Mishra · Nature biomedical engineering 2020 · Observational cohort study · n=5,300 (32 COVID-19 cases)

Pre-symptomatic detection of COVID-19 from smartwatch data.

Cited 480 times in the scientific literature.

Level 3 - non-randomized controlled study

Prospective observational cohort with retrospective algorithmic anomaly detection

PubMed 33208926 · doi:10.1038/s41551-020-00640-6 · record verified 2026-08-30

What was done

Physiological and activity data (heart rate, daily steps, and sleep duration) were collected from consumer smartwatches within a cohort of nearly 5,300 participants. The study analyzed data from 32 individuals confirmed to have COVID-19. A two-tiered warning system based on elevations in resting heart rate relative to individual baselines was retrospectively modeled to evaluate potential real-time pre-symptomatic detection.

What was found

Of 32 individuals with COVID-19, 26 (81%) showed alterations in heart rate, steps, or sleep duration. Among 25 cases with detectable physiological changes and available symptom data, 22 were flagged before or at symptom onset, including 4 cases detected at least 9 days prior. The modeled resting heart rate alert system detected 63% of COVID-19 cases prior to symptom onset.

Why it matters

This demonstrates that continuous data from consumer smartwatches can capture physiological signals prior to COVID-19 symptom onset, suggesting potential utility for early respiratory infection surveillance.

Limits

The study is limited by a small sample of infected cases (n = 32). The real-time warning algorithm was evaluated retrospectively. The abstract does not report specificity, false-positive rates during non-infection periods, or the ability to distinguish COVID-19 from other infections.

Cited by