Using the Group-Based Dual Trajectory Model to Analyze Two Related Longitudinal Outcomes

In this study, we used the group-based dual trajectory model to analyze two related outcomes jointly. The two outcomes are 1) the social contact with non-substance abusing friends (recoded as dichotomous) and 2) the substance abuse treatment scale rating (SATS). Annual data from the New Hampshire Dual Disorder 10-year Follow-up Study (n=223) were used for the analysis. Group- based logit and censored normal models are specified for contact and SATS, respectively. Four trajectory groups were identified for contact, and three trajectory groups were for SATS. Membership probability in each identified trajectory group for each outcome was computed, and the interrelationship across the trajectory groups between the two outcomes was examined. Some key individual-level covariates were included as predictors of conditional probabilities linking trajectories between the two outcomes. The results indicate that participants with no or minimal social contact were more likely to be in the group with lower SATS scores, meaning that they were at an earlier stage of substance abuse treatment (engagement or persuasion). Conversely, participants in the active social contact group were more likely to be in the group with higher SATS scores, meaning that they were at a later stage of substance abuse treatment (relapse prevention and recovery). Gender and education are statistically associated with the transition probability from contact to SATS. The results of the joint analysis provide an opportunity to understand more about the underlying phenomena as well as their dynamic linkage.