Multiple substance use patterns in adolescents-A multilevel latent class analysis.

Background: Multiple substance use among adolescents is associated with a number of negative consequences. Therefore, we aim to investigate multiple substance use patterns among young adolescents and identify possible multilevel predictors.

Methods: We analyzed a longitudinal sample of 2490 German students (51% male; Mage=13.32, SD=0.57) at 45 schools in four German states (Bremen, Hesse, North Rhine-Westphalia, Schleswig-Holstein), who completed two assessments in fall, 2010 and fall, 2013. We conducted multilevel latent class analysis for follow-up data on a variety of outcomes, and tested our final 3-class-model for possible baseline predictors. Follow-up substance use measures included lifetime use, current use, and amount of substance for cigarettes and alcohol. Baseline covariates comprised age, gender, socio-economic status, bullying, victimization, peer and parental use, type of school, and health certification of school.

Results: We identified three latent classes: non-users (n=1541; 61.9%), experimenters (n=722; 29.0%), and multiusers (n=227; 9.1%). Experimental consumption was predicted by higher baseline age (OR=1.71; 1.31-2.24), paternal drinking (OR=2.89; 1.23-6.79), and school type (OR=2.57; 1.83-3.61), while multiuse was predicted by peer smoking (OR=2.94; 1.80-4.80) and drinking (OR=2.13; 1.32-3.44), maternal drinking (OR=6.26; 2.02-19.43), bullying (OR=1.69; 1.15-2.48), higher age (OR=1.92; 1.40-2.62), and school type (OR=4.76; 2.75-8.24) compared to the non-users class.

Conclusions: Prevention and further research on multiple substance use need to concentrate on social influence models and behavior-related interventions, especially at schools without a college-preparatory track.