, 30).

The protective effects varied systematically across demographic characteristics. Education level showed relatively stable protective effects (OR: 0.933–0.967), suggesting that the benefits of social engagement transcend educational background (31, 32). Gender analysis revealed slightly stronger protective effects in males compared to females (OR: 0.960 vs. 0.945) (33, 34). Geographic differences were notable, with rural residents demonstrating stronger protective effects compared to urban dwellers (OR: 0.967 vs. 0.955) (35–37), which is different with Feng’s research (32). This discrepancy might be attributed to differences in the types of social activities available and commonly practiced in rural versus urban settings (34, 38).

Our research findings extend previous studies by demonstrating that the level of structure and organization during social activities may be a crucial determinant of cognitive protection. The results suggest that activities combining social interaction with structured community engagement may optimize cognitive benefits (39). This is particularly evident in the strong effects observed for community organization participation and training programs. This perspective is supported by findings from the Survey of Health, Ageing and Retirement in Europe (40), which similarly demonstrated that structured social activities were associated with better cognitive outcomes among older adults across European countries (41).

The study found that depression plays a key mediating role in how social activities influence cognitive function. According to our results, the relationship between social activities and depression showed significant non-linear patterns, with depression scores decreasing as social activity participation increased (16). The relationship was more pronounced at higher social activity levels (>20 points) in the crude model, though this became more gradual after adjusting for covariates. As social activity scores increased from 0 to 40, depression scores decreased from approximately 0.95 to 0.85, indicating a modest but significant effect of social activities on depressive symptoms (42).

The relationship between depression and cognitive impairment demonstrated strong non-linear features (P-non-linear <0.001), with a steeper increase in cognitive impairment risk observed at lower depression scores (0–10 points) and a more gradual increase at higher scores (10–30 points). As depression scores increased from 0 to 30, the probability of cognitive impairment increased substantially from approximately 0.60 to 0.90, with this pattern remaining consistent across all adjusted models.

This complex mediating relationship is particularly evident in the structural equation modeling results, where different types of social activities showed varying degrees of indirect effects through depression. Volunteer work demonstrated the strongest indirect effect through depression (β = 0.213, p < 0.001), followed by training programs (β = 0.180, p < 0.01) and community organization participation (β = 0.140, p < 0.001). This suggests that social activities help reduce depressive symptoms by enhancing social connections and offering emotional support (43), which aligns with previous research findings in Canada (44). The significant mediating effects across multiple types of structured social activities underscore the crucial role of mental health in the pathway between social engagement and cognitive function.

This study analyzed the relationship between participation frequency in nine different types of social activities and cognitive function. Structural equation modeling showed that structured community activities, particularly community organization participation (β = 1.110) and training programs (β = 1.020), had the most significant positive correlations with cognitive function and indirectly influenced cognition by reducing depression levels (indirect effects β = 0.140 and β = 0.180 respectively). These findings align with a longitudinal study conducted in Chicago (45). Volunteer work also demonstrated substantial effects (direct effect β = 0.833, indirect effect β = 0.213) (46–48). In contrast, casual social interactions and caregiving activities showed no significant effects on either cognitive function or depression. These findings support the importance of structured social engagement, suggesting that activities involving organized community participation can more effectively maintain and promote cognitive function in older adults.

Depression consistently played a crucial mediating role in the relationship between most social activity frequencies and cognitive function. The mediating effects were particularly strong for volunteer work (β = 0.213) and training programs (β = 0.180), highlighting the importance of assessing and managing mental health when formulating cognitive function protection plans for community-dwelling older adults (16, 37). The study provides empirical evidence for developing targeted interventions that combine emotional well-being promotion with structured social engagement to comprehensively maintain cognitive function in the older adult population.

Our findings have significant implications for both public health practice and theoretical research. From a practical perspective, the study highlights the role of structured social activities as crucial intervention measures for preventing cognitive decline, aligning with recent Chinese policy initiatives. The State Council’s “Opinions on Strengthening Aging Work in the New Era” emphasizes promoting high-quality social participation among older adults (49), while “The 14th Five-Year Plan” outlines strategies for active aging (50). Our results, particularly regarding the benefits of cognitive-stimulating activities, directly support these policy directions and suggest that community programs should prioritize activities combining social interaction with cognitive engagement. Additionally, the study recommends integrating mental health management into cognitive function protection plans to establish comprehensive intervention models, as demonstrated by the significant mediating effects of depression across multiple activity types.

From a theoretical perspective, the study identifies consistent relationships between social activities and cognitive function and reveals the crucial mediating role of depression in cognitive protection. The findings enhance our understanding of how social engagement influences cognitive health through both direct and indirect pathways, with structured activities showing particularly robust effects. These findings provide theoretical and practical support for fostering healthy aging in an increasingly aging society.