Purpose: Substance abusers are likely to be addicted to multiple substances at the same time, and the patterns of substance abuse are different in each case. Understanding the structure of substance abuse patterns for treatment participants is essential to design an effective intervention strategy and to provide adequate services. However, less is known about the types and patterns of multiple substance uses among treatment program participants. Applying latent class analysis, the recent development of statistical tool with categorical data, the study focused on identifying meaningful clusters of substance abusers of treatment program participants, examining the nature of each latent class, and finding important predictors of substance abuse patterns.

Methods: The study used the data from the National Institute on Drug Abuse's Drug Abuse treatment Outcome Study (DATOS) collected from 96 treatment programs in 11 large U.S. cities to reflect typical community-based treatment services available to the public (N=8636). Latent class analysis was applied to determine the latent structure of substance use patterns for the weekly uses of each substance (heavy alcohol, cocaine, crack, marijuana, heroine and other drugs). Latent class analysis in this study had the following structured steps; 1> model selection, 2> assignment of cases, 3> latent class interpretation, and 4> model testing. Finally, multinomial logit model on class membership followed to determine the effects of social demographic variables (race, sex, age, education, marital status, etc.) and risk/protective variables (mental health, victimization trauma, social support, community involvement, etc.) on substance abuse patterns

Results: After multiple latent class analyses examined, the unrestricted three-class model produced the best fit (BIC=57523.94, entropy=0.95) and interpretable results. Weekly uses of heavy alcohol, crack, and cocaine were the most important variables (ANOVA, p<.001) to distinct each group profile (1> no alcohol and less substance group, 42%; 2> medium alcohol and high cocaine, crack, 20%; 3> high alcohol and medium substance, 38%). By multinomial logit regression, social demographic variables except race did not influence the latent structure of drug abuse patterns. However, most of risk/protective variables are significant (p<.05) to predict a class membership.

Implications for practice: Recognition of common patterns of substance abuse assists social workers to develop a theoretical understanding of substance abuse, to design effective prevention interventions, and to implement targeted services to each class of substance abusers. Findings also indicate that assessing risk/protective factors are important to identify substance abuse patterns and to design a targeted effective intervention strategy.