An Algorithm to Classify Real-World Ambulatory Status From a Wearable Device Using Multimodal and Demographically Diverse Data: Validation Study (Preprint)

Abstract

BACKGROUND Measuring physical activity amounts and patterns using wearable sensor technology in real-world settings can provide critical insights into health status. OBJECTIVE We trained an algorithm that classifies binary ambulatory status (yes or no) on accelerometer signal from a wrist-worn Biometric Monitoring Technology (BioMeT) and tested its analytical validity and generalizability. METHODS BioMeT algorithm validation traditionally relies on large numbers of self-reported labels or on periods of high-resolution monitoring with reference devices. We used both methods on data collected from two distinct studies for algorithm training and testing, one with precise ground-truth labels from a reference device (N=75), and the second with participant-reported ground-truth labels from a more diverse, larger sample (N=1691); in total, we collected data from 16.7 million 10-second epochs. We trained a neural network on a combined dataset, measuring performance in multiple held-out testing datasets, overall and in demographically-stratified subgroups. RESULTS The algorithm was accurate classifying ambulatory status on 10-second epochs (AUC = 0.938; 95% CI, 0.921-0.958) and on daily-aggregate metrics (daily Mean Absolute Percentage Error [MAPE] = 18%; 95% CI, 15-20%), without significant performance differences across subgroups. CONCLUSIONS Our algorithm can accurately classify ambulatory status using a wrist-worn device in real-world settings, with generalizability across demographic subgroups. CLINICALTRIAL NA