Training a Machine Learning Model Using Wrist Biometric Data to Detect Opioid Use and Measure Withdrawal in People with Opioid Use Disorder

What this Trial Is About

Researchers are evaluating a machine learning model that uses biometric data collected from a wrist-worn device to identify acute opioid use events and measure opioid withdrawal levels in individuals dependent on opioids. The study focuses on patients undergoing medication for opioid use disorder MOUD induction, aiming to improve detection accuracy and withdrawal quantification compared to current measures. Participants will wear the OpiAID Strength Band Platform, a Samsung Galaxy Watch, continuously for 14 days except during charging or water activities. The machine learning model will be trained to detect MOUD events during the induction phase and to quantify withdrawal severity based on physiological data. The study includes a non-inferiority analysis comparing the new withdrawal measure to the Short Opiate Withdrawal Scale SOWS. During the 14-day monitoring period, participants will respond to daily prompts and complete the SOWS questionnaire on the watch. Researchers will collect time-stamped biometric data to evaluate classification success and withdrawal quantification accuracy. The study tracks how well the model detects opioid dosing events and correlates withdrawal levels with time since the last opioid dose. Participants are expected to wear and charge the device daily and comply with study procedures throughout the monitoring period.

A Study to Train a Machine Learning Algorithm for an Evaluation of the Use of Biometric Data Captured at the Wrist for the Identification of Acute Opioid Use Events and the Quantification of Opioid Withdrawal in Opioid Dependent Individuals