Modeling the Diffusion of Family-Based Prevention: Strengthening Causal Inference of Social Network Effects on Adolescent Smoking

Introduction: Despite a substantial evidence-base around preventing tobacco use, initiation of smoking among minors remains a serious public health issue. While it is known that the family plays an important role in preventing smoking during adolescence, dissemination of family-based substance use prevention remains a major barrier. The current study addresses the question of whether the effects of a family-based prevention program can spread beyond the people directly receiving the program and if those effects are sustained across time. By using data from the PROSPER project, in which an evidence-based program was delivered with fidelity under real world conditions, we tested the diffusion of effects through adolescent social networks.

Methods: The participants were 5,083 middle school students in the 13 treatment communities of the PROSPER project that collected social network data. A family-based substance use prevention program (Strengthening Families Program 10-14; SFP) was offered to all participants in sixth grade (17% of the sample attended), and all students received a school-based prevention program in seventh grade. A propensity modeling approach was used to account for any potential confounding effects that could affect both peer selection (which is not randomizable) and cigarette use. The outcome model, weighted by the propensity score, tested the impact of having friends who received SFP in sixth grade on cigarette use in twelfth grade.

Results: The mean number of friends a student had who received SFP was 0.66 (SD = 0.92, range 0-6). 44% of the sample had at least one friend who received SFP. 63% had ever smoked a cigarette by the end of high school. Having friends at the beginning of sixth grade who attended a family-based program was significantly associated with lower odds of ever having smoked by the end of twelfth grade (OR = .90, 95% CI: 0.84, 0.97). Each additional friend represented a 2.45% decrease in the absolute likelihood of smoking by the end of high school.

Conclusions: The current study found significant effects of diffusion from participants in a family-based prevention program to nonparticipants through their social networks. Using a propensity model with a wide variety of covariates to account for potential peer selection effects, we found that having friends who received SFP in sixth grade was linked to lower odds of cigarette use by the end of twelfth grade. These findings suggest that the reach of family-based programs is broader than previously thought.