Can wearable devices detect atrial fibrillation earlier than routine care?

How do wearables detect atrial fibrillation, and how accurate are they?

Wearable devices detect AF using two main methods: photoplethysmography (PPG), which uses light to measure pulse patterns, and single-lead electrocardiography (ECG), which records the heart's electrical activity. Many consumer smartwatches (like Apple Watch, Samsung Galaxy Watch, Fitbit) and dedicated medical wearables (like the Duranta patch or KardiaMobile) use these technologies. In a head-to-head comparison of five popular devices, sensitivity for detecting AF ranged from 58% to 85%, and specificity from 69% to 79% [2]. However, a newer study testing four devices (KardiaMobile 6L, Apple Watch, FibriCheck, Preventicus) found 100% sensitivity and 96-99% specificity [4]. The difference likely reflects improvements in algorithms and study design—the second study used a controlled setting where patients performed measurements in a specific order, which may reduce errors from movement or poor contact.

A ring-type wearable using PPG and deep learning achieved 96.9% accuracy, 99.0% sensitivity, and 94.3% specificity in a study of 100 patients [7]. A medical-grade wearable with an upper-arm PPG sensor and deep neural network reached 96.0% sensitivity and 99.0% specificity over 24 hours of monitoring in hospitalized patients [8]. These results show that accuracy varies by device and setting, but the best-performing wearables can match or exceed the accuracy of standard ECG for detecting AF when used correctly.

Can wearables really find AF earlier than a doctor's visit?

Yes, because AF is often intermittent (paroxysmal) and symptomless, so a single ECG at a clinic visit can easily miss it. Wearables that monitor continuously or repeatedly over days or weeks have a much better chance of catching brief episodes. In a prospective study of 241 stroke patients with no prior AF diagnosis, a wearable patch detected new AF in 8.7% of patients during 7 days of monitoring [1]. These patients would have been missed by routine care, which typically involves a single ECG or short monitoring. Another large trial (PATCH-AF) is currently testing yearly 7-day wearable monitoring in high-risk primary care patients aged 65+ and expects results in 2025 [9]. The rationale is that repeated prolonged monitoring will detect AF that standard care misses, potentially preventing strokes.

A deep-learning model called WARN can predict the transition from normal rhythm to AF an average of 30.8 minutes before it starts, using R-R interval data from wearables [6]. This means wearables could not only detect AF when it happens but also warn users before an episode begins, which routine care cannot do. While this is still early-stage research, it points to a future where wearables provide real-time early warning.

What are the limitations? When might wearables not work well?

Wearables are not perfect. A major issue is 'inconclusive' readings—when the device cannot determine the rhythm. In one study, 17-26% of tracings from five smart devices were inconclusive, requiring manual review by a clinician [2]. Another study found that 7-15% of measurements had insufficient quality [4]. This means that relying solely on a wearable's automated diagnosis could lead to missed AF or false alarms. Frequent premature beats (extra heartbeats) can also fool PPG-based algorithms, causing false positives. In the DoubleCheck-AF study, all seven false-positive PPG readings came from patients with frequent premature contractions, but adding a 6-lead ECG from the same device corrected the diagnosis in most cases [3].

Another limitation is that most studies have been done in controlled settings (clinics or hospitals) or with motivated volunteers. Real-world performance may be lower due to movement artifacts, poor device fit, or user non-compliance. A meta-analysis of nine studies (1,581 participants) found that wrist-worn devices had high sensitivity (96%) but specificity varied significantly between brands (from 81% to 99.6%), meaning some devices generate more false alarms [5]. False positives can lead to unnecessary worry, extra tests, and healthcare costs. Finally, these devices are not a substitute for medical diagnosis—any alert should be confirmed by a doctor with a standard 12-lead ECG.

About These Sources

This answer is built on 9 peer-reviewed studies — published from 2020 to 2025, 3 from 2024 or later, 2 in Q1 journals, collectively cited 209 times — selected as the most relevant from 12 studies that passed quality screening, drawn from 71 papers retrieved from a database of over 500 million.

Sources used in this answer

Related research questions