Association between social media use and depressive symptoms in middle-aged and older Chinese adults

Depression has become a leading cause of years lived with disability in China and appears to have been exacerbated during the COVID-19 pandemic. Online convenience samples suggest a high burden, but they under-represent middle-aged and older adults with limited internet access. The study aimed to: (1) estimate the prevalence of depressive symptoms among middle-aged and older Chinese adults in 2020 using a nationally representative survey; and (2) assess longitudinal associations between social media use and the development or reduction of depressive symptoms over two years. The rationale includes potential benefits of social media in fostering social connection, access to services (including mobile payments), and support networks in a largely cashless Chinese society where lack of smartphone access can increase isolation among older adults.

Prior large-scale online surveys reported substantial depressive symptom burden in China during COVID-19 but used convenience sampling that likely excluded many older adults. Literature on social media and mental health indicates both potential benefits (social support, engagement with services) and harms (cyberbullying, unhealthy social comparison), with adverse associations more consistently observed in adolescents and young adults. In China, social media is tightly integrated with mobile payments and online services, making digital inclusion particularly salient for older adults. Evidence among adults ≥30 years (e.g., PSID) suggested social media use could be associated with reduced depressive symptoms, but longitudinal evidence in middle-aged and older Chinese adults was lacking.

Design and data source: Longitudinal analysis using the China Health and Retirement Longitudinal Study (CHARLS), a nationally representative, biannual panel of adults aged ≥45 years in China, employing a four-stage, stratified, cluster probability sampling design. In 2020, 17,364 participants from 10,204 households were surveyed (response rate 84.3%). Ethical approval was obtained (Tulane University IRB #2020-2091), and participants provided informed consent. Participants and analytic cohorts: Social media activities were measured in 2018 and 2020. Two analytic cohorts from 2018 were formed: (1) non-depressed at baseline (n=9950), of whom 9121 (91.67%) were followed in 2020 to assess incident depressive symptoms; (2) depressed at baseline (n=5922), of whom 5302 (89.53%) were followed in 2020 to assess conversion to non-depressed status. Dropouts were older, more rural, lower educated, more often divorced, drank alcohol, had lower physical activity, more comorbidities, and very low social media use. Measures: - Depressive symptoms: Assessed with CESD-10 (10 items; score 0–30). Primary threshold: CESD-10 ≥10 defines depressive symptoms; secondary threshold ≥12 used for comparison. CESD-10 psychometrics validated in CHARLS subsamples. - Social media activities: Four questions captured seven activities: chatting, reading news, watching videos, playing games, mobile payments (e.g., Alipay/WeChat Pay), using WeChat, posting WeChat Moments. A binary composite (any vs none) was created. Methods/devices: desktop, laptop, tablet, cellphone. - Social media transition groups (for baseline-depressed participants): (a) Consistently inactive (no activity in 2018 and 2020); (b) Transition to active (none in 2018, ≥1 in 2020), with thresholds for ≥1, ≥2, and ≥3 activities; (c) Consistently active (≥1 in both years). - Covariates: Demographics and socioeconomic factors (age, sex, education [illiterate, elementary, middle, high school+], living area [urban/suburban/rural], marital status, experience of the Down to the Countryside movement), lifestyle (smoking status, alcohol drinking, physical activity via IPAQ-like instrument producing indices for vigorous/moderate/walking based on duration×days, sleep hours, daytime napping minutes), and number of self-reported physician-diagnosed chronic conditions (hypertension, dyslipidemia, diabetes, cancer, chronic lung disease, liver disease, CVD, stroke, kidney disease, GI diseases, emotional/nervous/psychiatric problems, memory-related diseases, arthritis, asthma). Statistical analysis: - Descriptives compared baseline characteristics by incident depressive symptoms status (for non-depressed cohort) and by conversion to non-depressed status (for depressed cohort) using chi-square, t-test, or Mann–Whitney U where appropriate. - Prevalence estimation for 2020 used survey procedures (SAS PROC SURVEYFREQ) accounting for complex design and non-response; results stratified by 10-year age groups, sex, and living area; province-level prevalence mapped via GIS. - Longitudinal associations: Multivariable logistic regression assessed (i) baseline social media activity measures vs. incident depressive symptoms in 2020 among those non-depressed in 2018, and (ii) social media transition groups vs. conversion to non-depressed status among those depressed in 2018. Models adjusted for all covariates listed above. Beta changes in CESD scores between 2018 and 2020 were also examined for activities/methods. Two-sided p<0.05 considered significant. Analyses conducted in SAS 9.3.

- Prevalence (2020): Overall depressive symptoms 36.0% (95% CI: 34.8–37.1%). Using CESD-10 ≥12 threshold: 27.4% (95% CI: 26.3–28.4%). Prevalence increased with age, peaking at 70–80 years; females consistently higher than males. Women in rural areas had the highest burden: 50.3% (95% CI: 48.6–52.0%). Western provinces showed higher prevalence; Qinghai, Gansu, Chongqing, and Hubei approached or exceeded 50%; Qinghai was 78.2%. - Incident depressive symptoms among those non-depressed at baseline (n=9121): 22.7% developed depressive symptoms over two years. Any social media activity at baseline associated with 24% lower odds of incident depressive symptoms (OR=0.76; 95% CI: 0.64–0.90). Specific activities: chatting OR=0.81 (0.66–0.98), reading news OR=0.76 (0.63–0.92), watching videos OR=0.85 (0.70–1.03; not significant), playing games OR=0.59 (0.42–0.84), mobile pay OR=0.67 (0.54–0.84), WeChat OR=0.75 (0.63–0.89), sharing Moments OR=0.79 (0.65–0.96). Device method: cellphone OR=0.78 (0.66–0.93) was significant; desktop, laptop, tablet were not significant. Corresponding reductions in CESD score changes were observed for most significant activities/methods (e.g., mobile pay β=−0.57, p=0.002; WeChat β=−0.40, p=0.008; cellphone β=−0.34, p=0.02). - Conversion to non-depressed among those depressed at baseline (n=5302): 36.2% became non-depressed by 2020. Social media transitions: compared with consistently inactive (reference), transition to active (≥1 activity) OR=1.14 (0.99–1.33; Ptrend=0.08); for thresholds, ≥2 activities OR=1.17 (1.00–1.37; Ptrend=0.04) and ≥3 activities OR=1.24 (1.05–1.46; Ptrend=0.01). Consistently active showed OR=1.31 (1.04–1.64), OR=1.33 (1.06–1.67), and OR=1.36 (1.09–1.72) across increasing thresholds, indicating greater likelihood of remission with higher or sustained engagement.

Findings show a heavy burden of depressive symptoms among middle-aged and older Chinese adults during the COVID-19 period, with disproportionate impact on women, rural residents, and western provinces. Longitudinal analyses indicate that engagement in social media is associated with reduced risk of incident depressive symptoms and greater likelihood of remission among those already depressed. These results address the research gap left by prior online convenience samples and support the potential protective role of social media for older adults, contrasting with reports of harm in adolescents/young adults. Potential mechanisms include enhanced social connection, access to peer support and services (including mobile payments and online shopping), and reduction of social isolation, especially in a cashless economy. Geographic disparities (e.g., Qinghai) may reflect environmental (high altitude, low temperature/oxygen, strong UV, limited greenness), socioeconomic (lower income, limited healthcare access), and lifestyle factors. Public health implications include targeting high-burden groups for mental health resources and digital inclusion efforts to expand beneficial social media access.

This nationally representative, longitudinal study documents substantial depressive symptom burden among middle-aged and older Chinese adults and demonstrates that engaging in social media is associated with lower incidence of depressive symptoms and higher remission among those depressed. The work supports integrating digital inclusion strategies (e.g., facilitating smartphone/internet use) into public health approaches for depression prevention and management in older populations. Future research should evaluate dose/duration effects of social media use, long-term outcomes beyond two years, causal mechanisms, and intervention trials promoting safe, supportive, and accessible digital engagement.

- Social media exposure lacked duration/intensity measures, precluding assessment of potential dose-response and addiction. - Follow-up limited to two years; long-term effects remain unknown. - Depression not clinically diagnosed; CESD-10 used for screening, which may introduce misclassification. Symptom fluctuation over time could cause non-differential misclassification, likely biasing associations toward the null. - Attrition bias: dropouts differed systematically (older, rural, lower education, more comorbidities, lower social media use), which may affect generalizability despite high follow-up rates. - Potential residual confounding cannot be fully excluded despite adjustment for extensive covariates.