Validating Heart Period Estimates From the Garmin Vivosmart 5 Smartwatch and Mindware ECG Across Various Tasks
Year: 2026
Type: preprint
Abstract: Consumer wearable devices, like smartwatches, may offer a more practical way to collect cardiac data in everyday life. Yet, to be used in a research study, prior work suggests comparing the accuracy of smartwatches against a gold-standard electrocardiography (ECG) device across various conditions. In the present work, we tested the accuracy of the Garmin Vivosmart 5, a wrist-worn smartwatch, against the research-grade Mindware Mobile ECG in collecting heart period data. Participants wore both devices as they completed 7 tasks, including physical and psychological stressors. We assessed how closely the two devices' heart period estimates agreed within each task, the magnitude and direction of the differences between devices, and whether the two devices detected the same task-related changes from baseline task estimates. Agreement between the devices was assessed using absolute-agreement intraclass correlations (ICCs) for each task. We found that agreement was good to excellent during seated tasks (ICC $= 889-.997$), but poor during movement tasks (ICCs $< .2$). Averaged across all tasks and participants, the differences in heart period between Garmin and MindWare was small and not significant, $M = 12.3$~ms, 95\% CI $[-8.6, 33.3]$. However, we also found that the magnitude and direction of the differences between Garmin and MindWare were significantly different across tasks, $F(6, 85.4) = 5.16$, $p < .001$, with the largest differences occurring during the movement tasks. Our findings suggest that the Garmin Vivosmart 5 can reliably estimate heart period (provide similar heart period estimates to the Mindware ECG) especially during seated tasks, while tasks with heavy movement present larger discrepancies that should be accounted for.
Source: PsyArXiv (OSF Preprints)
Authors: Dasha A. Yermol, Katie Hoemann
Language: English
Subfield: Biomedical Engineering
Field: Engineering
Domain: Physical Sciences
Open Access status: green
OpenAlex ID: W7234963459
