Quantifying At-Home Physiotherapy Participation: SPARS vs Self-Reported Diaries

Abstract

The completion of at-home physiotherapy exercise is key to many rehabilitation protocols. This study compares at-home upper extremity physiotherapy participation as measured based on data captured with a smart watch to that recorded in self-report diaries. Daily at-home exercise participation (sessions) was recorded for 53 patients with rotator cuff pathology during their first 2 weeks of a 12-week physiotherapy rehabilitation program. Exercise participation was measured using a physical therapy monitoring system that uses smart watch (accelerometer/gyroscope) data analyzed via a convolutional neural network trained on labeled patient-specific in-clinic data and compared to patient reported diaries. A high level of agreement between diary exercise participation and the measurements derived from the smart watch data (ICC=0.72, n=53) was found, with an AUROC=0.99 for binary identification of exercise periods on labeled clinic data. However, overall patient diaries reported more exercise performed (0.96 additional days on average) than measured by the ML algorithm. ML and accelerometer/gyroscope data collected by embedded sensors in a smartwatch represents an accurate and objective alternative to self-reported diaries for monitoring patient at-home participation. Lower levels recorded by the ML algorithm may indicate some limitations in the technology to fully capture participation or potential over-reporting of participation within diaries. As self-reported diary completion decreases over time, physical therapy monitoring technology may represent an acceptable method for longer term assessment of exercise participation.