Methods: 77 schools were randomly allocated to deliver the GBG for two years (n=38) or continue with their usual practice (n=39). Participants were N=3,084 children aged 6-7 at baseline. Conduct problems and emotional symptoms were assessed using the Strengths and Difficulties Questionnaire (SDQ; Goodman, 1997). Psychological wellbeing, bullying, peer and social support, and school connectedness were assessed via the Kidscreen-27 survey (Kidscreen Group, 2006). Following Kellam et al (1994), our at-risk subgroup were boys who scored in the borderline or abnormal range of the SDQ conduct problems scale at baseline. Dosage data (total number of minutes played in the GBG) was used as a proxy for intervention compliance.
Results: Analysis is on-going at the time of submission. In brief, intention-to-treat (ITT) and subgroup analyses are being conducted using two-level (school, child) hierarchical-level models with fixed effects and random intercepts in MLWin. Missing data are being addressed via multiple imputation in REALCOM-Impute. Complier average causal effect (CACE) estimation in MPlus will be undertaken to robustly ascertain the effects of moderate (50th percentile) and high (75th percentile) compliance on intervention effects.
Conclusions: Our findings will contribute significantly to the GBG evidence base, and more broadly, the field of school-based prevention, by reporting not just 'what works', but 'for whom' and 'how/why'.