The Effectiveness of Life Skills-Based School Violence Prevention Programs in Secondary Education: A Systematic...

Background: School violence is a major public health concern, with bullying alone affecting up to 50% of adolescents worldwide. Life skills-based programs have been proposed as a promising preventive approach, yet evidence for their effectiveness in secondary education remains limited. This systematic review and meta-analysis aims to synthesize the existing evidence on the effectiveness of life skills-based violence prevention programs among adolescents aged 12-18. Methods: A systematic literature search was conducted across five databases (PubMed, Scopus, ERIC, CINAHL, and PsycInfo) in accordance with PRISMA 2020 guidelines (INPLASY202650168). Interventional studies published between 2005 and 2025 that targeted secondary school students were eligible. Risk of bias was assessed using the Cochrane RoB2 (for RCTs/cRCTs) and ROBINS-I tools. A three-level random-effects meta-analytic model with Robust Variance Estimation was employed to pool Hedges' g effect sizes. Results: Nineteen studies met the eligibility criteria, with 18 contributing 56 effect sizes to the meta-analysis. The pooled effect size was statistically significant (g = -0.2652, 95% CI [-0.4879, -0.0425], p = 0.0196), indicating an overall small reduction in violent behaviors. However, a high degree of variation in effects was observed across studies (Q(55) = 2263.58, p < 0.001; I² = 99.23%). None of the examined moderators-including outcome type, study design, and risk-of-bias tier-significantly explained this heterogeneity. The sensitivity analysis, excluding one highly influential study, yielded a smaller pooled effect (g = -0.1422, 95% CI [-0.2246, -0.00598]) while maintaining high heterogeneity (I² = 92.82%). Conclusions: Although life skills-based programs demonstrate a statistically significant overall average effect in reducing school violence among adolescents, this finding must be interpreted with extreme caution given the high degree of variation in effect sizes across contexts and the fact that prediction intervals span across zero. Because moderator analyses revealed no significant differences across intervention types, study designs, or outcome categories, no specific program type can be identified as consistently more effective. Future primary research must prioritize prospectively registered trials with multi-informant assessments to establish true intervention effectiveness.