Clarifying Intervention Impacts on Social Networks: The Roles of Baseline Levels of Problem Behavior and Network Centrality

Evidence-based substance use prevention programs targeting middle school youth can alter peer networks in ways that reduce peer influence towards problem behavior through at least two distinct mechanisms. Specifically, programs could (a) decrease the influence potential (indegree centrality) of youth displaying high initial levels of problem behavior; or (b) have a disproportionate beneficial impact on the problem behavior of potentially influential youth. Prior analyses documented that the PROSPER intervention altered peer networks in ways that reduced influence toward problem behavior (Osgood et al., 2013) but did not clarify which of these mechanisms is more likely. We compare the plausibility of these two alternative explanations.

Data come from the PROSPER intervention trial that included 11,000+ youth from 28 communities, half of which received evidenced based interventions delivered to families (6^th grade) and in schools (7^th grade). Youth were surveyed fall and spring of 6^th grade and spring of 7^th, 8^th and 9^th grades. In each survey youth named up to 7 friends and reported on their own problem behavior (indexed here as a composite measure of substance use attitudes/behavior and delinquency). Influence potential was defined as indegree centrality (# of friendship nominations received).

In preliminary analyses examining whether the program decreased the influence potential of youth displaying problem behavior, we split the sample based on initial (fall 6^th) levels of problem behavior (25% highest vs remaining 75%) and examined intervention-control differences in indegree centrality across all subsequent assessments. No clear pattern of differences emerged, suggesting that youth with high initial levels of problem behavior did not become less influential after the intervention. Next, we split the sample based on initial (fall 6^th) levels of indegree centrality and examined intervention-control differences in problem behavior across subsequent assessments. A clear pattern suggested that intervention effects on problem behavior were stronger among youth who were initially high in indegree centrality: in 7^th- 9^th grades, treatment-control differences in problem behavior were 40%-80% larger among youth who were high in initial indegree centrality. Final multilevel models will include individual and community covariates and will consider program impact as a joint function of initial centrality and problem behavior.

We found that youth with greater initial influence potential exhibited fewer problem behaviors after the intervention. This suggests that the PROSPER program impact on peer influence may be best explained by the disproportionately strong impact on the problem behavior of youth who were initially high in influential potential.