Latent class analysis of substance abuse patterns

This chapter discusses use of latent class analysis (LCA) as a tool for identifying substance use patterns in cross-sectional data. LCA serves as an exploratory and data reduction tool that helps clarify the nature of substance use and may provide insight concerning effective prevention strategies. LCA is well suited to categorical data such as typically are collected in substance use research. Use of LCA can be divided into three steps: (1) model comparison and selection, (2) assignment of cases to latent classes, and (3) interpretation of the latent classes. Quantitative indices of model fit may assist model comparison and selection. Latent classes can be interpreted by examining probabilities of substance use in each latent class and by examining differences on exogenous variables. Limitations, extensions, and software for LCA are discussed. An example illustrates use of LCA with actual data collected from a current substance abuse prevention study.