Pedagogical implications of peers, leisure and institutional predictors of risk behaviour among university students

Introduction Risk behaviours among university students co-occur, yet they are usually studied separately. Drawing on social learning theory and ecological frameworks, this study examined the contributions of peer dynamics (peer risk behaviours, quality of peer relationships), leisure-time activities (STEAM, sports, virtual activities, serious leisure engagement) and institutional relationships (relationships with teachers, media engagement) to four risk-behaviour outcomes: delinquent behaviour, online risk behaviour, disordered eating and risky sexual behaviour. Methods A convenience sample of 439 Croatian university students (72.2% female) completed an anonymous online questionnaire. Owing to the non-normal distribution of the outcomes, Spearman's rank-order correlations were used in the bivariate analyses. A multiple regression model was estimated for each outcome, and a latent risk-behaviour construct indicated by the four outcomes was tested with structural equation modelling (SEM) with robust standard errors. Results Peer risk behaviours were the strongest predictor of all four outcomes in the regression analyses ( B = 0.079–0.332; p < .01) and in the SEM ( β = 0.643; p < .001). Peer relationship quality was negatively associated with online risk behaviour, whereas its association with disordered eating was not robust in the sensitivity analyses. STEAM activities were unexpectedly positively associated with all four outcomes, and relationships with teachers were negatively associated with disordered eating. The SEM fit was acceptable but not excellent ( χ ²(26) = 114.90; robust CFI = .912; RMSEA = .088; SRMR = .041), and all four outcomes loaded significantly but unevenly on the latent factor ( λ = .47–.85). A one-factor measurement model reproduced the covariation among the outcomes closely ( χ ²(2) = 0.69, p = .71); the misfit of the full model reflected outcome-specific associations of the predictors, mainly with online risk behaviour and disordered eating. Discussion Peer deviance emerged as the strongest proximal correlate of risk behaviour, a pattern consistent with peer socialisation but equally compatible with peer selection, whereas serious leisure engagement and institutional climate were only weakly and outcome-specifically associated with it. Because of the cross-sectional design, the findings are interpreted as associations rather than causal effects. Implications for prevention programmes in the university context are discussed.