Characterising the longitudinal relationship between social media use and psychiatric diagnoses in secondary care in adolescents in England

BackgroundPrevious research has not directly considered the link between social media use and demand for youth mental health services. In this study, we address this gap by examining the longitudinal relationship between social media use and psychiatric diagnosis in secondary healthcare. MethodsWe used Hospital Episode Statistics (HES) data linked to the Millennium Cohort Study, representing residents of England born in 2000–2002 (N = 12,586 with linkage consent rate 49.2%). We applied Cox proportional hazards models to separately model the hazard of having a psychiatric diagnosis recorded in HES in the three years following social media use recorded at ages 11 and 14. To select theoretically relevant covariates, we used directed acyclic graphs co-produced with experts with lived experience of mental distress and social media use. We used a combination of multiple imputation and inverse probability weighting to handle missing data, and sampling weights to render the sample representative of the target population.OutcomesWe did not find evidence for a longitudinal association between social media use and psychiatric diagnoses in either time period. However, post hoc analyses investigating non-linear relationships suggested a significantly increased hazard of psychiatric diagnosis among participants who reported no social media use at age 14 relative to those with moderate use.InterpretationClaims that social media are a major driver of increasing demand for youth mental health services are currently not evidenced. Future research should critically consider the potential risks, as well as benefits, of interventions into social media use in young people.FundingWellcome Trust and Economic & Social Research Council.