Peer effects on health-related behaviours of adolescents in Scotland using social network analysis: a cross-sectional...

Previous research suggests that adolescent norms and behaviours may be influenced by peers. The aim of this study was to investigate social clustering of health outcomes among school friendship groups.

Cross-sectional surveys were collected from Oct 26, 2022, to March 30, 2023, in four secondary schools in Scotland's central belt, and all Secondary 2 (12-13 years) and Secondary 4 (14-15 years) students were invited to take part. Schools self-selected into the study, between 6% and 27% had a free school meal registration (Scotland average 25%). The survey asked about health and about friendships in school. The outcomes of interest were binary indicators of: mental health and self-esteem using validated scales, smoking, drinking without parents knowing (DWPK), and trying drugs. Ethics approval for the study was given by the University of Glasgow (200190035) and all participants gave consent via an online form. We used Auto-Logistic Actor Attribute Models (ALAAMs) to model the association between features of individuals' social networks and their health outcomes. We specified a model for each health outcome separately including parameters: indegree, outdegree, and simple contagion, and combined using meta-analysis.

Response rate was 74% (n=1097; 50% boys, 46% girls, 4% other). Based on self-report measures, 40% participants had poor mental health, 15% had low self-esteem, 6% smoked regularly, 4% tried drugs, and 18% were drinking without parents knowing. Preliminary unadjusted analysis found evidence of social contagion for mental health. Odds of poor mental health for each additional friend with poor mental health was 1·15 (95% CI 1·05-1·26). There was no evidence of contagion for self-esteem (1·13, 0·95-1·34), smoking (1·14, 0·46-2·82), DWPK (0·88, 0·71-1·10), and having tried drugs (0·91, 0·38-2·19). Some networks had low or zero prevalence of the outcomes, increasing the uncertainty in the pooled estimate for the contagion parameter.

A cross-sectional study cannot differentiate between social contagion and selecting similar friends, and low prevalence and social desirability bias might have masked associations. However, the unique combination of social network data with advanced statistical modelling gives initial findings on the potential communicable nature of mental health and health behaviours in adolescence. Preliminary results indicate preventive approaches in schools could benefit from social network methods.

Medical Research Council (MRC) and Chief Scientist Office (CSO).