Toward a Just-in-Time Adaptive Intervention to Reduce Emerging Adult Alcohol Use: Testing Approaches for Identifying When to Intervene

Abstract

Background: To identify critical periods for just-in-time adaptive interventions (JITAIs), we measured time-varying correlates of drinking (e.g. stress, mood) daily to predict near-term alcohol use. Methods: Emerging adults (aged 17–24; n = 51) who reported past-month alcohol use used SARA, an app use designed to assess substance use, for 30 days. Participants completed daily process measures of stress, mood, hopefulness, free time, fun, and loneliness. Candidate variables for prediction of next-day drinking included a contextual factor (day of the week), between-person factors (age, sex), and within-person factors (daily process measure responses) as well as daily process measure noncompletion. We compared two approaches to predict next-day use. From the daily process measure responses, Approach 1 used the current day’s survey responses; whereas, Approach 2 used the deviation of daily responses from the participant’s average response in prior days. Backward model selection identified candidate variables to include in the logistic model. Each model’s discriminatory power was determined using the area under the curve (AUC). Toward identifying critical periods for interventions, decision rules for when next day alcohol use was likely are reported for the better performing approach. Results: Approach 1 included day of the week, hopefulness, stress, and participant sex (AUC = 0.76). Approach 2 included day of the week, and deviation in hopefulness rating (AUC = 0.71). Decisional cutpoints are provided for the better performing model. Conclusions: Approach 1 provided better prediction than Approach 2. Decisional tools for identification of near-term alcohol use in emerging adults open the door for JITAIs to reduce drinking and prevent consequences of use. Abbreviations JITAI: Just-in-time adaptive intervention; ROC: receiver operating characteristic; AUC: area under the curve; MRT: micro-randomized trial