Fellow and Senior Data Scientist
The objective of this methodological study is to develop and systematically test predictive models of an individual's alcohol consumption. The ability to forecast alcohol consumption in patients can potentially help individualized treatment programs. For binge drinkers, predicting the timing of their next episode provides additional awareness that could help to control, or even to prevent, the entire episode. For heavy drinkers, predictive models can forecast the windows of treatment opportunity when the patient is most sober and responsive to intervention. For those who are attempting to quit drinking, forecasting the next relapse episode could be used to trigger motivational interviewing or other timely interventions that will help to prevent the relapse. In this project, we will analyze two datasets containing long time series' (i.e., 2 years and 6 months) of individual daily records of alcohol use, stress, and other factors. We will adapt innovative predictive models developed to forecast future alcohol consumption, as well as identify and explain a variety of daily use patterns. The data on individual alcohol consumption is very limited, and this is the first study aimed at building forecasting models using such data. This study offers the field a "next step," with an innovative analysis approach that can possibly offer clinical implications for relapse prevention, increased treatment efficiency, and enhanced understanding of the factors driving the variety of daily patterns of use.
NATIONAL INSTITUTE ON ALCOHOL ABUSE AND ALCOHOLISM