Using Latent Class Analysis to Examine Polysubstance Use Patterns in Adolescents Aged 10-18: A Systematic Review

Over half of substance use disorder (SUD) cases develop from substance use during adolescence. Studies have used latent class analysis (LCA) to determine adolescent polysubstance use patterns. This systematic review aims to provide an updated characterization of studies that examine patterns of adolescent polysubstance use.

This paper is part of a two-part systematic review examining patterns of polysubstance use. This review focuses on the adolescent literature, which includes study samples with mean age up to 18 years.

A majority of the N=54 studies were conducted in the United States. Sample sizes (N=98 to N=418,702), number of substances used as indicators (3 to 15), and number of latent classes extracted (2 to 6) varied per study. Analyses with larger sample sizes tended to extract a greater number of latent classes (Spearman's rho = .274, p = .022). The mean ages of samples (Spearman's rho = .255, p = .181) nor the utilization of more substances as indicators (Spearman's rho = 0.021, p = .861) were associated with the number of classes extracted. The Bayesian Information Criterion was the most used statistical fit index for the latent classes.

Substantial heterogeneity existed regarding study methodologies, LCA solutions, and statistical measures. Analyses focusing on niche population samples tended to highlight patterns involving less commonly used substances among adolescents. Future studies should expand on the types of substances used as indicators to explore how polysubstance use patterns are influenced by unique geographical communities.