Smartwatch AF Alerts in the ED: What the Notification Proves

At 0300 a 68-year-old long-haul trucker with hypertension and type 2 diabetes checks in because his watch woke him with a possible atrial fibrillation notification. He feels fine now. His heart rate is 76, his blood pressure 142/84, and the triage 12-lead shows sinus rhythm. On his phone is a single-lead tracing he recorded at 0140, still labeled by the app.

The notification on its own proves nothing. The tracing can, once you read it yourself, and what it shows decides whether this is a stroke-prevention visit or a referral for monitoring.

What the Watch Actually Measured

Consumer watches can do two different things. The notifications tested in the large screening studies came from photoplethysmography (PPG), an optical sensor that infers rhythm from the timing of pulse waves at the wrist [2,3]. The 2023 ACC/AHA/ACCP/HRS atrial fibrillation guideline states that PPG may prompt a person to obtain an electrocardiographic tracing but is not reliable enough to establish an AF diagnosis [1].

Some watches can also record a single-lead ECG on demand. That is an electrical tracing, and the guideline treats it as one. Its first recommendation on rhythm monitoring is that, in a person without known AF, the initial diagnosis be made by a clinician visually interpreting the electrocardiographic signals, regardless of the device that recorded them [1]. The device's automated label does not count.

So the first question is practical: did the watch record light or electricity, and can you see the tracing?

How Often the Notification Is Right

Two large remote studies, both run with the device makers, set expectations.

The Apple Heart Study enrolled 419,297 people who reported no AF [2]. Over a median of 117 days, 2,161 (0.52%) received an irregular pulse notification. Among the 450 who returned an analyzable ECG patch, applied on average 13 days after the notification, AF was present in 34%. When a new notification fired while the patch was recording, AF was on the ECG at the same moment 84% of the time [2].

The Fitbit Heart Study enrolled 455,699 adults, median age 47 and 71% female [3]. Irregular rhythm detections occurred in 1% overall and in 4% of participants aged 65 or older. Among the 1,057 notified participants with an analyzable follow-up patch, AF was present in 32.2%, but when a detection fired during patch wear, 221 of 225 were concurrent AF, a positive predictive value of 98.2%, and 97.0% in those aged 65 or older [3].

Both findings hold at once because AF is often paroxysmal. A notification can be right on Tuesday and the patch can be clean the following week. A sinus 12-lead at 0300 does not mean the 0140 alert was false. It means you have not seen the rhythm yet.

What ED Studies Show

Screening cohorts are not your waiting room. People who come to the ED after an alert are a filtered group.

In a retrospective single-center study of 424 ED patients who presented after a smartwatch irregular rhythm alert, 44.6% needed pharmacologic or electrical treatment [4]. Here most alerts (60.8%) came from the watch's ECG function rather than PPG. Palpitations were the most common complaint (56.1%) and 8.5% had no symptoms. Agreement between the watch's call and the final rhythm diagnosis from the department 12-lead was poor for both AF and supraventricular tachycardia, and no clinical factor predicted a false AF alert [4]. The reference ECG was taken in the department, after the alert, so some disagreement reflects rhythms that had already changed.

In 721 patients in an ED's critical care areas, a watch lead-I recording performed similarly to the standard ECG's lead I in separating normal from abnormal rhythm, and heart rates agreed well on average (mean difference 0.81 ± 6.12 beats per minute). The two differed significantly when identifying abnormal findings within lead I [5]. The comparison was against lead I of the standard ECG, not the full 12-lead, but the point stands: a single lead can show the rhythm and cannot do the 12-lead's job.

Practice varies. In a survey built on hypothetical asymptomatic patients with Apple Watch alerts, 95 of 636 invited physicians answered (14.9%), 25 of them emergency physicians [6]. Emergency physicians chose rhythm monitoring in 46% of cases, against 84% for primary care and 94% for cardiology, and chose BNP more often (20% against 8% and 2%) [6].

Anticoagulation: Whose Evidence Applies

If the 12-lead shows AF, this is clinical AF and the guideline applies as written. Anticoagulation is recommended when the estimated annual thromboembolic risk is 2% or higher, for example a CHA2DS2-VASc score of 2 or more in men or 3 or more in women, and is reasonable at 1% to under 2% [1]. The trucker scores 3. The 2024 ESC guideline scores the same risk with CHA2DS2-VA, which drops sex as a criterion; for this man it gives the same answer, and it offers only a weak recommendation for anticoagulating device-detected subclinical AF [10].

If his own watch strip, read by you, shows AF, the guideline's diagnostic rule is met by a clinician's read of the tracing [1]. The open question is how much a brief episode found only on a consumer device matters. The closest randomized evidence comes from patients with implanted pacemakers and defibrillators. ARTESIA randomized 4,012 patients with device-detected episodes of 6 minutes to 24 hours to apixaban or aspirin: stroke or systemic embolism fell from 1.24% to 0.78% per patient-year, while major bleeding rose from 0.94% to 1.71% [7]. NOAH-AFNET 6 randomized 2,536 patients aged 65 or older with episodes of at least 6 minutes and at least one more stroke risk factor to edoxaban or placebo and stopped early. The composite of cardiovascular death, stroke or embolism did not differ significantly, stroke ran near 1% per patient-year in both arms, and death or major bleeding rose [8]. Neither trial studied wearable-detected AF, and a PubMed search for this post found no randomized trial of anticoagulation for AF seen only on a watch.

If the strip is uninterpretable, missing or sinus, and the 12-lead is sinus, the notification alone does not meet the diagnostic standard [1]. That patient needs a monitor and a follow-up appointment. In a 2026 randomized trial at two Dutch centers, 437 patients aged 65 or older with elevated stroke risk were assigned to six months of smartwatch screening (optical notifications plus single-lead ECGs, each tracing read by a remote team within 24 hours) or usual care; new AF was found in 9.6% against 2.3% [9]. Its primary outcome was AF detection; the published abstract reports no stroke outcome.

What to Do on the Shift

Get the 12-lead. Ask to see every stored tracing, read each one as you would a referral strip, and record your own interpretation in the chart. Treat the patient's rate, symptoms and hemodynamics the way you would without the watch. For the patient in sinus rhythm, the useful disposition is ambulatory ECG monitoring with a named clinician who will act on the result. For patients with known AF, the guideline calls a consumer ECG device with a good tracing reasonable for detecting recurrences [1], so the watch has a place after discharge too.

The software that fired this alert is a narrow algorithm, and it carries the same lesson as any AI ECG second reader: it starts the question, a clinician answers it. The automation bias problem applies here too: the app's label arrives first and sounds certain, which is exactly when a clinician is most likely to adopt it unread.

Key Takeaways

• PPG notifications are not diagnostic; the 2023 guideline requires a clinician's own read of an ECG tracing, and a watch's single-lead strip can qualify [1].
• About one in three notified people who returned a usable patch showed AF on it (applied 13 days later on average in the Apple study), yet concurrent notifications matched AF 84% to 98% of the time [2,3]. A sinus 12-lead does not clear the alert.
• In one ED cohort of 424 alert-driven visits, 44.6% needed drug or electrical treatment, and the watch's label agreed poorly with the department ECG [4].
• AF clearly shown on an ECG tracing follows CHA2DS2-VASc thresholds (2 or more in men, 3 or more in women) [1], though the weight of a short, watch-only episode is less certain; the short-episode trials (ARTESIA, NOAH-AFNET 6) enrolled patients with implanted pacemakers or defibrillators [7,8].

FAQ

Is a smartwatch AFib alert accurate?

It depends on what fired. In the two largest studies, a notification that fired while a reference ECG was recording matched AF 84% of the time with Apple and 98.2% with Fitbit [2,3]. Among notified people who returned a usable patch, only about a third had AF on it, partly because many episodes are paroxysmal and partly because some notifications are false [2,3]. The guideline does not accept a PPG notification as a diagnosis [1].

Can AF be diagnosed from an Apple Watch ECG?

A clinician can diagnose AF from a single-lead watch tracing they read themselves; the guideline's rule is a clinician's visual read of the electrocardiographic signal, whatever the device [1]. The app's automated label is not the diagnosis. Document your own interpretation and confirm on a 12-lead when you can.

Should a patient with a smartwatch AFib alert start a blood thinner?

Only after AF is confirmed on an ECG tracing a clinician has read, and then by the same CHA2DS2-VASc thresholds used for any clinical AF, for a tracing that clearly shows AF [1]. How much weight a short, watch-only episode deserves is less certain. A notification with no confirming tracing is a reason to arrange rhythm monitoring. The trials of anticoagulation for brief device-detected episodes used implanted devices and showed more bleeding with treatment [7,8].

Dr. Chet's Take

I have practiced emergency medicine for more than 25 years and work at a Level I academic trauma center, and this article gets the order of operations right. The notification is a reason to look. The guideline is clear that the diagnosis belongs to the clinician who reads the tracing, and that is the right place for it. As TelEmergency medical director, I also think about the small rural department where the physician sees the patient through a video link. A watch strip held up to a camera is a picture of a picture. It is still worth reading, and it deserves less confidence than a clean 12-lead in your hand. The core argument holds: the watch starts the question, and an ECG somebody actually read is what answers it.

That being said, the article makes the single-lead read sound easier than it is. One lead on a small screen does not show you much atrial activity, and tremor, motion artifact and frequent premature atrial beats can all look irregularly irregular. The honest answer is that most of us read far fewer one-lead strips than 12-leads, and we are less practiced at it than we think. If you cannot say with confidence that the strip shows AF, call it uninterpretable and chart exactly that. A diagnosis of AF is not a small thing to write. It follows a patient into every future chart, every future procedure and every future anticoagulation discussion, and it is very hard to take back once it is there.

If you are the attending, put the patient's own watch to work while he is in your department. Ask him to record a watch tracing while he is on your monitor and read the two side by side, so you learn what his device shows when you know the rhythm. If he stays in sinus rhythm, tell him plainly that the alert has been neither confirmed nor cleared, so he understands why the follow-up matters. If you are leading an emergency department, write down two things this month: who reads watch strips when the reading matters, and how a patient discharged in sinus rhythm gets a monitor and a clinician who will read its result. A monitor ordered with nobody assigned to read it is a referral to nowhere. Decide those two answers before the next alert walks in, because the next one is already on somebody's wrist.

— Chester Shermer, MD, FACEP | Emergency Medicine, 25+ Years Clinical Experience | State Surgeon

Continue Your Training

If you're an emergency physician (or any clinician treating patients daily) trying to understand how AI will actually impact your clinical practice — not just the hype — I put together a free practical guide. You can download the AI in EM Survival Guide here.

Browse All Courses at Global MedOps Command

Relevant Reading on Global MedOps Command:

How to Avoid Becoming an AI Casualty — Dr. Shermer's guide to navigating AI tools in clinical and operational settings without compromising judgment or patient outcomes.

Emergency Department Efficiency Playbook — Practical systems for throughput, triage optimization, and operational efficiency.

Read more from Dr. Shermer on Medium

Connect with Dr. Shermer: LinkedIn — Chester "Chet" Shermer, MD, FACEP.

References

1. Joglar JA, Chung MK, Armbruster AL, et al. 2023 ACC/AHA/ACCP/HRS Guideline for the Diagnosis and Management of Atrial Fibrillation: A Report of the American College of Cardiology/American Heart Association Joint Committee on Clinical Practice Guidelines. Circulation. 2024;149(1):e1-e156. PubMed 38033089
2. Perez MV, Mahaffey KW, Hedlin H, et al. Large-Scale Assessment of a Smartwatch to Identify Atrial Fibrillation. N Engl J Med. 2019;381(20):1909-1917. PubMed 31722151
3. Lubitz SA, Faranesh AZ, Selvaggi C, et al. Detection of Atrial Fibrillation in a Large Population Using Wearable Devices: The Fitbit Heart Study. Circulation. 2022;146(19):1415-1424. PubMed 36148649
4. Alpar S, Tatlıparmak AC. Evaluating smartwatch-based detection of supraventricular tachycardia and atrial fibrillation in the emergency department. Am J Emerg Med. 2025;95:101-106. PubMed 40440816
5. Paslı S, Topçuoğlu H, Yılmaz M, et al. Diagnostic accuracy of apple watch ECG outputs in identifying dysrhythmias: A comparison with 12-Lead ECG in emergency department. Am J Emerg Med. 2024;79:25-32. PubMed 38330880
6. Demkowicz PC, Dhruva SS, Spatz ES, et al. Physician responses to apple watch-detected irregular rhythm alerts. Am Heart J. 2023;262:29-37. PubMed 37084933
7. Healey JS, Lopes RD, Granger CB, et al. Apixaban for Stroke Prevention in Subclinical Atrial Fibrillation. N Engl J Med. 2024;390(2):107-117. PubMed 37952132
8. Kirchhof P, Toennis T, Goette A, et al. Anticoagulation with Edoxaban in Patients with Atrial High-Rate Episodes. N Engl J Med. 2023;389(13):1167-1179. PubMed 37622677
9. van Steijn NJ, Blommestijn IS, Blok S, et al. Enhanced Detection and Prompt Diagnosis of Atrial Fibrillation Using Apple Watch: A Randomized Controlled Trial. J Am Coll Cardiol. 2026;87(14):1714-1728. PubMed 41569211
10. Van Gelder IC, Rienstra M, Bunting KV, et al. 2024 ESC Guidelines for the management of atrial fibrillation developed in collaboration with the European Association for Cardio-Thoracic Surgery (EACTS). Eur Heart J. 2024;45(36):3314-3414. PubMed 39210723

Related simulation training: EM-Sim · Books by Dr. Shermer