Comorbidity of psychiatric and substance use disorders in late adolescence: a cluster analytic approach.

Cluster analysis was used to identify subgroups of youths with past-year substance and/or psychiatric disorders (N = 110, mean age 19.0 years). Data for this study came from a community-based, prospective longitudinal investigation of speech/language (S/L) impaired children and matched controls who participated in extensive diagnostic and psychosocial assessments at entry into the study at 5 years of age and again at follow-up. Clustering variables were based on five DSM diagnostic categories assessed at age 19with the University of Michigan Composite International Diagnostic Interview. Using Ward's method, the five binary variables were entered into a hierarchical cluster analysis. An iterative clustering method (K-means) was then used to refine the Ward solution. Finally, a series of analyses of variance (ANOVAs) were run to analyze group differences between clusters on measures of Global Assessment of Functioning (GAF), criminal involvement, anxiety and depressive symptomatology, and frequency of drug use and heavy drinking. The analysis yielded eight replicable cluster groups, which were labeled as follows: (a) anxious (20.9%); (b) anxious drinkers (5.5%); (c) depressed (16.4%); (d) depressed drug abusers (10%); (e) antisocial (16.4%); (f) antisocial drinkers (10%); (g) drug abusers (8.2%); (h) problem drinkers (12.7%). These groups were differentiated by external criteria, thus supporting the validity of our cluster solution. Cluster membership was associated with a history of S/L impairment: A large proportion of the depressed drug abusers and the antisocial cluster group had S/L impairment that was identified at age 5. Clarification of the developmental progress of the youths in these cluster groups can inform our approach to early intervention and treatment.