P1-440 Latent class analysis of student substance use

Introduction Data were collected from a student population to investigate how substance use is related to perceived substance use of friends and the wider University population, during the period February to May 2009. The outcome measure was use of any of nine substances in the previous 3 months, giving differing patterns of use (trajectories) for each student. 4309 students were available for analysis. Methods We used latent class analysis (LCA) to classify the students into trajectory subgroups and investigated how these trajectories were associated with covariates of interest. The emerging classes contained types of students rather than all individuals. Model fit was explored comparing log-likelihood statistics with consideration of model parsimony. Results The model with two latent classes was preferred, containing one “low risk” class (3313, 90.6%) and one “high risk” class (344, 9.4%). Students in the “high risk” class perceived a higher proportion of others to be using more than the median amount of each substance, and had a higher mean knowledge score on a substance use quiz. Conclusion The latent class structure was informative, with students well differentiated into the two substance use classes. We have thus identified factors that are associated with heavier substance use. While educational campaigns that employ “scare tactics” are unlikely to be successful, a campaign highlighting the disparity between perceived and measured levels of risk could be developed and be targeted towards the different classes of substance use seen here

Class (philosophy) · Covariate · Environmental health · Latent class model · Population · Statistics · Substance use · Artificial Intelligence · Clinical Psychology · Computer Science · Demography · Mathematics · Medicine · Mental Health Research Topics · Online Learning and Analytics · Sensory Analysis and Statistical Methods