Abstract
BACKGROUND: Successful recovery from alcohol use disorder requires long-term lapse risk monitoring. Self-monitoring is difficult, given the dynamic, complex interplay of the many risk factors over time. An automated recovery monitoring support system embedded with a machine learning lapse prediction model could improve sustained, adaptive, and personalized self-monitoring by delivering daily support messages. OBJECTIVE: We propose to optimize the components included in daily support messages to increase engagement with a recovery monitoring support system. METHODS: The participants will include 304 US adults with moderate to severe alcohol use disorder. Participants will complete daily surveys and provide geolocation data for 17 weeks. Participants will receive daily support messages, starting in week 2, that convey a combination of individualized information from a lapse prediction model. Manipulated message components include (1) lapse probability and lapse probability change, (2) an important model feature, (3) a risk-relevant recommendation, and (4) message personalization on tone preference. RESULTS: The National Institute on Alcohol Abuse and Alcoholism funded this project (R01AA031762) on August 9, 2024, with a funding period from August 20, 2024, to July 31, 2029. The institutional research board of the University of Wisconsin-Madison Health Sciences approved this project (IRB #2024-0869). Enrollment will begin in December 2025. CONCLUSIONS: Message components that either increase engagement or improve clinical outcomes will be recommended for use in future recovery monitoring support systems and digital therapeutics. INTERNATIONAL REGISTERED REPORT IDENTIFIER (IRRID): PRR1-10.2196/81697.