Wearable consumer devices that track health metrics like step count, blood pressure, and heart rate—some of which include single-lead ECGs—have grown in popularity in recent years. Researchers have been exploring the possibility of using data from these devices to passively screen people for arrhythmias, including Afib.
There are questions, however, not only about ECG watch accuracy compared to a standard 12-lead ECG—which may have implications for managing healthcare costs—but also about how physicians should incorporate the growing torrent of patient-created data into their decision-making.
How Accurate Are Consumer Devices in Diagnosing Arrhythmias?
Smartwatches, such as those made by Apple, Samsung, Withings, Fitbit, and others, reportedly have high diagnostic accuracy when it comes to identifying Afib. The ECG app on the Apple Watch, for example, was shown to have 99.3% specificity for classifying sinus rhythm and 98.5% sensitivity for classifying Afib in a clinical trial using 12-lead ECG as the gold standard.1
However, other studies have introduced some skepticism around the accuracy of smartwatch-measured ECG. A study showed that a single-lead ECG measured using a KardiaBand paired with an Apple Watch had only "moderate diagnostic accuracy" in diagnosing Afib compared to a 12-lead ECG in the hospital, though interpretation of unclassified tracings by electrophysiologists did boost performance.2
Another study assessed the Apple Watch ECG's accuracy in detecting Afib in patients who had undergone cardiac surgery. The study showed a sensitivity of 41% for the abnormal rhythm notification from the Apple Watch Series 4 and a sensitivity of 96% for the single-lead ECG downloaded as a PDF, compared to telemetry. The unreadable rate was 31%, however, and the investigators concluded that "physicians should exercise caution before undertaking action based on electrocardiographic diagnoses generated by this wrist-worn monitor."3
The accuracy of consumer wearables is also questionable compared to that of Holter monitors in ambulatory patients with Afib. Research revealed that even though the Fitbit Charge HR and Apple Watch Series 3 had high accuracy compared to Holter monitors in sinus rhythm (92%-95%), these devices under-reported heart rate with Afib, especially when the heart rate was over 80 bpm. Accuracy fell to 56% for the Fitbit Charge HR and 77% for the Apple Watch.4
A major concern about ECG watch accuracy revolves around the people most likely to use the devices, who tend to be young and healthy. In a response to the findings of the WATCH AF trial, researchers cited data indicating that just 4.6% of smartwatch users in the US were older than 65, which means the bulk of wearers would have a very low risk of Afib.5
With the assumption that demographics and Afib prevalence were similar between smartwatch users and the general population, the positive predictive value of watch-measured ECG for detecting Afib would be just 33%. With two-thirds of positive results being false-positive, that could lead to a lot of inappropriate testing and treatment. Another physician made a similar point, calculating a positive predictive value of only 19.6% in people younger than 55.6
To learn more about the power of the ECG in today's clinical landscape, browse our Diagnostic ECG Clinical Insights Center.
Handling the Flood of Data
There are concerns beyond inaccurate data when it comes to consumer-grade ECGs. As consumer wearables become more popular, physicians are going to see an ever-growing mountain of data provided by their patients, which will raise questions about how this information can be efficiently incorporated into clinical decision-making while avoiding burnout.
One way to make this data more workable might be to integrate ECG recordings into electronic medical records (EMRs), potentially with the assistance of artificial intelligence (AI) technology. Researchers speculated that "creation of easy-to-interpret ECG app data summaries with the help of AI in the [EMR]" might help clinicians parse this large amount of data.7
GE Healthcare is working on this issue by finding ways to use AI and machine learning to optimize interpretation of ECG data, both inside and outside the hospital, as well as collect the information in one place for easy access by physicians.
Ashutosh Banerjee, Global General Manager of Cardiology Solutions and Diagnostic Cardiology for GE Healthcare, explains: "What we are trying to do is integrate data from various sources, bring them into one ECG management system or one EMR, and make sure that we can provide insights to the clinicians with that set of algorithms that we have. Those insights need to be actionable, and that's what we are trying to do."
References:
- Apple. ECG App Instructions for Use (IFU). www.apple.com. https://www.apple.com/legal/ifu/ecg/
- Rajakariar K et al. Accuracy of a smartwatch based single-lead electrocardiogram device in detection of atrial fibrillation. Heart. Apr 2020; vol. 106: 665-670. https://heart.bmj.com/content/106/9/665
- Seshadri D R et al. Accuracy of Apple Watch for detection of atrial fibrillation. Journal of the American Heart Association. Feb 2020; vol. 141 (iss. 8): 702-703. https://www.ahajournals.org/doi/10.1161/CIRCULATIONAHA.119.044126
- Toner L et al. The accuracy of smartwatches compared to Holter monitors for heart rate monitoring in atrial fibrillation: A pilot study. Heart Lung and Circulation. Jun 2019: vol. 28. https://www.heartlungcirc.org/article/S1443-9506%2819%2930830-3/fulltext
- Cheung C C et al. Watch out: The many limitations in smartwatch-driven AF detection. Journal of the American College of Cardiology: Clinical Electrophysiology. Apr 2019; vol. 5 (iss. 4): 525-526. https://www.jacc.org/doi/10.1016/j.jacep.2019.02.008
- Sheridan K. Here's the data behind the new Apple Watch EKG app. www.statnews.com. https://www.statnews.com/2018/09/13/heres-the-data-behind-the-new-apple-watch-ekg-app/
- Isakadze N and Martin S S. How useful is the smartwatch ECG? Trends in Cardiovascular Medicine. Oct 2020; vol. 30 (iss. 7): 442-448. https://www.sciencedirect.com/science/article/pii/S1050173819301495
