Real-time Alerting System for COVID-19 Using Wearable Data.
Level 3 - non-randomized controlled study
Prospective observational cohort study evaluating a diagnostic alerting algorithm.
PubMed 34189532 · doi:10.1101/2021.06.13.21258795
What was done
Researchers developed an open-source, real-time alerting system that monitors smartwatch data (including resting heart rate and step counts) to detect individual physiological anomalies indicative of early infection. The system was tested in a cohort of 3,246 participants and evaluated alongside over 100,000 survey annotations tracking COVID-19 diagnoses, symptoms, and non-COVID events.
What was found
The system generated alerts in 78% of pre-symptomatic and asymptomatic COVID-19 cases, detecting signals a median of 3 days prior to symptom onset. Non-COVID events (such as other respiratory infections, stress, alcohol consumption, and travel) also triggered alerts, with a mean alert period of 1.9 days compared to 4.3 days for COVID-19 cases.
Why it matters
Continuous passive monitoring via consumer wearables can identify physiological anomalies days before clinical symptoms emerge, offering a scalable method for early infection detection.
Limits
The abstract does not report key diagnostic performance metrics such as specificity, false positive rates, or positive predictive value. Non-infectious factors like stress and alcohol trigger alerts, which may limit specificity, and the study was evaluated as a preprint.
Cited by
- supports Resting heart rate increases on average about 4 days prior to symptom onset in people infected with COVID-19.