Using decision trees to examine risk profiles for cannabis use among large samples of underage youth before and after cannabis legalization in Canada

Background: This paper compares risk profiles for cannabis use among large samples of youth in the school years preceding (2017-18, T₁) and four years following (2021-22, T₂) cannabis legalization in Canada. Methods: COMPASS Study data from students across 85 secondary schools that participated in both the T₁ and T₂ waves were used. A novel classification tree approach examined current cannabis use (past 30-day), modelling complex interactions among multiple risk factors simultaneously in the T₁ and T₂ samples. Results: At T₁, 15.0 % of students reported current cannabis use, compared to 12.3 % of students at T₂. The classification tree at T₁ identified six unique risk profiles. The highest risk group (Pr = 0.269) was large (30.4 % of the sample) and comprised students who placed lower value on getting good grades and spent 45 min or more per day texting. The classification tree at T₂ identified 11 unique risk profiles. The highest risk group (Pr = 0.27) was large (18.8 % of the sample) and comprised students who again placed lower value on getting good grades but also reported not eating breakfast daily and having elevated anxiety. Discussion: Cannabis never use increased and current cannabis use slightly decreased among underage youth in a 4-year period spanning cannabis legalization. The relative importance ranking of risk factors for predicting current cannabis use changed considerably from T₁ to T₂. This suggests that prevention efforts need to adapt over time to target the relevant risk factors associated with cannabis use.