Diffusion of Intervention Effects: Explaining the Impact of a Family-Based Substance Use Prevention Program on Non-Participants

Introduction: We use social network analysis to test whether and how diffusion occurs when a substance-use prevention program is delivered to only a fraction of the targeted population. Several prior studies provide evidence that effects from behavioral interventions may diffuse through social networks, but more work is needed to test whether having more friends who participated in an intervention, or amount of indirect exposure to the intervention, is linked to non-participants’ outcomes. More work is also needed to identify specific mediating processes that explain how indirect exposure may improve non-participants’ behavioral outcomes. Accordingly, we test whether there is a link between indirect exposure and non-participants’ behaviors, and which program effects on participants account for diffusion of intervention effects to their friends.

Method: As part of the PROSPER intervention trial, all families of 6th graders in 14 school districts were invited to a family-based substance use intervention (Strengthening Families Program 10-14 [SFP10-14]), but less than 20% of families attended any sessions. We focus on 5,449 students who did not participate in SFP10-14 (51% Female; 81% White, 7% Latino, 12% Other). All students completed surveys in Spring of 6^th, 7^th, 8^th, and 9^thgrades, reporting on past month drunkenness and cigarette use. Respondents also named up to 7 friends, which we used to compute the proportion of each individual’s friends (cumulatively across waves) who attended SFP. Students also indicated how often they spent time with each friend without adult supervision. Based on the SFP-participating friends’ self-reports, we computed for each non-participant the cumulative mean of their friends’ family relationship quality (average of 5 subscales), parents’ discipline style (5 items), anti-substance use attitudes (average of 5 subscales), and past month drunkenness or cigarette use.

Results: Across waves, 40-46% of non-participants had at least one SFP-attending friend. Multilevel modeling analyses indicated that there was a significant negative association between amount of indirect exposure and drug use: Non-participants with more SFP-attending friends were less likely to use drugs than their peers. Friends’ family context was a weak mediator of this association, whereas unsupervised time with friends and friends’ substance use attitudes were stronger mediators. Friends’ drug use and non-participants’ own substance use attitudes were also significant mediators.

Conclusions: Although intervention developers strive to reduce diffusion or “contamination” in evaluation studies, the field should also consider, conceptualize, and test how interventions may facilitate diffusion.