Your wearable can see illness coming

In 5,262 people, 81% of COVID cases showed wearable-detectable changes, a median of 4 days before symptoms. The signal exists. It has nowhere to go.

By Hydra Editorial

In a Stanford study of 5,262 people, 81% of confirmed COVID-19 cases showed measurable changes in resting heart rate, daily step count or sleep. Detection came a median of four days before symptoms appeared. The signal was already there. What happens to it next is the unsolved part.

How early can a wearable detect an infection?

The clearest evidence comes from Mishra and colleagues, published in Nature Biomedical Engineering in 2020. Of 32 participants with confirmed COVID-19 and sufficient wearable data, 26 (81%) showed physiological alterations. Among 25 cases with a known symptom-onset date, 22 (88%) were detected at or before symptoms began, and four were flagged at least nine days ahead.

The detection method was not exotic. A real-time algorithm watched for sustained elevations in resting heart rate relative to each person’s own baseline, and would have caught 63% of cases pre-symptomatically. Mount Sinai’s Warrior Watch study found a comparable pattern using heart rate variability in healthcare workers, signalling infection up to seven days before a positive swab.

One honest limitation: 32 confirmed cases is a small denominator. The direction of the finding is well replicated, the precision of any single percentage is not.

Why doesn’t that signal reach a doctor?

Because detection and action are different problems, and only one of them has been solved.

A four-day head start is only useful if something happens with it. The physiological deviation that a wrist-worn sensor picks up lands in a consumer app, and stops there. It does not enter the medical record, it is not visible at the next appointment, and it carries no context about medication, family history, recent travel or a lab result from three weeks ago.

That context is what turns a number into information. A resting heart rate six beats above baseline means one thing during a heat wave, another during a hard training block, and another entirely in someone with a known cardiac history. The measurement is cheap. The interpretation is what is missing.

The bottom line

  • In a 5,262-person cohort, 81% of confirmed COVID-19 cases showed wearable-detectable physiological change, a median of four days before symptoms.
  • The detecting signal was deviation from a personal baseline, not a population threshold.
  • The constraint is not sensing. It is that continuous physiological data and clinical records still sit in separate systems.

Frequently asked questions

Can a smartwatch detect illness before you feel sick?
In published research, yes, for some infections. A 2020 Nature Biomedical Engineering study found 81% of confirmed COVID-19 cases showed changes in resting heart rate, steps or sleep, with a median lead time of four days. This is a research finding about populations, not a diagnostic capability.
What wearable metric gives the earliest warning?
Resting heart rate performed best in the Stanford analysis, because it is measured continuously during sleep and has a stable personal baseline. Heart rate variability and skin temperature showed similar patterns in other cohorts. In each case the signal is the deviation, not the absolute value.

Sources

  1. Mishra, T., Wang, M., Metwally, A.A. et al., 2020. Pre-symptomatic detection of COVID-19 from smartwatch data. Nature Biomedical Engineering, 4, 1208 to 1220.
  2. Mount Sinai Health System, 2021. Wearable Devices Can Detect COVID-19 Symptoms and Predict Diagnosis: the Warrior Watch Study. Mount Sinai, press release.
  3. Alavi, A. et al., 2025. Automatic detection of persistent physiological changes after COVID infection via wearable devices. Scientific Reports.
Hydra Editorial

Hydra's editorial team. We write from primary research and cite every study we rely on, with its year and journal. We quantify claims wherever the evidence allows and state plainly where it does not: where a finding rests on a small or single trial, we say so in the text rather than in a footnote.