The Apple Heart Study enrolled 419,297 people. It is a screening tool, not a diagnosis.

The Apple Watch's irregular-rhythm notification is a screening prompt, not a diagnosis, and the study Apple ran to support it makes the distinction unusually concrete. In the Apple Heart Study, among the small group of wearers who were notified of an irregular pulse and then wore a medical ECG patch, atrial fibrillation was confirmed in 34% [s1]. The other two-thirds who returned an analysable patch did not have it documented over the recording window.

What the study measured

Apple recruited 419,297 participants through an iPhone app over eight months, all of whom reported not having atrial fibrillation at the outset [s1]. The watch's optical sensor ran an algorithm that looked for irregular pulse intervals; if it flagged possible atrial fibrillation, the participant was offered a telemedicine visit and mailed an ECG patch to wear for up to seven days [s1]. The study's stated goals were to estimate how many notified participants had atrial fibrillation on the patch and the positive predictive value of the pulse-interval alert [s1].

Two numbers frame everything else. Over a median of 117 days of monitoring, 2,161 participants — 0.52% — received a notification [s1]. That is the first thing the marketing tends to omit: for the overwhelming majority of wearers, the feature does nothing, which is the correct behaviour in a young, mostly healthy population where atrial fibrillation is uncommon.

The confirmation rate

Among the 450 participants who returned ECG patches with analysable data — applied on average 13 days after the notification — atrial fibrillation was present in 34% (97.5% confidence interval, 29 to 39) [s1]. In those aged 65 or older, the figure was 35% (97.5% CI, 27 to 43) [s1].

A 34% confirmation rate is easy to misread in either direction. It does not mean the watch was wrong 66% of the time. Atrial fibrillation is intermittent; a patch applied nearly two weeks after the alert can easily miss an arrhythmia that was present when the watch flagged it. When the researchers looked at concordance in real time, the positive predictive value of the alert was higher: 0.84 (95% CI, 0.76 to 0.92) for observing atrial fibrillation on the ECG at the same time as a subsequent notification [s1]. For the tachogram-based signal it was 0.71 (97.5% CI, 0.69 to 0.74) [s1].

The honest summary is that a notification meaningfully raises the odds that atrial fibrillation is present, without confirming it. That is what a screening test does.

The worried-well problem

The study also recorded what people did after being notified. Of 1,376 notified participants who returned a 90-day survey, 57% contacted a health care provider outside the study [s1]. In a screening programme aimed at hundreds of millions of watch wearers, a low notification rate multiplied across that population still generates a large number of clinic visits, many of them prompted by alerts that will not be confirmed as atrial fibrillation. The downstream testing, cost and anxiety are real, and they fall partly on people who turn out not to have the condition.

The Apple Heart Study could not quantify that harm well, because it had no unmonitored control group and enrolled a self-selected, smartphone-owning population that skewed younger than the group in whom atrial fibrillation matters most. It was a single-arm estimate of a device's behaviour, not a trial of whether screening improves health.

What it does not establish

Nothing in the study shows that finding atrial fibrillation this way prevents strokes or extends lives. That question — whether earlier, device-driven detection changes outcomes — is separate, and the evidence for opportunistic screening in the general population has been mixed. A risk-guided approach that screens only the highest-risk patients has performed better in trials than screening everyone, which is the opposite of how a consumer wearable is used.

Apple's own framing has been consistent on the narrow point: the notification and the on-demand ECG are described as tools that may prompt a conversation with a doctor, not as diagnostic devices [s2]. That framing is regulated for a reason. The watch can tell a wearer their pulse looked irregular over some interval. It cannot tell them what to do about it, and a third of the time the finding will hold up on a proper recording — which is worth knowing, and worth not overstating. The feature belongs in the same category as the newer blood-pressure and sleep-apnea notifications: a signal that warrants clinical follow-up, not a verdict.