Mobile phone sensors and supervised machine learning to identify alcohol use events in young adults: Implications for just-in-time adaptive interventions

• Published in: Addictive behaviors Vol. 83; pp. 42 - 47
• Main Authors: Bae, Sangwon, Chung, Tammy, Ferreira, Denzil, Dey, Anind K., Suffoletto, Brian
• Format: Journal Article
• Language: English
• Published: England Elsevier Ltd 01.08.2018; Elsevier Science Ltd
• Subjects: Adult; Adults; Alcohol; Alcohol use; Alcoholic beverages; Alcoholism - diagnosis; Alcoholism - prevention & control; Artificial intelligence; AWARE app; Biosensing Techniques - instrumentation; Biosensing Techniques - methods; Cell Phone; Cellular telephones; Drinking behavior; Ecological Momentary Assessment; Female; Humans; Just in time; Learning algorithms; Machine learning; Male; Monitoring, Ambulatory - instrumentation; Monitoring, Ambulatory - methods; Prospective Studies; Sensors; Smartphone sensors; Supervised Machine Learning; Surveys and Questionnaires; Typing; Young Adult; Young adults; Smartphone sensors; Alcohol; AWARE app; Machine learning
• ISSN: 0306-4603; 1873-6327; 1873-6327
• DOI: 10.1016/j.addbeh.2017.11.039

Real-time detection of drinking could improve timely delivery of interventions aimed at reducing alcohol consumption and alcohol-related injury, but existing detection methods are burdensome or impractical. To evaluate whether phone sensor data and machine learning models are useful to detect alcohol use events, and to discuss implications of these results for just-in-time mobile interventions. 38 non-treatment seeking young adult heavy drinkers downloaded AWARE app (which continuously collected mobile phone sensor data), and reported alcohol consumption (number of drinks, start/end time of prior day's drinking) for 28days. We tested various machine learning models using the 20 most informative sensor features to classify time periods as non-drinking, low-risk (1 to 3/4 drinks per occasion for women/men), and high-risk drinking (>4/5 drinks per occasion for women/men). Among 30 participants in the analyses, 207 non-drinking, 41 low-risk, and 45 high-risk drinking episodes were reported. A Random Forest model using 30-min windows with 1day of historical data performed best for detecting high-risk drinking, correctly classifying high-risk drinking windows 90.9% of the time. The most informative sensor features were related to time (i.e., day of week, time of day), movement (e.g., change in activities), device usage (e.g., screen duration), and communication (e.g., call duration, typing speed). Preliminary evidence suggests that sensor data captured from mobile phones of young adults is useful in building accurate models to detect periods of high-risk drinking. Interventions using mobile phone sensor features could trigger delivery of a range of interventions to potentially improve effectiveness. •Mobile phone sensor data is useful in building accurate models to detect periods of drinking.•Useful sensor features relate to activity/movement, phone use/calls, and keystrokes.•Interventions could use phone sensor features to trigger remote support when it is most needed.