The emergence of personal monitoring devices
In recent years there has been an increasing array of direct-to-consumer devices capable of detecting AF, predominantly utilizing smart phones and smart watches. Survey data from the Pew Research Center (www.pewresearch.org) suggests that the proportion of US adults with a smart phone has roughly doubled since 2011 to over 80% and the proportion who use a smartwatch or fitness tracker now exceeds 20%, with ongoing expansion each year as more products enter the marketplace. The resulting growth in patient driven heart rhythm monitoring without direct physician oversight has empowered patients to investigate their own rhythm abnormalities either as a screening tool or for disease management. Current methods available to monitor for AF include smart phone finger pulse wave photoplethysmography (PPG) using downloadable applications and hardware already present in modern smartphones, external electrodes that communicate with an app downloaded to a smart phone used to generate an mobile single or six-lead electrocardiogram (iECG), or smart watch PPG with or without single lead iECG confirmation (by touching the crown of the watch with opposite hand to create a single lead ECG).
Personal monitoring devices have been shown to have a relatively high accuracy in detecting AF. The SEARCH-AF study used iECGs to screen 1000 pharmacy customers for AF using 12 lead ECG confirmation, finding a sensitivity of 98.5% and specificity of 91.4% for AF, diagnosing new AF in 1.5% of customers, all of which had a CHADS-VASc score of 2 or more and therefore would conceivably benefit from anticoagulation to reduce stroke risk28,29. Chan and colleagues screened 1013 primary care patients with a smartphone camera-based PPG algorithm against iECG tracings reviewed by two cardiologists and found a sensitivity of 92.9% and specificity of 97.7%, though lower sensitivities are reported outside the research setting30. The WATCH AF trial31compared the accuracy of an automated wrist watch PPG algorithm in diagnosing AF in 672 hospitalized patients as compared to a single lead iECG interpreted by cardiologists as a reference. Although 21.8% of PPG datasets were not interpretable, the remaining datasets had a sensitivity of 93.7% and specificity of 98.2% for detecting AF. Contact-free facial PPG using subtle beat-to-beat variations of skin color has similarly shown promise in detecting AF with similar sensitivity32. As a proof of concept on a larger community scale, the Apple Heart Study recruited 419,297 participants with smart watches using an irregular pulse notification algorithm and mailed a 7-day continuous Holter monitor patch to those with an irregular pulse. They found that 0.52% of participants received notifications of an irregular pulse, of which 34% had AF diagnosed on the monitoring patch that was placed on average 13 days later33. Of the 86 participants that had an irregular pulse notification subsequently while wearing a patch, the positive predictive value of the irregular pulse notification was 0.84 in detecting AF confirmed on the patch.
Advancements in wearable AF-sensors raises the possibility that assessment of AF duration and burden over long time horizons will no longer require implantable devices. In a study of 24 patients Wasserlauf and colleagues compared the accuracy of a simultaneous deep convolutional neural network PPG algorithm with single lead iECG confirmation using KardiaBandTM (AliveCor, Mountain View, CA) with simultaneous ICM recordings34. The smart watch algorithm had a sensitivity of 97.5% in detecting AF episodes > 1 hour and had a high correlation (R2 0.996) with the duration of AF recorded on the ICM. With studies suggesting that stroke risk is highest for the 30 days following an episode of AF35 and that hours and not minutes of AF is likely necessary to increase stroke risk in those with few vascular risk factors, wearable AF-detection devices raise the possibility that stroke risk and bleeding risk can be reduced by targeted, time-delimited, “pill-in-pocket” anticoagulation in those with infrequent episodes of paroxysmal AF, either spontaneously or as the result of AF ablation or surgery.