Differential Effects of the Communities That Care Prevention System on Youth Problem Behaviors for Movers and Stayers: An Inverse Probability Weighting Approach

Background: In studies of community-level interventions, participants who move from their communities during the study period compared to those who remain may receive less exposure to intervention programming. Thus, intervention effects may be stronger among “stayers” compared to “movers.” Because moving occurs post-randomization, testing this hypothesis in randomized trials without accounting for other factors associated with moving status may induce bias. Previous intent-to-treat findings demonstrated effects of the Communities That Care (CTC) prevention system on reducing prevalence of youth problem behaviors in a longitudinal panel. This current study used inverse probability weights (IPWs) to examine whether effects of CTC were stronger among those who stayed in their community compared to those who moved during the study period. Applying weights according the inverse probability of one’s own moving status, such that those with higher propensity are down-weighted and those with a lower propensity up-weighted, creates a “pseudo-population” with balanced distribution of confounders across moving status.

Method: Data were from a community-randomized trial of 24 towns designed to test the efficacy of CTC. A cohort of 4407 youth was followed from 5^th to 8^th grade whether they remained in their original community (86%) or not (14%). IPWs for moving status were calculated with the denominator for the weights being the predicted probability of one’s own moving status prior to the 8^th grade according to baseline and 6^th grade covariates at the individual- (including demographics, substance use, known risk and protective factors) and community-level. A stabilized version of the IPWs was created by specifying the numerator of the weights as the predicted probability of moving status according to only baseline demographic covariates and study condition. The IPWs were applied to multilevel logistic regression models that examined effects of CTC on substance use by mover-stayer status.

Results: Weighted analyses showed effects of CTC on substance use in 8^th grade were stronger among those who remained in their original study community compared to those who moved between baseline and 8^th grade. For example, among stayers, CTC had statistically significant effects on reducing past month prevalence of use of alcohol, marijuana, and smokeless tobacco, as well as past two-week binge drinking. Among those who moved from their communities, all odds ratios were near 1 and not statistically significant.

Conclusions: Findings were consistent with a dose-response effect where CTC had more benefit for those who stayed in their community and received greater exposure to CTC programming. These findings strengthen support for CTC as an efficacious system for preventing youth problem behaviors.