Along with alcohol use disorders come many negative related consequences, some examples being physical assault, academic consequences, death, and more. In hopes of informing intervention targets, researchers have developed many unique predictors of protective and risk factors for these adverse alcohol-related problems. Use of machine learning in behavioral research has been increasing over the past few years. Machine learning has been known to be a possible tool to help treatment providers diagnose and tailor treatment to patients. This study examines the convergence and stability of decision trees across multiple independent samples of college students.
The first sample of this study was recruited as part of the Protective Strategies Study Team (PSST) and focused on examining what is the precursor of harm reduction for alcohol and cannabis use. The sample consisted of 7,217 students who completed an online study survey. A second sample was recruited from the Addictions Research Team (ART). ART examined the precursor of several alcohol and cannabis-related outcomes and consisted of 5,497 students. Negative Alcohol-related Consequences were measured by assessing the Alcohol Use Disorder Identification Test (AUDIT) and Brief Young Adult Alcohol Consequences Questionnaire (BYAACQ). Predictor variables such as demographics, alcohol use, and other psychosocial indicators were also measured. This study consisted of 4 distinct decision trees with 71 potential predictor variables.
Results of this study revealed nine unique and noticeable predictors of negative alcohol-related consequences. These consisted of two negative reinforcement drinking motives (coping with depression and conformity motives), four alcohol use indicators (binge drinker frequency, typical quantity, heaviest quantity, drunk frequency), a substance use indicator, one subtype of harm reduction, behavior, and an indicator of mental health. These are all critical targets to consider for treating alcohol-related problems among college students.
Takeaway: This study revealed nine unique and noticeable predictors (such as coping with depression and binge drinking) of negative alcohol-related consequences.