Methods: Generalized linear mixed-effects models with random intercepts for schools were used to examine interactions between ease of access from different sources in predicting past 30-day alcohol and cannabis use at follow-up. Predictors included perceived access from home, adults, peers, dealers, and stores/dispensaries. Covariates included age, gender, race, food insecurity, and group assignment (though intervention effects were unlikely given implementation timing). Inverse probability weighting adjusted for missing data.

Results: For alcohol, a significant home x adult access interaction was found. Compared to adolescents who perceived difficult access in both contexts, those with easy access from both home and adults had 2.85x higher odds of subsequent alcohol use (p<0.001). Among single-source access, easy home access alone (aOR=3.31, p<0.001) and easy adult access alone (aOR=2.67, p<0.001) predicted subsequent alcohol use. For cannabis, home x adult access followed the same trend (aOR=2.78, p<0.01 for easy access on both fronts). A significant peer x adult access interaction for cannabis use also emerged. Compared to difficult access in both contexts, easy adult access alone was associated with the highest estimated odds of subsequent cannabis use (aOR=7.70, p<0.001).

Conclusions: Perceived easy home and adult access were the strongest solo predictors of alcohol and cannabis use, but their combination did not confer added risk beyond single-source access. Findings indicate family-based prevention to target home and adult access with additional focus on peer dynamics in cannabis prevention.