Longitudinal Study on Crack Cocaine Use and Dependence: Application of General Growth Mixture Modeling

In longitudinal studies, latent growth mixture modeling (GMM), which combines the conventional growth model with the latent class growth analysis (LCGA), has been increasingly used to study population heterogeneity in regard to developmental trajectories. The GMM can be further expanded to a more general latent variable modeling framework -- general growth mixture modeling (GGMM), which is the statistical framework used in Mplus (Muthén & Muthén, 1998- 2006). In GGMM, the model identified latent trajectory classes are related to antecedents (e.g., individual background characteristics) and consequence (e.g., distal outcome). This study applies GGMM to study crack cocaine use development among a sample of 430 crack cocaine users in Dayton, Ohio, over an 8-year period. In the study, trajectories of crack cocaine use are modeled and classified into a finite number of latent classes; individual background variables, such as gender, ethnicity, age, education, and early onset of crack cocaine use, are used to predict the class membership; and the latent trajectory classes are used to predict the likelihood of crack cocaine dependence in the end of the 8-year observation period. Mplus is used for modeling.