PROJECT SUMMARY/ABSRACT Medications for opioid use disorder (MOUD), including the partial opioid agonist buprenorphine, provide a treatment option for opioid use disorder (OUD) that significantly reduces morbidity and mortality. Even with successful buprenorphine initiation, however, adherence is paramount to prevent return to non-medical opioid use and its associated risks. Current methods of determining buprenorphine adherence are limited by their retrospective nature and recall bias. We propose to develop a novel artificial intelligence-assisted wearable sensor system, MINDER, which will continuously monitor physiologic changes, and will use machine learning algorithms to accurately identify buprenorphine use. The MINDER system will be comprised of a custom wearable sensor (MINDER-band), a companion mobile app and a clinician facing portal. The MINDER-band, which is a low profile, upper arm band with a user-driven design, continuously records physiologic data. We will use the band to curate a high-quality dataset of MOUD ingestions and subsequently use machine learning to evaluate the ability of the sensor to detect MOUD (specifically buprenorphine) ingestion events. Finally, we will deploy the MINDER system in real-world MOUD treatment settings to understand usability factors. The investigative team brings together complementary expertise in toxicology/addiction medicine, mobile health (Carreiro, Smelson), machine learning, human computer interaction (Venkatasubramanian), novel on-body wearable sensors, and medical device development (Mankodiya, Solanki). The specific aims of the project are to: 1) Understand the requirements, barriers, and facilitators for an ML driven buprenorphine adherence support system, 2) Develop and test a novel wearable sensing system, MINDER, designed for individuals in buprenorphine treatment, 3) Curate a high quality annotated dataset for machine learning-based modeling of buprenorphine adherence, 4) Model the buprenorphine ingestion data collected from the MINDER-band to build the ML algorithms infrastructure for the MINDER system. Upon completion, the MINDER system will be ready for clinical deployment. This study will lay the groundwork for novel just-in-time adaptive behavioral interventions to personalize OUD treatment, improve buprenorphine adherence and its success, and ultimately reduce morbidity and mortality from OUD.

NARRATIVE STATEMENT Medications for opioid use disorder such as buprenorphine provide an effective treatment option that significantly improves outcomes; however impact is limited by adherence challenges. We propose to develop a novel wearable sensor system, MINDER, which will continuously monitor physiologic changes to detect buprenorphine use and will use machine learning algorithms to accurately identify adherence to buprenorphine use in laboratory and field settings. The results of this study will form the foundation for a wearable, sensor- based, real-time monitoring system for individuals being treated with MOUD, and to develop precision medicine interventions for OUD.