Arash Alavi · Nature Medicine 2021 · prospective cohort study · n=3,318

Real-time alerting system for COVID-19 and other stress events using wearable data

Cited 154 times in the scientific literature.

Level 3 - non-randomized controlled study

Prospective cohort study evaluating an alerting algorithm without randomized controls

OpenAlex W3215572702 · doi:10.1038/s41591-021-01593-2 · record verified 2026-08-28

What was done

Researchers developed an online smartwatch-based detection system tracking continuous physiological and activity signals (heart rate and step counts) to detect early aberrant patterns. The system was prospectively tested in a cohort of 3,318 participants to evaluate its ability to trigger real-time alerts for pre-symptomatic and asymptomatic SARS-CoV-2 infections, with participant survey data used to evaluate other potential non-infection triggers.

What was found

Among 84 participants infected with SARS-CoV-2, the alerting system successfully flagged pre-symptomatic or asymptomatic infection in 67 individuals (80%). Signals occurred at a median of 3 days before symptom onset. While other events (such as other respiratory infections, psychological stress, alcohol intake, and travel) also generated alerts, they did so at a lower mean frequency of 1.15 alert days per person compared to 3.42 alert days per person in COVID-19 cases.

Why it matters

This demonstrates the feasibility of using continuous, consumer-grade wearable data to provide early advance warnings of viral respiratory infections several days before clinical symptom onset.

Limits

The study is observational and relies on self-reported surveys for symptom timing and non-infection triggers. The algorithm lacks specificity for SARS-CoV-2, as baseline physiological stressors (alcohol, travel, stress) also produced false-positive alerts, and 20% of confirmed infections were missed.

Cited by