Social-Media-Based Mental Health Interventions: Meta-Analysis of Randomized Controlled Trials

The study addresses the global mental health crisis, where more than 1 in 8 adults and adolescents live with a mental disorder and access to care is limited by stigma, workforce shortages, and affordability. Social-media-based interventions offer scalable, low-cost, and accessible delivery of psychoeducation, peer support, and therapeutic interactions. Prior syntheses have focused on broader digital and internet-based interventions or specific populations (eg, cancer patients), leaving a gap regarding rigorously designed social-media-based mental health interventions in general populations. The meta-analysis poses two research questions: RQ1 examines the overall impact of social-media-based RCTs on alleviating negative mental health outcomes (depression, anxiety, stress, negative affect, psychological distress) compared with care as usual or waitlist; H1 hypothesizes effectiveness. RQ2 investigates moderation by recruitment type (clinical vs nonclinical), age group, control group type (waitlist vs care as usual), intervention delivery (self-guided vs human-guided), program duration, program orientation (social vs task-oriented), and sex composition; H2 hypothesizes larger effects in clinical populations, younger age groups, more passive controls, human-guided, social-oriented programs, majority-female samples, and longer duration.

Past meta-analyses have extensively examined online, digital, eHealth, computer therapy, and internet-based mental health interventions, as well as social-media-based interventions in specific contexts (eg, patients with cancer) and scoping reviews for children or young people. However, no prior meta-analysis has focused specifically on rigorously designed social-media-based mental health interventions for general populations. Conflicting findings in prior work regarding moderators such as recruitment type, age, control group type, guidance, and duration informed the selection of potential moderators to re-examine in this domain.

Registration: The meta-analysis was preregistered on OSF. Deviations from protocol included broadening the population to include all ages with age used as a moderator, narrowing interventions to social-media-based only (excluding broader internet/mobile formats) to reduce heterogeneity, and focusing on negative mental health outcomes (depression, anxiety, stress, negative affect, psychological distress) to avoid heterogeneity with positive outcomes.

Search Strategy: A comprehensive search (completed and updated by April 2025) combined database queries (ERIC, PsycINFO, Scopus, PsycArticles, Communication and Mass Media Complete, PubMed, ProQuest), targeted hand searching via Paperfetcher, and forward/backward citation tracking using CitationChaser on relevant reviews. The search yielded 11,658 records and was managed in Covidence.

Eligibility Criteria: Inclusion required RCTs with at least 30 participants per condition at baseline; social-media-based delivery (eg, Facebook, Instagram, WhatsApp, WeChat) excluding social media abstinence; baseline group differences <0.25 SD on mental health measures; differential attrition <15%; interventions delivered by nonresearchers; quantitative measures of negative mental health outcomes with sufficient data to compute Hedges g; full-text available in English; published in or after 2005; primary studies (not secondary analyses); exclusion of one-item measures; exclusion of single-session interventions.

Screening and Coding: Titles/abstracts and full texts were double-blind screened by at least two authors. Eligible studies were double-coded using Google Spreadsheets, with disagreements resolved in weekly meetings to consensus. Procedures, coding sheets, and analysis code were shared on GitHub.

Analytical Plan: Analyses used R (metafor package). Effect sizes were standardized mean differences (Hedges g), using inverse-variance weights with random-effects meta-regression due to heterogeneity. Moderators (grand-mean centered) included recruitment type (clinical vs nonclinical), age group (adolescents, early adulthood, middle adulthood, late adulthood based on mean age), control group type (waitlist vs care as usual/active), delivery personnel (self-guided vs human-guided), program duration (weeks), sex composition (>70% female vs <70%), and program orientation (social-oriented vs task-oriented). Publication bias was assessed via selection modeling (weightr package). Risk of bias used the JBI Critical Appraisal Checklist for RCTs, focusing on allocation concealment, blinding of participants, deliverers, outcomes assessors, and follow-up completeness. Two independent reviewers assessed quality; a third resolved disagreements.

Moderator Coding: Recruitment type distinguished general/nonclinical samples from clinically identified samples. Age categories were determined by mean age. Control coding distinguished waitlist vs care as usual. Delivery coded self-guided vs guided by humans. Program duration standardized to weeks. Sex coded as ≥70% female vs otherwise. Program orientation coded as social vs task-oriented.

• Included studies: 17 RCTs; total sample size = 5624; 22 distinct intervention programs; 73 effect sizes (depression 31, anxiety 27, stress 12, negative affect 2, psychological distress 1).
• Overall effectiveness: Meta-regression indicates social-media-based interventions are effective (ES=0.32, P<.001; 95% CI 0.24–0.45; I2=88.10; τ²=0.13). Forest plot random-effects mean ES=0.32.
• Outcome subgroups: Depression ES=0.31 (P<.001; n=31); Anxiety ES=0.33 (P=.04; n=27); Stress ES=0.69 (P=.02; n=12). Negative affect (n=2) and psychological distress (n=1) were limited.
• Heterogeneity: High heterogeneity (I2≈88%). Prediction intervals indicate wide variability across potential future observations.
• Moderator effects: • Sex: Samples with >70% female showed significantly higher effects (β=1.40, P=.01; marginal mean ES=1.81 vs 0.41). • Delivery: Human-guided programs outperform self-guided (β=-0.72, P=.02; marginal mean ES=1.35 vs 0.63). • Program orientation: Social-oriented programs outperform task-oriented (β=-0.76, P=.03; marginal mean ES=1.20 vs 0.44). • Control group: Interventions with care as usual/active controls more effective than waitlist (β=-0.49, P=.02; marginal mean ES=1.37 vs 0.88). • Age: No significant moderation; adolescents vs middle adulthood comparison approached significance (P=.06). Late adulthood had largest marginal mean (limited data). • Clinical vs nonclinical: Not a significant moderator (P=.17). • Duration: Not a significant moderator (P=.26).
• Bias assessments: Risk of bias generally low (mean JBI appraisal score 9.29/13). Selection model indicated possible upward adjustment due to under-reporting of nonsignificant results. Sensitivity analysis using 50% female threshold yielded consistent findings.

Findings support H1: rigorously designed social-media-based RCTs reduce negative mental health symptoms across depression, anxiety, and stress. Moderator analyses inform H2 partially: greater effectiveness was observed for majority-female samples, human-guided programs, social-oriented content, and care-as-usual controls; age, clinical recruitment, and program duration did not significantly moderate effects. These results underscore the importance of social interaction and human support in digital therapeutic contexts, consistent with theories of therapeutic alliance and tend-and-befriend responses. The unexpected lower effects with waitlist controls may reflect participant engagement with alternative treatments during waiting periods, particularly in the post-pandemic telehealth environment. Compared with meta-analyses of broader digital or CBT-only interventions, social-media-specific effects are somewhat smaller, potentially due to relative novelty and variability in implementation fidelity. Contextual observations include female-only recruitment patterns in some predominantly Muslim countries, suggesting cultural and access-related factors that may make social-media delivery particularly acceptable to women. Overall, the results address the research questions by quantifying average benefits and identifying key design features that enhance effectiveness, highlighting relevance to scalable, accessible mental health care.

Social-media-based mental health interventions are effective, accessible, scalable, and cost-efficient tools for reducing depression, anxiety, stress, and related symptoms. The meta-analysis contributes new evidence focused on rigorously designed RCTs in general populations, identifying human guidance, social-oriented program design, and majority-female samples as conditions associated with larger effects. Integration of social-media-delivered interventions into routine mental health services and public health systems is recommended, particularly for populations facing barriers to in-person care. Future research should design more rigorous multi-arm RCTs, examine interactions between program orientation and participant characteristics (eg, sex, personality, comfort with technology, specific symptom domains), improve reporting on race/ethnicity to evaluate equity, and explore strategies to optimize effectiveness for male-majority samples.

• Limited number of included studies (17), constraining statistical power and degrees of freedom in meta-regression.
• Small subgroup sample sizes for certain outcomes and moderators (eg, psychological distress n=1, negative affect n=2, late adulthood n=1), warranting cautious interpretation.
• Publication/selection bias suggested by weight-function modeling, indicating potential upward adjustment of mean effects due to under-reporting of nonsignificant findings.
• Limited reporting on racial distribution across studies impeded analyses of equity-related moderators.