B1630
Title: Detection of medication taking using wrist-worn commercially available wearable device
Authors: Quy Cao - University of Pennsylvania (United States) [presenting]
Abstract: Medication non-adherence is a persistent and costly problem across healthcare. Measures of medication adherence are ineffective. Methods such as self-reporting, prescription claims data, or smart pill bottles have been utilized to monitor medication adherence. Still, these methods are subject to recall bias, lack real-time feedback, and are often expensive. A method is proposed for monitoring medication adherence using a commercially available wearable device. Passively collected motion data was analyzed based on the Movelets algorithm, a dictionary learning framework that builds person-specific chapters of movements from short frames of elemental activities within the movements. The Movelets method is adapted and extended to construct a within-patient prediction model that identifies medication-taking behaviours. Using 15 activity features recorded from wrist-worn wearable devices of 10 patients with breast cancer on endocrine therapy, it is demonstrated that medication-taking behaviour can be predicted in a controlled clinical environment with a median accuracy of 85\%. These results in a patient-specific population are exemplars of the potential to measure real-time medication adherence using wrist-worn commercially available wearable devices.
