Consequences of Adolescent Substance Use Latent Class Membership: Associations with Young Adult Distal Outcomes

Introduction: Latent class analysis has been shown to be useful for identifying profiles of co-occurring risky behaviors, such as substance use behaviors. In addition, use of substances at an early age is widely documented as a risk factor for later substance use, abuse, and comorbid behaviors. Linking substance use latent class membership to a distal outcome requires an approach that can yield unbiased estimates of that association. Therefore, in this study we model adolescent substance use as a latent class variable, and employ a new, model-based approach for LCA with distal outcomes to estimate the association between adolescent substance use class membership and various young adult health outcomes. Unlike commonly used classify-analyze approaches to this problem, including pseudo-class draws, this model-based approach does not attenuate the effects of interest.

Methods: This study used data from 731 adolescents from the 9^th grade cohort of the National Longitudinal Study of Adolescent Health. Adolescents that were missing on all substance use indicators or on any outcome variables were deleted from the sample. A latent class model of substance use was estimated using indicators of past-year cigarette, alcohol, and marijuana use at Wave II. Then we estimated associations between substance use latent class membership and outcomes at Wave IV, when our sample reached young adulthood, using the LCA_distal SAS macro in conjunction with PROC LCA.

Results: Four classes of adolescent substance use emerged: Non-users (50.0%) were unlikely to report any substance use in the past year; Smokers (16.5%) were likely to have used cigarettes and may have smoked cigarettes regularly in the past year; Drinkers (13.4%) were likely to use alcohol, to drink five or more drinks in a row, and to have been drunk in the past year; and Heavy Users (20.1%) were likely to report cigarette, alcohol, and marijuana use behaviors in the past year. Preliminary results indicate that adolescent substance use class membership significantly predicted regular cigarette use and lifetime illegal drug use in adulthood (p < .0001 for both outcomes). Adolescents in the Smokers class had the highest probability of regular cigarette use in adulthood compared to other classes. In addition, adolescents in the Heavy Users class had the highest probability of using illegal drugs by adulthood compared to other classes, and adolescents in the Non-users class had the lowest probability of using illegal drugs by adulthood.

Conclusion: These results demonstrate a model-based approach to examining the effect of latent class membership on distal outcomes. Findings will be useful for better understanding the consequences of particular adolescent substance use behavior profiles during young adulthood.