Clinical Validation of Wearable Monitoring Devices

Abstract

Wearable monitoring devices have moved from consumer wellness tools to clinical instruments with regulatory clearance for cardiac rhythm detection, blood oxygen estimation, and continuous glucose approximation. Their clinical promise rests on continuous, passive data capture outside hospital settings -- enabling detection of paroxysmal arrhythmias, nocturnal hypoxaemia, and glycaemic excursions that episodic clinic measurements routinely miss. Whether this promise translates to diagnostic accuracy meeting clinical standards, and whether accuracy holds across the demographic groups most likely to benefit, remains incompletely characterised. We conducted a prospective clinical validation of five wearable monitoring device categories across 14,284 patients at 12 European clinical sites in Madrid, Stockholm, Berlin, Amsterdam, and Vienna between 2020 and 2023. Device categories covered single-lead ECG patches, photoplethysmography-based smartwatches, continuous glucose monitors, wrist-worn pulse oximeters, and multiparameter biosensor patches. Performance was benchmarked using our Wearable Clinical Validation Score (WCVS), integrating diagnostic accuracy against clinical reference standards, demographic equity, data completeness, clinical workflow integration, and patient acceptability. Multiparameter biosensor patches achieved the highest WCVS (0.906), with single-lead ECG patches scoring 0.891 and demonstrating the strongest arrhythmia detection sensitivity (94.2% for paroxysmal atrial fibrillation). Pulse oximeter accuracy showed significant skin-tone-dependent bias - overestimating SpO2 by 2.4-4.8 percentage points in patients with Fitzpatrick scale 5-6 skin tones -- representing the most clinically consequential equity finding across all device categories