Multi-level patterns predict cannabis use onset among youth

Early cannabis initiation during youth is associated with elevated risk for harmful substance use, mental disorders, and cognitive impairments. To account for the complexity behind cannabis use initiation, we performed a data-driven analysis across 151 measurements spanning seven domains from the individual, microsystem, and exosystem level of influences: biobehavior, cognition, brain MRI, family, peer, neighborhood and legal factors. Data were from 450 cannabis-naïve youths from the National Consortium on Alcohol and NeuroDevelopment in Adolescence (NCANDA) (baseline age: 12-21 years). Within an 8-year period, 292 transitioned to first use and 163 to weekly use of cannabis. Random Survival Forest predicted age of first onset (C-index = 0.68; 95 % CI: [0.65,0.71]) and weekly onset (C-index = 0.69; 95 % CI: [0.67-0.71]) with an accuracy significantly higher than chance (i.e., C-index = 0.5). Its prediction patterns consisted of factors from all three levels of influence. The predictive pattern of first onset comprised 13 factors across six domains including lower positive thinking during stress coping, which correlated with earlier use (R²=0.023, p = 0.0090). Three variables were shared with the predictive pattern of weekly use onset: cannabis outlet density, access to alcohol at home, and more positive social expectations of alcohol use forecasting earlier onset (Initial Use: R²=0.031, p = 0.0027; Weekly Use: R²=0.023, p = 0.0090). Weekly use onset was predicted by only four factors suggesting that while many influences contribute to a youth trying cannabis, only a few key factors appear to facilitate escalation to habitual use, some of which represent promising targets for prevention programs.