Resting heart rate, heart rate variability and what wearables can actually tell you

Wearable devices have put continuous physiological monitoring on a great many wrists, including a large number belonging to people over sixty-five. The data is real. The interpretation is where things go wrong.

Resting heart rate

Of everything a consumer device measures, this is among the most reliable and the most useful. Optical sensors measure heart rate accurately at rest, and the value is in the trend rather than the number.

Normal resting heart rate spans roughly 60 to 100 beats per minute, with fitter people at the lower end. Population studies consistently associate higher resting heart rate with increased cardiovascular and all-cause mortality, independent of fitness — though causality is not established and part of the association reflects underlying disease.

The practically useful signal is a change from your own baseline. A sustained rise of five to ten beats per minute over days, without an obvious explanation, commonly reflects infection, dehydration, poor sleep, alcohol, overtraining, anaemia, thyroid dysfunction, or the early stages of an illness you have not yet noticed. Several studies found consumer wearables detected physiological changes preceding symptomatic viral infection.

A sustained fall can reflect improving fitness, or a new medication such as a beta blocker, or something that needs attention if accompanied by dizziness or fainting.

Atrial fibrillation detection

This is where wearables have made a genuine clinical contribution, and also where the questions get complicated.

Irregular rhythm notification algorithms detect the irregular pulse pattern of AF, and single-lead ECG functions on several devices can record a trace. The Apple Heart Study and similar work demonstrated that these can identify previously unknown AF in a meaningful number of users.

Since AF is frequently asymptomatic and substantially raises stroke risk, finding it matters. The complications are twofold. False positives are common, particularly with movement artefact, ectopic beats, or poor contact — and they generate anxiety and investigation. And the clinical significance of brief, screen-detected episodes in people with no symptoms is genuinely unsettled: it is not established that very short episodes carry the same stroke risk as clinically diagnosed AF, or that anticoagulating them improves outcomes.

A notification is a reason for a proper assessment, not a diagnosis and certainly not a reason to start medication.

Heart rate variability

HRV measures beat-to-beat variation in the interval between heartbeats, reflecting the balance between sympathetic and parasympathetic activity. Higher variability generally indicates better autonomic function.

It is a legitimate research measure, and consumer implementations are considerably cruder than the ones used in the literature — different algorithms, different measurement windows, different postures. Absolute values are not comparable between devices or between people, and HRV varies substantially with age, sex, posture, breathing rate and time of day.

Used as a personal trend it has some value: a marked drop from your own baseline often precedes noticing that you are unwell or overreached. Used as an absolute score to be optimised, it is mostly a source of anxiety.

What the devices do badly

Blood pressure. Most consumer wearables that claim it are not validated to clinical standards, and several require calibration against a cuff, which rather defeats the purpose. Use a validated upper-arm cuff.

Blood oxygen. Wrist pulse oximetry is considerably less accurate than fingertip devices, and both are less accurate at lower saturations and on darker skin — a limitation that received substantial attention during the pandemic. Not adequate for clinical decisions.

Sleep staging. Devices infer sleep stages from movement and heart rate. Compared with polysomnography, they are reasonable at distinguishing sleep from wake and poor at staging. Total sleep time is roughly useful; the REM and deep sleep breakdown is close to fiction, and people lose sleep worrying about it — a phenomenon researchers have named orthosomnia.

Step counts are reasonably accurate at normal walking speeds and less so at the slow, shuffling gait of some older adults, where they systematically undercount. Devices worn on the wrist also miss steps taken while pushing a trolley or using a frame.

Calorie estimates are unreliable, frequently by a wide margin.

Where they help most

Honestly, the largest documented benefit of wearables is behavioural rather than diagnostic. Meta-analyses of activity trackers find modest but real increases in physical activity — on the order of a thousand extra steps a day — and modest weight loss. For an older adult, an extra thousand steps daily is a genuinely worthwhile change.

Fall detection, available on several devices, has plausible value for people living alone, though real-world sensitivity and false alarm rates are imperfect.

The reasonable posture is to treat a wearable as a prompt and a trend-tracker rather than a diagnostic instrument — and to take any alarming reading to someone with a proper machine before drawing conclusions from it.