Are there distinctive trajectory groups in substance abuse remission over 10 years? An application of the group-based modeling approach

Abstract

This paper introduces to mental health services and evaluation researchers a method for analyzing longitudinal data: a group-based modeling approach for applied research. We present the basic formulations of the model and illustrate its application by analyzing the 10-year longitudinal data on substance abuse remission from the New Hampshire Dual Disorders Study. The basic features of the approach are: identifying latent groups with distinctive trajectories, estimating the proportion of the population in each group, linking individual-level covariates to group membership, assigning individuals to different groups, and creating group profiles based on posterior probabilities. We also discuss cautions and the controversies surrounding this approach. The major findings include four groups with distinct trajectories of remission from substance abuse.