Predicting Success in Digital Cannabis Self-Help Tools: a Machine Learning Study

Background: For individuals who wish to reduce their cannabis use without formal help, there are a variety of self-help tools available. Although some are proven to be effective in reducing cannabis use, effect sizes are typically small. More insight into predictors of success among frequent cannabis users with the desire to reduce/quit their cannabis use could help identify factors that contribute to successful outcomes.

Methods: We analyzed data from a randomized controlled trial comparing the effectiveness of the digital cannabis intervention ICan to four online modules of educational information on cannabis. Success was defined as reducing the grams of cannabis used in the past 7 days by at least 50% at 6-month follow-up. To train and evaluate the machine learning models we used a nested k-fold cross validation procedure. 

Results: Of the 253 participants included, 124 participants successfully reduced their cannabis use, 129 participants were unsuccessful. The results show that the two models applied had comparable AUROC values of .61 (Random Forest) and .57 (Logistic Regression). Not identifying yourself as a cannabis user, not using tobacco products, high levels of depressive symptoms, high levels of psychological distress and high initial cannabis use values were the relatively most important predictors for success. 

Conclusions: Our study finds modest prediction accuracy when using machine learning models to predict success among cannabis users with the desire to reduce/quit cannabis use and with interest in digital self-help tools. However, the study provides valuable insight into the characteristics that are most strongly associated with success.