Can a wearable catch illness before you feel it? The evidence is intriguing and shaky

One of the most striking claims made for consumer wearables is that they can detect illness before a person notices any symptoms — a rise in resting heart rate or skin temperature overnight signalling an infection or a flare-up days in advance. The published evidence contains some genuinely eye-catching results in that direction, but it also comes with an unusually large asterisk: most of the studies carry a high risk of bias, rely on algorithms no one outside the company can inspect, and have not shown that the early warning changes what happens to a patient [s1][s2].

The eye-catching results

A systematic review of smart rings in clinical medicine, spanning 107 studies and roughly 100,000 participants, catalogued the predictive claims [s1]. Among them: detection of COVID-19 an average of 2.75 days before symptoms, with 82% sensitivity; prediction of inflammatory-bowel-disease flares up to seven weeks early, with 72% accuracy; and detection of bipolar mood episodes three to seven days ahead, with 79% sensitivity [s1]. The underlying signals — heart rate, heart-rate variability, temperature — were themselves measured accurately in these studies, with heart-rate agreement very high (r² = 0.996) [s1].

Taken at face value, those numbers describe something valuable: a passive monitor that notices the body changing before the mind does. The physiological logic is sound — infection and inflammation do raise resting heart rate and shift temperature and autonomic balance before overt symptoms.

Why to read them cautiously

The same review that lists these results is candid about their quality. Sixty-five percent of the included studies carried a moderate-to-high risk of bias [s1]. Eighty-nine percent relied on proprietary algorithms that cannot be independently reproduced, and only 35% reported the diversity of their participants — so it is often unclear whether a result generalises beyond the sample that produced it [s1]. Adherence also decayed steeply, from 80% of users still wearing the device at three months to 43% at twelve, which matters for a tool whose value depends on continuous, long-term wear [s1].

A separate umbrella review sharpens the concern about real-world usefulness. Synthesising 42 reviews across more than 30 brands, it found highly variable review quality, wide heterogeneity in device performance, and limited evidence for clinical utility — with only two biometrics, heart rate and atrial-fibrillation detection, enjoying comparably stronger support [s2]. In other words, the signals that feed an early-warning algorithm are well measured, but the leap from "the number moved" to "this person is getting sick, and acting on it helped" is where the evidence thins [s2].

The false-alarm problem

An early-warning system's usefulness turns on how often it cries wolf. A resting-heart-rate rise can mean a brewing infection — or a late meal, alcohol, a hard workout, a hot bedroom, or poor sleep. Without validation in the messy conditions of everyday life, a sensitive detector generates alerts that are mostly false, and the burden of chasing them down can outweigh the benefit of the occasional true catch. This is the recurring theme across the wearable literature we examined in digital-biomarker research: strong measurement, weak proof that the derived alert should change behaviour.

Detection is not the same as benefit

Even a true early warning only helps if there is something useful to do with it. For COVID-19, an alert 2.75 days before symptoms could in principle prompt earlier testing and isolation — a plausible benefit during a period of high transmission [s1]. For a chronic condition like inflammatory bowel disease, a flare predicted seven weeks out is more ambiguous: it is a long lead time in which the prediction might not come true, and acting on it — escalating treatment pre-emptively — carries its own risks [s1]. None of the reviewed studies were designed to show that the early alert improved outcomes, which is the question that would justify wearing the device for that purpose [s1][s2]. Demonstrating a signal moves before symptoms is a first step; demonstrating that catching it changes the course of the illness is the one that has not been taken.

The adherence data underline the practical gap. A tool that predicts flares weeks ahead is only useful if worn continuously for months and years, yet wear fell to 43% by twelve months in the pooled studies [s1]. An early-warning system most people have taken off is not an early-warning system.

What it means

The honest summary is that pre-symptomatic illness detection is a real and active area of research with some promising early findings, not a proven consumer feature [s1][s2]. For someone wearing a ring or watch, a sustained, unexplained rise in resting heart rate or temperature over several days is a reasonable prompt to rest and pay attention — the same common-sense reading of the underlying HRV and heart-rate signals — but it is not a diagnosis, and a normal reading is no guarantee of health. The technology may earn a clinical role as the evidence matures and the algorithms open up; on current evidence, the claim is ahead of the proof.