Effects of a 14-day social media abstinence on mental health and well-being: results from an experimental study

The study examines whether abstaining from social media for 14 days impacts mental health and well-being, focusing on screentime, problematic smartphone use (PSU), depression and anxiety, body image, Fear of Missing Out (FoMO), and loneliness. Motivated by rising global social media use and mental health concerns, the work combines a longitudinal and experimental design with follow-up to explore causal effects. Hypotheses: H1: PSU is positively associated with total screentime, depression and anxiety symptom severity, and FoMO. H2.1: Screentime is positively (weakly) associated with depression/anxiety, FoMO, and loneliness. H2.2: Screentime is negatively associated with body image. H3: Screentime decreases in the experimental (abstinence) group. H4: Depression and anxiety scores, and PSU scores decrease in the abstinence group. H5: Body image improves in the abstinence group. Research questions: RQ1: How does FoMO change over time (daily)? RQ2: What is the impact of abstinence on loneliness (daily)? RQ3: Are changes observed during abstinence stable after the intervention (14-day follow-up)?

Background summarizes links between PSU and adverse outcomes (poor sleep, impaired performance, musculoskeletal and visual issues) and positive associations with depression, anxiety, and FoMO. The Compensatory Internet Use Theory (CIUT) and I-PACE model provide frameworks for understanding overuse as coping and dispositional influences. Social media use is a major component of smartphone activity; problematic social media use (PSMU) has been linked to depression, anxiety, stress, sleep problems, cognitive failures, and lower meaning in life, with some longitudinal evidence of bidirectional links with depression and predictors of insomnia and suicide-related outcomes. Prior abstinence studies (often 7 days) report mixed effects on well-being, stress, anxiety, FoMO, loneliness, and sleep; effect sizes are generally small to moderate. Body image concerns have been associated with social media via exposure to idealized bodies and social comparison, though multiple factors influence body image beyond social media. The present study extends prior work by using 14 days of abstinence, daily assessments, and explicit focus on body image.

Design: Longitudinal experimental study with randomization and daily assessments during a 14-day intervention, plus a follow-up 14 days post-intervention. Period: October 2022 to February 2023. Recruitment via university mailing lists, flyers, social media, and eBay marketplace. Inclusion criteria: age 18+, German proficiency, smartphone and social media use. Platform: SurveyCoder; daily links emailed at 4 pm. Randomization to four groups: (1) control: use as usual; (2) social media abstinence: instructed to uninstall Instagram, Twitter, Facebook, Snapchat, TikTok; (3) gaming abstinence: uninstall gaming apps; (4) combined abstinence (social media + gaming). Participants could continue using smartphones for other purposes; app deinstallation was not objectively monitored. At intervention end, participants could reinstall apps and reported future social media intentions. Due to sample composition and data cleaning, analyses in the main manuscript focused on women and only compared control vs social media abstinence. Sample: Initial N=196 with complete datasets; after exclusions n=165; analyses restricted to female participants due to 83.6% female and body image focus. Gaming abstinence (n=21) and combined abstinence (n=31) excluded from main analyses (see Supplement). Final analyzed sample: n=86 female participants (control n=35; social media abstinence n=51). Groups comparable in age (mean control 23.17±6.99; abstinence 24±4.63), education, and occupational status. Measures: - FoMO: TS FoMO (trait and online-specific state; 12 items, 5-point Likert; α=0.76–0.83 and 0.77–0.79), at baseline, end, follow-up; daily single-item FoMO (FoMOsf). - PSU: Smartphone Addiction Scale—Short Version (SAS-SV; 6-point Likert; sum scores; α=0.81–0.86) at baseline, end, follow-up. - Depression and anxiety: PHQ-4 (0–3; sum; α=0.81–0.87) at baseline, daily, end, follow-up. - Loneliness: UCLA 3-item scale (1–3; sum; α=0.78–0.89) at baseline, daily, end, follow-up. - Body image: BIAS-BD (Body Image Dissatisfaction: difference between actual vs ideal BMI figure); MBSRQ-AS subscales (appearance evaluation, appearance orientation, body area satisfaction, overweight preoccupation, self-classified weight; 34 items; α=0.66–0.92). - Screentime: self-reported from smartphone screentime feature (minutes/day). Baseline and follow-up averaged last 7 days. - Fear of COVID-19 (FCV19S; α=0.83) at baseline as covariate. Additional questionnaires (IPAQ, PANAS-positive, PSS-4, SWLS) collected but analyzed in Supplement. Analysis: Conducted in R 4.1.3 using lme4 and lmerTest. Baseline descriptive stats and Holm-corrected correlations. Multilevel models: For outcomes measured at three time points (baseline, end, follow-up), spline models with knot at end of intervention assessed pre- and post-knot changes; covariates included baseline FCV, PSU, and BID (for MBSRQ-AS outcomes); group coded 0=control, 1=abstinence; random intercepts and, where possible, random slopes. Daily outcomes modeled with linear, quadratic, and cubic time trends; model comparisons via AIC, BIC, likelihood ratio tests; covariates included baseline PSU, FCV, BID, and the respective baseline outcome. Ethics: Approved by the Ethics Committee of University of Ulm (252/22); informed consent obtained. Data availability: OSF https://osf.io/qdp8r/.

Baseline associations (H1, H2.1, H2.2): - PSU positively associated with depression/anxiety and FoMO; not associated with screentime (text summary). - Screentime weakly positively associated with depression/anxiety, FoMO, and loneliness; weak negative associations with some body image facets (appearance evaluation, body area satisfaction) and positive with appearance orientation; correlations largely small and non-significant in this sample. Three-time-point spline models: - Screentime: In model including covariates and group, change from baseline to end of intervention differed between abstinence vs control (T1*Group b = -38.905, t = -1.685, one-sided p = 0.047), indicating decreased screentime in the abstinence group relative to control. - Body Image Dissatisfaction (BID): Change from baseline to end differed between groups (T1*Group b = -0.95, t = -1.900, one-sided p = 0.029), indicating larger BID decrease in the abstinence group. - Depression/anxiety: Overall pre-knot decrease (b = -0.60, t = -2.52, p = .013); no significant group differences and no post-knot change. - PSU: Overall pre-knot decrease (b = -4.023, t = -5.583, p < .001); FCV positively related to PSU; no group differences; post-knot nonsignificant. - Loneliness: Overall pre-knot decrease (b = -0.66, t = -4.751, p < .001); no group differences; post-knot nonsignificant. - FoMO state: Pre-knot decrease (b = -0.22, t = -3.813, p < .001); no group differences; post-knot stable. - FoMO trait: Pre-knot decrease (b = -0.52, t = -7.30, p < .001); no post-knot change; control showed significant pre-knot decrease; values at knot point (end) were lower in abstinence group. - Appearance evaluation: Increased pre-knot (b = 0.988, t = 2.736, p < .001), significant for control in model with covariates; no group differences. - Body area satisfaction: Near-significant pre-knot increase (b = 0.698, t = 1.899, p = .0597); not significant with covariates; no group differences. - Overweight preoccupation: Decrease pre-knot in control with covariates; no group differences. - Appearance orientation and self-classified weight: No changes. Daily models: - Screentime: Linear/quadratic trends not supported overall; when adding group, interactions for linear (b = 8.56, t = 1.800, p = .072) and quadratic (b = -0.61, t = -1.727, p = .084) changes were near-significant, suggesting different patterns in abstinence group. - Loneliness: Best modeled by cubic trend; significant linear, quadratic, and cubic components; no significant group differences; model with cubic trend fit best (AIC ~ 3596.9–3597.2; BIC ~ 3647.2–3667.6). - Depression/anxiety (daily): Model with group showed an almost significant decrease for control (b = -0.04, t = -1.834, p = .067) and a significant interaction with abstinence group (b = 0.06, t = 2.429, p < .05), indicating different trajectories. - FoMO (daily): Best modeled by cubic trend with significant linear, quadratic, and cubic components; no group differences; cubic model fit best (AIC ~ 2862.6; BIC ~ 2918.0). Follow-up stability (RQ3): Most spline models showed no significant post-knot (end to follow-up) changes, indicating stability of effects at 14-day follow-up, including screentime and BID changes observed between groups. Sample characteristics: n = 86 females (35 control; 51 abstinence). Baseline screentime mean ~ 245 minutes/day; groups comparable at baseline across key measures.

The findings partially support the hypotheses. PSU showed expected positive associations with depression/anxiety and FoMO but not with screentime, underscoring that screentime is an imperfect proxy for PSU and aligning with CIUT and I-PACE perspectives on coping and dispositional factors. Screentime decreased in the abstinence group relative to control, consistent with H3 and prior abstinence literature; day-to-day fluctuations and compensatory use may explain the lack of clear daily trends. Depression/anxiety and PSU decreased overall across the sample, but group differences were not observed in the spline models, thus H4 was not supported for depression/anxiety, and PSU reductions may reflect general intentions to reduce smartphone/social media use across participants. Daily depression/anxiety showed divergent trajectories between groups, potentially reflecting loss of a coping outlet for abstainers (CIUT), though effects were modest. Body image outcomes partly supported H5: BID decreased more in the abstinence group and appearance-related satisfaction tended to improve overall, suggesting reduced exposure to social comparison on visual platforms benefits body image. FoMO (trait and state) decreased during intervention and remained stable, but without group differences; daily FoMO and loneliness were best captured by cubic trends, indicating fluctuations independent of abstinence. Overall, most changes persisted to the 14-day follow-up, suggesting short-term stability. The results highlight heterogeneity in responses to abstinence and point to the importance of assessing motives and types of smartphone use when designing interventions.

A 14-day social media abstinence led to significant reductions in screentime and greater decreases in body image dissatisfaction compared to controls, while many mental health indicators (depression/anxiety, PSU, FoMO, loneliness) showed overall improvements over time without clear group differences. Daily FoMO and loneliness exhibited cubic fluctuation patterns, and most changes were stable at a 14-day follow-up. These findings inform the nuanced effects of social media abstinence and support incorporating body image outcomes into interventions. Future research should employ objective usage tracking, assess problematic social media use (PSMU) and motives of use, explore tailored abstinence or reduction strategies, and include diverse samples (including men and gaming-focused groups) to improve generalizability and power.

- Analyses restricted to female participants due to sample composition; gaming and combined abstinence groups excluded, resulting in smaller analytical sample and limited generalizability. - No objective verification of app deinstallation or usage; screentime self-reported from device feature may introduce transfer error. - PSMU was not measured; focus on total screentime and PSU may miss social-media-specific effects. - Potential participant self-selection and intention to reduce use across groups may confound group comparisons. - Single follow-up at 14 days limits inference on longer-term stability; limited statistical power means several directionally consistent effects did not reach significance. - Did not assess motives and specific uses of smartphones/social media, or concurrent life events that may affect outcomes.