Demographic and health profiles of medicinal, recreational, and dual cannabis use: A comparative regression and tree-based analysis

Introduction Cannabis use is increasing in the United States with expanding legalization. Recreational, medicinal, and dual use reflect distinct health and social profiles, yet complex predictors remain insufficiently characterized. Methods We conducted a cross-sectional analysis of 2018-2021 Behavioral Risk Factor Surveillance System data (N=499,817 U.S. adults; 42,732 [11.5%] reported past-month use). Fifty-eight candidate predictors were evaluated; 38 remained after preprocessing. Survey-weighted analyses characterized population-level differences across use patterns. Multivariate logistic regression and tree-based models (random forest and gradient boosting) were compared for classification performance using area under the receiver operating characteristic curve (AUC). SHapley Additive exPlanations were used to rank predictive features. Predictive models were unweighted and assessed within the analytic sample. Results Among users, 27.1% reported medicinal, 36.4% recreational, and 36.5% dual use. Medicinal users were older (mean 48.6 years), female (49.9%), and had higher prevalence of arthritis and disability. Recreational users were younger (mean 32.8 years), male (64.7%), and more likely to binge drink (44.7%). Dual users exhibited elevated mental health distress (31.5%) and cigarette use (38.2%). Regression and tree-based models showed similar discrimination (AUC ~0.74-0.75 for medicinal/recreational; ~0.62-0.63 for dual). Conclusions Cannabis use subtypes demonstrate distinct profiles. Regression and tree-based approaches provided comparable classification performance.