Person-centered approaches to modeling trajectories of concurrent substance use across adolescence and adulthood: individual and family background predictors

Informed by theoretical approaches that emphasize variation in the developmental pathways of substance use (e.g., Moffit, 1993; Muthén & Muthén, 2000), the current study examined two person-centered approaches to assessing concurrent substance use across adolescence and adulthood (ages 16 to 28). Person-centered approaches have the advantage of capturing heterogeneity within a sample thus allowing for the explicit assessment of different developmental pathways of substance use for subsections of a larger population. Furthermore, trajectories of concurrent substance use have seldom been modeled in the extant literature, partially due to the complexity of data and models required to do so. Instead, studies have primarily relied on one indicator or one specific substance over time, which limits the extent to which those models accurately reflect individuals’ lived experiences. The analytical sample for the current study was drawn from the Center for Education and Drug Abuse Research (CEDAR, 2015) dataset and included 722 predominantly White male participants, approximately half of whom had fathers with diagnosed substance use disorders (SUD). Substance use was assessed across five waves of data from age 16 to age 28. Two approaches to modeling concurrent substance use trajectories were assessed: the multiple-indicator multilevel (MIML) growth mixture model (GMM) and the parallel processes latent class growth analysis (LCGA) model. Each model identified heterogeneity in substance use over time. Furthermore, family background and individual predictors differentially predicted membership into the profiles providing some evidence of at-risk versus normative patterns of substance use over time. Results indicated both the MIML GMM and parallel processes mixture model were appropriate methods for modeling concurrent substance use over time. Whereas results from the multiple-indicator multilevel growth mixture model indicated approximately 75% of the sample being classified as increasing low users, results from the parallel process mixture model indicated only 56% of the sample was classified as predominantly increasing low alcohol-only users. The typologies identified via these two different approaches are an important first step in assessing concurrent substance use trajectories from adolescence into adulthood and advance research that has been limited to a focus on modeling only one substance at a time. Furthermore, the ability of this study to identify at-risk versus normative patterns of use while simultaneously accounting for concurrent substance use is especially helpful for clinicians working with individuals who use or abuse substances