Illness and HRV: What a Wearable Can and Cannot See

Why can illness move HRV?

Heart rate variability (HRV) is the beat-to-beat variation in the interval between heartbeats. Inflammation, the body's response to infection or injury, is thought to shift the balance of the autonomic nervous system: less vagal activity and more sympathetic activity. That is why researchers ask whether HRV from a wearable can reflect inflammation.

The only systematic review of this question pooled the direction of findings from eleven studies with 2,419 participants [S1]. When the inflammation marker CRP (C-reactive protein) was raised, SDNN was lower in 83% of comparisons [S1]. For RMSSD and inflammatory cytokines the results were mixed and mostly not significant [S1]. Devices that recorded an ECG gave more consistent results than optical pulse sensors [S1].

The reviewers draw a careful conclusion. At present, wearable HRV should be considered "an exploratory or adjunctive biomarker" [S1]. None of the included studies tested how accurately HRV can recognise inflammation [S1]. So the link is real at the level of groups, but it is a research finding, not a test.

What has picked up infection early?

This is the key point of this page: the best-studied early warnings of infection did not run on HRV. They ran on resting heart rate compared with the person's own baseline, together with daily steps and sleep (resting heart rate).

A retrospective analysis from Stanford looked at smartwatch data from 32 people with COVID from a cohort of nearly 5,300 participants [S2]. Most of them had changes in heart rate, daily steps or time asleep around the illness [S2]. An alert system based on a sharp rise in resting heart rate against the personal baseline could have flagged 63% of the COVID cases before symptoms began [S2]. This is a single study with few infected people, and the alerts were simulated afterwards, not sent in real time.

The same group then tested real-time alerts in a prospective study of 3,318 participants, 84 of whom were infected with the coronavirus that causes COVID [S3]. The system used heart rate and steps from smartwatches [S3]. It sent alerts before or without symptoms in 67 of the 84 infected people (80%), and the first signals came at a median of 3 days before symptom onset [S3]. Infection was confirmed by tests, not by the watch.

These are three studies from one research group, so they are not independent confirmations of each other. In the first study, Fitbit promoted the study and donated devices, and the senior author of all three studies co-founded and advises several health-technology companies [S2] [S3] [S4]. Wearable studies of breathing rate during COVID point in a similar direction (breathing rate).

What does an alert actually mean?

A common belief is that a drop in HRV or a device alert means you are getting sick. The prospective study shows why this does not follow. Other respiratory infections, and also events with no infection at all, such as stress, alcohol and travel, triggered alerts too [S3]. So the alert also fired in people who were not infected with the coronavirus, though less often: on average 1.15 alert days per person, compared with 3.42 alert days per person for COVID cases [S3]. An alert says that something deviates from your usual pattern, not what caused it.

A night after drinking is a typical example: HRV falls and resting heart rate rises, with no illness involved (alcohol and HRV). Short sleep, a late meal, hard training and travel move the same numbers (why HRV changes from day to day; HRV and heart rate during sleep). A single low HRV reading does not by itself mean you are stressed or unwell.

What about vaccination?

A vaccine is a planned challenge to the immune system, so it shows what a short immune reaction does to HRV. A systematic review found five observational studies that measured HRV after COVID vaccination [S5]. HRV, mostly RMSSD, changed for a short time and recovered within up to 3 days after vaccination [S5]. In some of the studies the change was larger in women than in men, and in younger than in older people [S5]. The studies were few and their quality was limited, and long-term HRV was not reported [S5]. A short dip in the days after a vaccination fits this pattern. The review does not evaluate vaccination itself, and neither does this page.

What about the time after an illness?

Some people have symptoms for months after an infection, which is called Long COVID. The Stanford group built machine-learning models from heart rate data of 126 people with an acute coronavirus infection [S4]. Adding heart rate features to symptoms gave an improvement of about 5% over symptoms alone in two measures of model accuracy (ROC-AUC and PR-AUC) [S4]. The authors see a possible objective biomarker in this [S4]. It is a single cohort without testing in new people, so the result is a research direction, not a way to recognise Long COVID with a watch.

What does the evidence show?

What we don't know. A review of the field, written by researchers who work with wearable data, says that viral infections can change heart rate, breathing rate, HRV, temperature, activity and sleep before symptoms [S6]. It also says that "the ability of wearable devices to detect viral infections in a real-world setting has yet to be proven" [S6]. Three of its authors were employees of physIQ, a company that analyses wearable data [S6].

By evidence class.

• Emerging. Wearable SDNN tends to be lower when CRP is raised [S1]. A rise in resting heart rate against the personal baseline preceded symptoms in many infections in research cohorts [S2] [S3]. Stress, alcohol and travel trigger the same alerts [S3]. HRV changes after vaccination are short and recover within days [S5]. Heart rate data added little to symptoms for Long COVID in one study [S4].
• Unknown. Whether a consumer wearable can reliably detect a viral infection in everyday life [S6].

What it does not tell you

• A wearable does not diagnose an infection. The reviewers call wearable HRV an exploratory biomarker [S1], and real-world detection has not been proven [S6].
• An alert does not tell you the cause. Infection, stress, alcohol and travel can produce the same change [S3].
• HRV was not the main early signal. The best-studied early warnings relied on resting heart rate, steps and sleep [S2] [S3].
• A normal number does not rule out illness. Many infected people in the studies had no alert [S3].
• It gives no advice on tests, treatment or vaccination.
• Population data are not a prediction for you. The studies describe groups of people; your own pattern may differ.

What can you do with a sharp dip?

If your HRV is clearly below your usual range and your resting heart rate is higher for several days in a row, treat it as a non-specific early signal that the body is under strain. It is a reason to rest and to watch how you feel. If you have symptoms, see a doctor rather than relying on your watch. A single low night is usually not a cause for concern (interpreting HRV).

In ONDA

ONDA builds a personal baseline from nightly values stored in Apple Health — from Apple Watch or another device that syncs heart data there [S7]. The window is 14 days, and ONDA compares each night with your own corridor — the average of your recent nights plus or minus one standard deviation — and flags a night only when it is at least 1.5 standard deviations outside and has changed by a minimum amount: resting heart rate up at least 5 bpm, HRV down at least 15%, breathing rate up at least 2 breaths per minute, with at most one signal every two days. Apple Health records HRV as SDNN, so ONDA's HRV trend is an SDNN trend (your HRV baseline). Such a signal is descriptive: it says that a night is outside your own corridor, not why. An infection, a short night, alcohol or travel can all produce it. ONDA does not identify illness, does not diagnose any condition and does not replace a doctor.