Right Place, Right Time: Validation of a Consumer-Grade Wearable for Heart Rate and Heart Rate Variability Across Sleep-Wake Cycles, Physical Activity, and Postures in an Ambulatory Study

Background: Consumer-grade PPG wearables are increasingly used in ambulatory studies to monitor heart rate (HR) and heart rate variability (HRV), yet their accuracy without raw data or ECG reference remains uncertain.Objective: To identify when and under what conditions HR and HRV data from a consumer device (Garmin Vivosmart 4) most closely align with ECG-derived values.Methods: Sixty-two participants wore both the Garmin and the VU-AMS Core (ECG reference) during daily life. Contextual data on movement (steps), sleep/wake status, and posture (via EMA) were collected. HR and HRV metrics (mean HR, mean HP, SDNN, RMSSD, LF, HF) were compared across matched 5-minute epochs using MAAPE, Spearman correlation, and Bland–Altman analyses. Machine learning models were also tested to identify whether Garmin-derived parameters could distinguish between high-quality and low-quality PPG data. Results: Agreement between devices was highest during rest, no steps, sleep, and sitting or lying postures. Mean HR and HP were estimated reasonably well, while HRV metrics, especially in the frequency domain, showed greater error and context sensitivity. Device agreement varied notably across individuals. Machine learning models achieved F1 scores of 0.76 (HR) and 0.69 (RMSSD), indicating limited ability to detect reliable epochs from Garmin data alone.Conclusions: Our findings emphasize the challenges and limitations of using consumer-grade PPG wearables in ambulatory research, particularly due to their sensitivity to movement, posture, and individual variability. We provide practical recommendations on when and how HR and HRV data from these devices can be used with higher confidence and when to be cautious.