BackgroundSocial media might influence adolescents' perceived social norms and subsequent engagement in behaviours with potentially negative effects on health (health-risk behaviours). We systematically reviewed studies investigating the relationship between social media and health-risk behaviours in adolescents (age 10–19 years). MethodsWe searched five databases (CINAHL, Embase, MEDLINE, APA PsycINFO, and SocINDEX), four preprint repositories (SSRN, SocArXic, PsyArXiv, and MedRxiv), and Google Scholar from Jan 1, 1997, to April 30, 2020, for studies reporting social media use (time spent, frequency, and exposure to health-risk behaviour content) and engagement in health-risk behaviours (PROSPERO: CRD42020179766). No geographical or language limits were applied, but only English-language studies were included, due to translation issues. Risk of bias was assessed using a modified Newcastle Ottawa Scale. We used random-effects meta-analysis; potential effect modification was explored with meta-regression and stratified analyses. FindingsOf 13 142 hits, 84 studies (n=571 175 participants) were included (79% [66 studies] cross-sectional): 55% (46 studies) with high risk of bias, 20% (17 studies) moderate, and 25% (21 studies) low. The most commonly investigated exposure was frequency of social media use (43 studies). Binary exposure meta-analyses showed that frequent social media use (versus infrequent) was associated with increased tobacco use (OR 1·8, 95% CI 1·4–2·3; n=236 300; four studies), alcohol use (1·5, 1·4–1·6; n=208 318; six studies), drug use (1·2, 1·0–1·5; n=59 823; three studies), gambling (2·8, 2·0–4·0; n=26 537; five studies), sexual risk (1·7, 1·4–2·1; n=46 483; eight studies), anti-social behaviour (1·7, 1·4–2·0; n=53 539; four studies), and multiple risk behaviours (1·8, 1·3–2·4; n=43 571; two studies). Increased time (2·3, 1·6–3·4; n=10 831; three studies) and exposure to health-risk behaviour content (2·4, 1·3–4·7; n=14 731; five studies) was associated with increased harmful alcohol use. Heterogeneity varied across different outcomes (I2 16–98%). Sensitivity analyses showed that effect sizes were generally larger for cross-sectional studies than cohort studies. For alcohol use, anti-social, and sexual risk behaviours, effects were mostly larger in studies adjusted for pre-specified critical confounders (age, sex, and socioeconomic position) and those at low or moderate risk of bias, while for gambling and tobacco use, larger effects were found in unadjusted studies, and those with higher risk of bias. InterpretationSocial media use is adversely associated with unhealthy adolescent behaviours, with exposure to harmful content potentially linked to increased risk. However, more robust longitudinal studies are needed. Given increasing use of social media by adolescents and targeting of social media by unhealthy commodity industries, action to mitigate risks is needed. FundingMedical Research Council, Chief Scientist Office, NHS Research Scotland, and Wellcome Trust.
Read full abstract