How wearables can detect illness before you feel sick

Your body whispers before it shouts. Wearable data can catch the whisper.

Published 2026-03-22 · Updated 2026-09-24 7 min read AI & Prediction ↗ Longevity Topic Guide

Woman checking health data on her smartphone

The science of pre-symptomatic detection

In 2020, Stanford researchers published a landmark study in Nature Biomedical Engineering (Mishra et al.) showing smartwatch data could detect COVID-19 infection in 63% of cases before symptoms, with lead times typically a few days, and in four cases at least nine. The key signals? Elevated resting heart rate, decreased HRV, and changes in skin temperature.

Mishra 2020: what the watch data held in 32 confirmed cases

26 of 32 cases (81 %) carried a change in heart rate, daily steps or time asleep

63 % of cases could have been flagged before symptoms began, by an alarm watching resting heart rate against each person's own baseline

Measured values from Mishra et al. 2020. The 32 cases sit inside a cohort of roughly 5,300 people wearing a smartwatch, and the detection was calculated backwards, over data that had already been collected.

Radin et al. (2020, Lancet Digital Health) demonstrated that wearable data improved influenza-like illness surveillance at population level. The pattern is consistent: your autonomic nervous system responds to infection before you consciously feel anything.

How it works: When your immune system activates, your ANS shifts, increasing resting heart rate, suppressing HRV, and altering sleep architecture. These changes are subtle but measurable, often days before symptoms appear.

What signals indicate illness is coming?

The shape the studies describe: heart rate up, HRV down, before anything is felt

above
baselineyour
baselinebelow
baseline

a week beforethree days beforefirst symptoms

resting heart rate
HRV

A schematic of the direction the studies report, not measured data. They give the size and the direction of the change, not one shared curve, and no two people move along the same line.

1. Resting heart rate elevation

An increase of 3–5 bpm above your personal baseline, sustained for 2+ days, is a common pre-illness signal.

2. HRV decline

A sustained decline of 10–15% below your baseline indicates your ANS is under stress from immune activation.

3. Sleep disruption

Less deep sleep, more awakenings, longer sleep onset, often precede illness.

4. Body Battery / energy depletion

A morning reading significantly below baseline (especially after normal sleep) can indicate immune system activation.

Signal	What the research watches for	What it does not tell you
Resting heart rate	3 to 5 beats above your own baseline, held for two days or more	Alcohol, a late night, a hot bedroom and a hard session all lift it the same way
HRV	A fall of 10 to 15 % below your own baseline	It also falls with stress, travel and short sleep, none of which is an infection
Sleep	Less deep sleep, more awakenings, longer to fall asleep, across several nights	One broken night is noise. A single reading carries no information here at all
Morning energy	A reading well under your usual one after a night of normal length	It is a number derived from the others, not a measurement of the immune system

The bands these two signals are read against

Resting heart rate, above your own baseline

0+2+4+6+8 bpm

HRV, below your own baseline

0-5-10-15-20 %

Neither band means anything against a population average; both are read against your own history, and only when they hold for two days or more. One morning outside them is a bad night, not a signal.

Why the nervous system moves first

The reason these signals arrive early is not that a watch can see a virus. It is that the same nerve which sets your heart rhythm also sits in the loop that governs inflammation. Tracey described that loop in Nature in 2002 as the inflammatory reflex: the vagus nerve senses inflammatory signals and answers them, in seconds, the way it answers a change in blood pressure.

Because the loop runs through the vagus, its state is partly readable from the beat of the heart. Williams and colleagues pooled 51 studies in 2019 and found the association in the expected direction: higher vagal HRV goes with lower measured inflammation. The correlations are modest, and the authors say so. What they establish is the link itself, which is what makes HRV worth watching at all.

So the chain is: immune activation, then an autonomic shift, then a change in heart rate and HRV and sleep. Each step costs time, which is why the change can show up in the data before it shows up as a sore throat.

What separates a signal from noise

Raw data alone produces too many false positives. What brings their number down is context:

Personal baselines: compared to your own history, not population averages
Activity context: the system knows if yesterday was a hard training day
Multi-signal correlation: single-metric spikes are filtered; risk rises only when multiple signals converge
Personal patterns: with half a year of history an individual pattern becomes visible where a population average shows nothing
Trend analysis: deterioration over 2–3 days matters more than a single bad reading

Evidence base: Mishra 2020 (Nature), Radin 2020 (Lancet), Williams 2019 and Tracey 2002. The thresholds these studies work with are age-adjusted.

What this research does not show

The findings above are real, and they are also narrow. Four things belong beside them.

The headline numbers rest on small groups. The 63 % comes from 32 confirmed cases inside a cohort of roughly 5,300 people, and it was calculated backwards over data already collected, not raised as a live alert to anyone.
A signal arrives with no cause attached. A raised resting heart rate means the body is working harder. An infection does that; so do alcohol, heat, a new medication, a hard session and a bad week. The studies count how often the change appears, not what produced it.
Sensitivity is paid for in false alarms. An alarm set to catch most infections also fires on days when nothing is wrong, and the papers that report high detection rates report that cost in the same tables.
A population is not a person. Radin's result improves the influenza estimate for a whole state from 200,000 devices. That is a different claim from telling one named person what is about to happen to them.

None of this is a diagnosis, and no consumer wearable is an approved diagnostic device. What the research describes is a pattern that tends to appear early, often enough to be worth understanding.

What to do when risk rises

Prioritize sleep: your immune system works best during deep sleep
Reduce training intensity: shift to light activity or rest
Stay hydrated: support your body's immune response
Monitor trends: 2–3 days of rising signals is meaningful

Sources

Mishra T. et al. Pre-symptomatic detection of COVID-19 from smartwatch data. Nature Biomedical Engineering 4, 1208-1220 (2020). nature.com
Radin J.M., Wineinger N.E., Topol E.J., Steinhubl S.R. Harnessing wearable device data to improve state-level real-time surveillance of influenza-like illness in the USA: a population-based study. The Lancet Digital Health (2020). thelancet.com
Williams D.P. et al. Heart rate variability and inflammation: a meta-analysis of human studies. Brain, Behavior, and Immunity 80, 219-226 (2019). doi.org
Tracey K.J. The inflammatory reflex. Nature 420, 853-859 (2002). nature.com

Start your 7-day trial