Wearable ECGs and Atrial Fibrillation: Can a Watch Detect a Dangerous Heart Rhythm?

In 2018, a 46-year-old construction worker in Arizona received a notification on his Apple Watch Series 4. The device, which he wore to track his steps and sleep, had detected an irregular heart rhythm and advised him to see a doctor. He had no symptoms. He felt completely fine. His cardiologist subsequently diagnosed paroxysmal atrial fibrillation and started him on anticoagulation therapy. Six months later, his electrophysiologist noted that, statistically, he had likely avoided a stroke.

Stories like this are no longer anecdotal curiosities. They are the clinical rationale behind a regulatory and technological revolution in cardiac monitoring. The Apple Watch, the AliveCor KardiaMobile, the Withings ScanWatch, and the Samsung Galaxy Watch ECG have collectively shifted how cardiologists think about rhythm surveillance: from episodic, hospital-triggered snapshots to continuous, passive, patient-initiated streams of data. The question is no longer whether a consumer wearable can record an electrocardiogram. It can. The question is what that recording can and cannot tell a clinician, and how the healthcare system should respond when a watch sends a patient walking into a clinic clutching a six-second rhythm strip.

This article unpacks the physiology of atrial fibrillation, the engineering of wearable ECG sensors, the landmark clinical evidence from the Apple Heart Study and the HEARTLINE trial, the regulatory path these devices have navigated, and the genuine limitations that still constrain what a wrist-worn sensor can diagnose. It also looks at what comes next: multi-lead wearables, AI arrhythmia classifiers, and the emerging question of whether continuous cardiac monitoring should be standard of care for anyone over 65. As explored in our broader coverage of biometric data and early disease detection, the wearable ECG story is a microcosm of a much larger shift in preventive medicine.

What Is Atrial Fibrillation and Why Does It Kill?

The heart is an electrical organ. Its mechanical pumping function depends entirely on coordinated electrical signals that originate in the sinoatrial (SA) node, propagate through the atria, pause briefly at the atrioventricular (AV) node, and then descend into the ventricles via the His-Purkinje system. This choreography produces the familiar P-QRS-T waveform on a standard ECG. The P wave represents atrial depolarisation. The QRS complex is ventricular depolarisation. The T wave is ventricular repolarisation. In a healthy heart, this sequence repeats with metronomic regularity, sixty to one hundred times per minute at rest.

In atrial fibrillation, the electrical architecture of the atria breaks down. Instead of a single coordinated wavefront originating from the SA node, hundreds of micro re-entrant circuits fire simultaneously and chaotically, generating electrical noise at 350-600 impulses per minute. The atrial myocardium quivers rather than contracts. Most of those impulses are blocked by the AV node, which acts as a gatekeeper, but enough get through to drive the ventricles at an irregular and often rapid rate. On an ECG, the P wave disappears, replaced by a chaotic baseline, and the R-R interval (the distance between consecutive heartbeats) becomes irregularly irregular.

Paroxysmal AFib, which terminates spontaneously within seven days, is particularly difficult to catch by traditional means. A standard 12-lead ECG records ten seconds of cardiac activity. A 24-hour Holter monitor captures one day. Paroxysmal AFib can come and go in minutes, without symptoms, days between episodes. A patient can have AFib-related stroke as their first presenting symptom. This diagnostic gap is precisely the clinical problem that continuous wearable monitoring is designed to solve. Approximately 33 million people worldwide have AFib, and the condition increases stroke risk by four to five times. Undiagnosed AFib is thought to explain a substantial fraction of cryptogenic (cause-unknown) strokes, which account for roughly a third of all ischemic strokes in younger adults.

How Wearable ECG Technology Actually Works

There are two fundamentally different sensing modalities in consumer cardiac wearables, and understanding the distinction is essential for interpreting what a device can and cannot detect: photoplethysmography (PPG) and single-lead electrocardiography (ECG). They are complementary rather than equivalent, and many modern devices use both.

Photoplethysmography: Detecting Pulse Irregularity with Light

PPG is the older and simpler technology. It works by shining a green (and sometimes infrared) LED into the skin on the underside of the wrist and measuring how much light is absorbed by the blood vessels below. Blood volume in capillaries increases with each heartbeat and decreases between beats, so the sensor captures the pulse waveform. From this signal, the device extracts heart rate and, crucially, beat-to-beat variability.

The irregular R-R intervals that characterise AFib produce a distinctive pattern in the PPG signal: the pulse intervals are chaotic rather than regularly spaced. Machine learning classifiers, trained on large datasets of confirmed AFib and normal sinus rhythm, can detect this irregularity with meaningful accuracy. The Apple Watch uses this principle for its irregular rhythm notification feature: it passively monitors PPG data and alerts the user if it detects an irregular rhythm consistent with AFib over five minutes of analysis. PPG-based detection is powerful for passive surveillance precisely because the sensor runs continuously without user action, but it cannot produce an actual ECG waveform. It detects the hemodynamic consequence of the arrhythmia (irregular pulse), not the electrical event itself.

Single-Lead Electrocardiography: Recording the Electrical Signal Directly

A conventional hospital 12-lead ECG places electrodes at ten standardised locations on the body (four limb leads, six chest leads) and records cardiac electrical activity from twelve different angles simultaneously. This provides rich spatial information about the heart: which wall is ischemic, whether a bundle branch is blocked, whether ventricular hypertrophy is present, and dozens of other diagnostically relevant findings.

Wearable ECGs take a more austere approach: they record a single electrical vector between two electrodes. In the Apple Watch ECG app (Series 4 and later), the user places a finger on the Digital Crown, which contains an electrode. A second electrode is built into the underside of the watch case against the wrist. The circuit is closed across the chest, creating a lead roughly equivalent to Lead I on a standard ECG (measuring electrical potential between the left arm and right arm). The resulting trace is a genuine ECG rhythm strip: it shows P waves, QRS complexes, and T waves, and a cardiologist can read it. For 30 seconds, the watch records this signal and classifies the rhythm as sinus rhythm, AFib, low heart rate, high heart rate, or inconclusive.

The AliveCor KardiaMobile uses a similar principle. The original KardiaMobile has two electrodes; the user places both thumbs on the device (or places it on a knee and uses a finger). The six-lead KardiaMobile 6L adds four chest electrode positions via a smartphone accessory, enabling a six-lead recording that approaches the informational richness of a standard clinical ECG, though it remains a subset of the full 12-lead. This expanded lead count allows the detection of additional rhythm abnormalities beyond AFib, which is why the KardiaMobile 6L has FDA clearance for six rhythm classifications rather than just AFib and normal sinus rhythm.

The Withings ScanWatch takes an integrated approach, embedding both PPG optical sensors and dry ECG electrodes into a traditional analogue watch design. The Samsung Galaxy Watch series, available in several markets, similarly offers on-demand single-lead ECG recording via the Galaxy Health app and has received regulatory clearances in multiple jurisdictions including South Korea and the European Union.

The Apple Heart Study: Mass-Scale Evidence

The landmark clinical evidence for wearable AFib detection came from the Apple Heart Study, a decentralised, app-based prospective cohort study conducted between November 2017 and July 2018 in partnership with Stanford Medicine. The study enrolled 419,297 participants through an iPhone app, making it one of the largest cardiovascular studies ever conducted and a template for how digital health research can scale.

The methodology was elegant in its simplicity. Participants wore their Apple Watch and the app passively monitored their PPG signal for irregular pulse. Participants who received an irregular rhythm notification were sent a telemedicine visit and mailed a seven-day ECG patch monitor to wear continuously. The patch provided ground-truth rhythm data. Of the 419,297 participants, 2,161 (0.52%) received irregular pulse notifications. Among the subset who returned patch data and had a notification concurrent with patch monitoring, the positive predictive value for AFib was 84%. Among participants over 65, the positive predictive value rose above 90%.

The study, published in the New England Journal of Medicine in November 2019, was notable for what it demonstrated and what it was careful not to claim. The authors emphasised that the 0.52% notification rate meant the vast majority of Watch wearers did not receive an alert, reflecting the algorithm's conservative threshold for flagging irregularity. The tool was designed for specificity as much as sensitivity: a high false positive rate in a healthy consumer population would generate enormous unnecessary clinical burden. The study also documented that of those who were notified and subsequently confirmed to have AFib, 57% were not previously aware of their diagnosis.

AliveCor KardiaMobile: The Cardiologist's Consumer Device

While Apple has the largest installed base and the most prominent public studies, AliveCor's KardiaMobile devices have arguably deeper clinical penetration. Founded in 2010, AliveCor pioneered the consumer ECG category and has an extensive body of peer-reviewed evidence spanning more than a decade. Cardiologists who use consumer ECG tools in their practice are, anecdotally, more likely to reference KardiaMobile than Apple Watch, partly because AliveCor's core product is a dedicated medical device, not a general-purpose smartwatch with ECG as one of many features.

The KardiaMobile 6L received FDA De Novo clearance in 2019 for detection of six rhythm classifications: normal sinus rhythm, AFib, bradycardia (heart rate below 50 bpm), tachycardia (heart rate above 100 bpm), premature atrial contractions (PACs), and premature ventricular contractions (PVCs). The six-lead configuration also received an additional clearance for determining PR interval, QT interval, QRS duration, and R-R interval: measurements that approach the basic quantitative outputs of a clinical ECG machine. For AF detection specifically, published validation studies report sensitivity of 93-98% and specificity of 84-99% against 12-lead ECG, varying by study design and patient population.

AliveCor has also developed KardiaCare, a subscription service that uses AI (backed by a team of board-certified cardiologists for overread) to provide rhythm analysis reports. The KardiaMobile is now used in clinical research, in cardiac rehabilitation programs, and as an adjunct monitoring tool in electrophysiology clinics. It represents the most validated consumer ECG platform in terms of regulatory clearance breadth and clinical literature depth.

The FDA Regulatory Pathway for Wearable ECGs

Understanding how these devices reached market requires a brief detour into FDA medical device regulation. The FDA classifies medical devices into three classes based on risk. Class I devices (low risk) require only general controls. Class II devices (moderate risk) require premarket notification under the 510(k) pathway, which demonstrates that the new device is substantially equivalent to a legally marketed predicate device. Class III devices (high risk) require the most stringent premarket approval (PMA), including clinical trials.

Wearable ECG devices that produce rhythm strips have generally navigated the De Novo pathway, which is used when a device is novel (no clear predicate exists) but low to moderate risk. De Novo classification simultaneously creates a new regulatory classification and grants clearance. This is the pathway AliveCor used for KardiaMobile and that Apple used for its ECG app (Apple Watch Series 4 ECG received De Novo clearance in September 2018, the same day the device was announced). Once De Novo clearance is established, subsequent similar devices can use the now-established classification as a predicate for 510(k) clearance, which is why additional wearable ECG products have reached the market more rapidly.

The FDA has been notably progressive in this space. The agency's Digital Health Center of Excellence has published guidance documents specifically addressing software as a medical device (SaMD) and AI-based medical devices, providing a framework for manufacturers to navigate clearance without defaulting to the slow, expensive PMA pathway. This regulatory posture has enabled the rapid commercialisation of wearable cardiac monitoring in a way that would have been difficult under older regulatory interpretations.

What Wearable ECGs Cannot Do: Critical Limitations

The clinical enthusiasm for wearable ECGs must be tempered by an honest accounting of their limitations. These devices are sensitive and specific for AFib. They are not general-purpose cardiac diagnostic tools, and understanding the gap between what patients believe them capable of and what they can actually detect is important for appropriate clinical framing.

Cannot Detect Myocardial Infarction

Acute myocardial infarction (heart attack) diagnosis requires ST-segment elevation or depression, which indicates ischemia in the ventricular myocardium. Detecting this requires multiple ECG leads viewing the heart from different spatial angles simultaneously. A single-lead ECG has one spatial perspective. It cannot reliably assess ST segments across the inferior, lateral, and anterior walls of the left ventricle. A patient experiencing a posterior STEMI (ST-elevation myocardial infarction), which classically shows ST depression in the anterior leads, would likely generate a normal or inconclusive result on a single-lead wearable ECG. Users who experience chest pain, diaphoresis, or jaw and arm pain should call emergency services, not reach for their Apple Watch.

Limited Arrhythmia Classification

The Apple Watch ECG currently classifies rhythm into five categories: sinus rhythm, AFib, bradycardia, tachycardia, and inconclusive. It does not detect ventricular tachycardia, ventricular fibrillation (though these would likely render the recording uninterpretable), second or third degree heart block, Wolff-Parkinson-White syndrome, or most ST-T wave abnormalities. The KardiaMobile 6L's six-rhythm classification is more comprehensive but still covers a small subset of the dozens of rhythms a trained electrophysiologist can identify on a 12-lead ECG. Reassurance from a "sinus rhythm" reading does not rule out serious pathology.

Signal Quality and Motion Artifact

Single-lead ECG recordings are susceptible to motion artifact, poor skin contact, tremor, and electrical interference. A significant proportion of recordings (15-25% in some studies) are classified as inconclusive, particularly in older adults with tremor or patients with poor peripheral perfusion. The PPG-based irregular rhythm notification has higher noise sensitivity due to motion: rigorous activity, for example, can produce pulse interval variability that is difficult to distinguish from AFib without additional processing. Both modalities perform better when the user is at rest.

The Population Health Consideration

Mass screening with any test carries the risk of false positives generating unnecessary downstream clinical workup and patient anxiety. The Apple Heart Study's 84% positive predictive value means that approximately one in six users who received an irregular rhythm notification did not have AFib. In a population of millions of Watch wearers, even a low false positive rate generates a large absolute number of inappropriate cardiology referrals. How the healthcare system absorbs that volume is a legitimate health economics and health systems question that has not been fully resolved.

These limitations do not undercut the technology's value. They contextualise it. Wearable ECGs excel at passive, continuous rhythm surveillance in a population that would otherwise receive none. That is a clinically meaningful capability that has genuine downstream benefit, particularly for the early detection of paroxysmal AFib. The key is integrating these tools appropriately into clinical workflows rather than positioning them as replacements for comprehensive cardiac evaluation. This is a theme central to the broader discussion of AI in chronic disease management: the technology augments the clinician, it does not replace the clinical encounter.

Clinical Workflow: What Happens When the Watch Flags AFib?

When a patient presents to a primary care physician or cardiologist with a wearable ECG notification or recording, the clinical pathway that follows is increasingly standardised but still evolving. Most cardiology societies, including the American Heart Association and the European Society of Cardiology, have issued guidance on integrating wearable cardiac data into clinical practice, though the specifics vary by institution.

The first step is confirmation. A single wearable ECG recording of AFib is considered a screen-positive result, not a definitive diagnosis. The physician will typically perform a 12-lead ECG in the office. If AFib is present on the 12-lead, the diagnosis is confirmed and the physician proceeds to standard assessment: determining AFib type (paroxysmal, persistent, long-standing persistent, or permanent), evaluating underlying causes (thyroid disease, structural heart disease, valvular pathology, sleep apnea, hypertension, alcohol use), calculating stroke risk using the CHA2DS2-VASc score, assessing bleeding risk using the HAS-BLED score, and deciding on rate control, rhythm control, and anticoagulation strategy.

If the 12-lead is in sinus rhythm at the time of the visit, the clinician faces a judgment call about the wearable recording. A high-quality, clearly readable wearable ECG trace showing characteristic AFib features (absent P waves, irregularly irregular R-R intervals) is increasingly accepted by cardiologists as sufficient for diagnosis when combined with a consistent clinical history. The AHA's 2019 AFib guidelines noted that rhythm documentation does not require a 12-lead ECG if a single-lead ECG strip clearly shows AFib. Many electrophysiologists are now comfortable initiating anticoagulation on the basis of a wearable ECG trace plus symptoms, particularly in patients with the Apple Watch Series 4 or later (FDA-cleared device) or KardiaMobile 6L.

For patients with paroxysmal AFib that is not captured during the office visit, extended rhythm monitoring may be ordered: a 24-48 hour ambulatory Holter monitor, a 7-30 day continuous patch monitor (Zio Patch by iRhythm is the most widely used), or an implantable loop recorder for cases where long-term surveillance is needed. Interestingly, cardiologists are increasingly recommending that patients with wearable-detected possible AFib continue to use their device to log rhythm during symptoms and bring a library of recordings to follow-up visits, effectively turning the patient into a contributor to their own longitudinal cardiac record.

Future Directions: Toward Continuous Multi-Lead Wearables

The single-lead limitation of current wearable ECGs is the most significant technical constraint, and it is the primary focus of next-generation development. Several research groups and companies are pursuing wearable architectures that could approach the diagnostic richness of a 12-lead ECG without requiring the patient to sit still with electrodes attached to specific anatomical landmarks.

One approach involves textile electrodes embedded in garments: shirts, bras, or vests that position electrodes at multiple standard lead locations and transmit data wirelessly. Several prototypes have demonstrated near-12-lead quality in ambulatory recordings, though commercial products with regulatory clearance remain limited. BioSerenity's Neuronaute garment (primarily designed for EEG, with cardiac capability) and Nuubo's nECG Holter shirt have demonstrated feasibility, and the regulatory pathways are being explored with the FDA.

Another direction is AI-enhanced single-lead interpretation. Deep learning models trained on millions of ECG recordings have demonstrated that information not visible to a human reader can be extracted from a single-lead trace. A 2019 study from Mayo Clinic published in Nature Medicine showed that an AI trained on standard 12-lead ECGs could detect low ejection fraction (a marker of heart failure) from a single-lead ECG with an area under the ROC curve of 0.93. Subsequent work has explored AI detection of electrolyte abnormalities, valvular disease, and even age and sex from a single lead. If such models can be validated and cleared, the single-lead wearable ECG becomes a significantly more powerful tool despite unchanged hardware.

The regulatory and clinical integration questions are perhaps as consequential as the technical ones. Should AFib screening with wearable ECGs be recommended for all adults over 65? The United States Preventive Services Task Force (USPSTF) does not yet endorse routine AFib screening (as of the most recent guidance), citing insufficient evidence that early detection improves outcomes relative to harms. However, the Apple Heart Study and HEARTLINE data, along with the growing body of outcome data showing that anticoagulation reduces AFib stroke risk by 60-70%, are generating pressure to revisit that position. The European Society of Cardiology's 2020 AFib guidelines took a more permissive stance, recommending opportunistic screening (pulse check or ECG rhythm strip) in patients over 65. The integration of wearable ECGs into clinical decision support frameworks, as discussed in our overview of clinical decision support systems, will be a central part of how this evidence gets translated into practice.

The trajectory is clear. The wearable ECG is not a novelty item or a wellness gadget with medical aspirations. It is a validated, FDA-cleared rhythm monitoring tool that has demonstrated its ability to detect the most common serious cardiac arrhythmia in an outpatient, consumer context. Its limitations are real and clinically important. Its capabilities are also real and clinically important. The task now is not to prove that the technology works: the Apple Heart Study did that. The task is to build the clinical infrastructure to respond appropriately when hundreds of millions of people are wearing devices that can tell them their heart is beating wrong.