Latent Classes of Polysubstance Use: Patterns By Grade Level in a Sample of Disadvantaged, Racially Diverse, Rural Youth

Background/Significance: Adolescent substance use is a critical public health issue, particularly among rural youth. Latent class analysis reveals that these substances form unique patterns of middle and high school youth polysubstance use, and that the high school pattern is non-linear and more complex than the middle school pattern.  These cross-sectional findings are consistent with results of prior longitudinal studies that show increasing severity of use as youth age. However, classifying students into middle and high school may obscure more complex patterns as students age, reflecting a gradual increase in the types and frequency of drug use. In this analysis, we answer the following question: are the polysubstance use latent class patterns exhibited for middle and high school students isomorphic to the corresponding grade levels, suggesting that these patterns emerge in a way that is associated with the transition from middle to high school? 

Methods: To answer this question, a latent class model of five substances—alcohol, cigarettes, marijuana, inhalants, and prescription drugs) was developed for grade levels 7 to 11 (6^th and 12^th excluded due to insufficient samples). The total sample was 6,030. We attempted to replicate the finding from an earlier paper pertaining to the level of schooling for that grade (e.g., the middle school model for 7^th grade) using the same fitting approach. 

Results: With one exception, the grade level models were simpler than the model for the corresponding level of schooling. Youth in 7^th, 8^th and 9^th grade all sorted into two major classes: abstainers or low frequency users, and youth who initiated use with alcohol, cigarettes, and marijuana (and less so with inhalants and prescription drugs). Within grade, the higher frequency users found in the middle school model were more similar and there were not enough of either to distinguish them. Abstainers were consistent across 7^th-9^th grades and on most substances were isomorphic to the middle school pattern. Only 10^th grade exhibited an overall pattern isomorphic to the high school pattern. In 11^th grade, experimental and high severity users were combined into a single class for a 2 class model.

Conclusions/Implications: Polysubstance use among adolescents is complex, but adheres to a pattern of increasing type and severity as the grade level increases. Alternatively, the 2 class solutions for 7^th through 9^th grades may not be more informative than a simple cumulative model adding binary indicators of lifetime risk. This analysis suggests that although transition to high school is a risk factor for more frequent and varied substance use, the pattern and frequency of use gradually increases during adolescence, with youth differentiating themselves in 10^th grade. The middle and high school models may obscure this gradual trend. These classes can inform timely grade level-specific interventions, and further analyses may support the use of predictive modeling to identify youth at risk for frequent substance use prior to initiation and provide them with targeted interventions. These conclusions are limited by the cross-sectional nature of the data. Future studies from a longitudinal perspective will better elucidate these trends.