A Component-Based Meta-Analysis of Family-Based Prevention Programs for Adolescent Substance Abuse

Use of controlled substances has serious implications for adolescent health and well-being, and research has consistently linked family-based factors with the initiation and escalation of substance use in adolescence. In response, numerous family-based prevention programs have been developed. Family-based programs work with family members to modify and manage emotions, cognitions, and behaviors, and create positive change in both individual behavior and interaction patterns among family members; these changes can, in turn, reduce adolescent substance use. Although these programs have proven to be successful, there is a lack of research that clarifies why they are successful. Thus, the field has little guidance as to how to optimize programs for greater efficiency or modify programs to target specific populations.

In this project, we applied meta-analytic techniques to examine aspects of family-based prevention programs most consistently associated with success at reducing adolescent substance use. Instead of comparing programs, we decomposed each program into a set of key topics or components that were addressed by program curricula (e.g., parental monitoring, family relations, peer risk, etc.). We then used the program components to predict effect sizes.

We conducted a document search in September 2012; our initial sample of 5071 studies was reduced to 128 studies that met all inclusion criteria. DV’s included outcome measures related to change in youth substance use. We coded sufficient information to calculate an effect size (ES) for each DV; a positive ES indicated a favorable outcome for the treatment group. For IV’s, we coded the amount of time that each program allocated to each component, in terms of youth-only, parents-only, and whole-family time. We developed a reliable coding system (ICC’s  > .90 for a randomly selected subsample) to capture the diversity of the different programs.

Since effect sizes (ES’s) were nested within programs, we used hierarchical linear modeling. In our final model, two youth-only components emerged as the best predictors of ES’s: improving family relations (B=.57, SE=.17, p < .001), and promoting more effective family problem solving (B=.29, SE=.12, p < .05). In an analysis of interaction effects, we found that the effects of the parent-only monitoring component were amplified by the youth-only family relations component (B=1.51, SE=.27, p < .001) and the youth-only substance use knowledge component (B=1.08, SE=.52, p < .05).

Our results provide a stronger understanding of how family-based programs work to prevent adolescent substance use, and our approach can be extended to additional outcomes and populations.