Therapeutic communities (TCs) for substance abuse treatment are based on mutual aid, with the community of peers considered to be the primary method of treatment. Clinical treatment consists of ongoing interactions with peers rather than periodic interaction with therapists. The most constant of these is peer feedback in the form of affirmations and corrections. Affirmations reinforce positive behavior, offer encour- agement to peers, and can counterbalance corrections. Corrections offer recipients the opportunity to take re- sponsibility for problematic behaviors and correct them, while demonstrating that their peers have responsible concern for their well-being. Over time, peer feedback forms a dynamic social network of affirmations and corrections. Therefore, improving our ability to intervene in the core TC clinical process of peer feedback requires an analytic approach that controls for complex network structures that change over time. This project implements a set of dynamic social network analysis approaches and has three specific aims:
Specific Aim 1 : Apply social network analysis to develop predictors of graduation from the TC based on peer exchange of affirmations and corrections. We will apply a recently developed dynamic social network model?Temporal Network Autocorrelation Model (TNAM)?to this problem.
Specific Aim 2 : Apply social network analysis to develop predictors of outcomes following graduation based on residents? peer exchange of affirmations and corrections prior to graduation. We will extend the TNAM to the case of survival analysis in order to conduct this analysis.
Specific Aim 3 : Learn which patterns of peer feedback exchange during treatment lead to network positions that predict outcomes that are more favorable.
This aim will lead to interventions to alter resident network trajectories. We will develop and apply a dynamic social network model?Temporal Generalized Exponential Random Graph Model (T-GERGM)?to analyze changes in social network positions over time. The research design for this project involves the development and programming of the necessary statistics and their application to a secondary analysis of a large de-identified clinical dataset of peer feedback interactions between TC residents drawn from six units at three TCs. The dataset includes demographic characteristics of the residents and graduation and reincarceration data. The multi-unit structure of the dataset allows for replication of findings at multiple sites. The resulting analyses will lead to an understanding of the feedback network predictors of successful TC outcomes along with a set of statistical tools to study social network outcome predictors in other mutual aid based programs.
This research is relevant to public health because the results will help us improve therapeutic communities and other peer treatment programs. This project uses social network analysis to study the relationship between resident participation in peer feedback and the outcomes of program graduation and reincarceration.
