Skip main navigation

Smartwatch Performance for the Detection and Quantification of Atrial Fibrillation

Originally publishedhttps://doi.org/10.1161/CIRCEP.118.006834Circulation: Arrhythmia and Electrophysiology. 2019;12:e006834

    Abstract

    Background:

    Atrial fibrillation (AF) burden and duration appear to be related to stroke risk. A wearable consumer electronic device could provide long-term assessment of these measures inexpensively and noninvasively. This study compares the accuracy of an AF-sensing watch (AFSW; Apple Watch with KardiaBand) with simultaneous recordings from an insertable cardiac monitor (ICM; Reveal LINQ).

    Methods:

    SmartRhythm 2.0, a convolutional neural network, was trained on anonymized data of heart rate, activity level, and ECGs from 7500 AliveCor users. The network was validated on data collected in 24 patients with ICMs and a history of paroxysmal AF who simultaneously wore the AFSW with SmartRhythm 0.1 software. The primary outcome was sensitivity of the AFSW for AF episodes ≥1 hour. Secondary end points included sensitivity of the AFSW for detection of AF by subject and sensitivity for total AF duration across all subjects. Subjects with >50% false-positive AF episodes on ICM were excluded.

    Results:

    We analyzed 31 348.9 hours (mean (SD), 11.3 (4.4) hours/day) of simultaneous AFSW and ICM recordings in 24 patients. The ICM detected 82 episodes of AF ≥1 hour while the AFSW was worn, with a total duration of 1127.1 hours. Of these, the SmartRhythm 2.0 neural network detected 80 episodes (episode sensitivity, 97.5%) with a total duration of 1101.1 hours (duration sensitivity, 97.7%). Three of the 18 subjects with AF ≥1 hour had AF only when the watch was not being worn (patient sensitivity, 83.3%; or 100% during time worn). Positive predictive value for AF episodes was 39.9%.

    Conclusions:

    An AFSW is highly sensitive for detection of AF and assessment of AF duration in an ambulatory population when compared with an ICM. Such devices may represent an inexpensive, noninvasive approach to long-term AF surveillance and management.

    WHAT IS KNOWN?

    • Atrial fibrillation (AF) duration and burden are increasingly recognized as predictors for stroke risk.

    • Smartphone and smartwatch technologies have been shown to accurately differentiate sinus rhythm from AF but provide only a brief rhythm assessment without information on AF duration or burden.

    WHAT THE STUDY ADDS?

    • A commercially available smartwatch with a Food and Drug Administration–cleared ECG sensor, application, and investigational algorithm is highly sensitive for detection of AF episodes lasting ≥1 hour in an ambulatory population and for assessment of AF duration when compared with an insertable cardiac monitor.

    Introduction

    Atrial fibrillation (AF) is the most common sustained arrhythmia in adults, has a lifetime risk of 25% to 33%, and is associated with heart failure, stroke, dementia, and death.1,2 It is well recognized that AF is often asymptomatic and may elude detection particularly in its paroxysmal form.3 Treatments including antiarrhythmic drugs and ablation may reduce AF burden but can also convert symptomatic episodes into asymptomatic episodes, making patient-reported symptoms an unreliable measure of the success of rhythm control interventions.4,5

    A plethora of monitoring techniques have been used for AF detection, but traditional monitors were not designed for long-term use and thus have limited sensitivity.6,7 Insertable cardiac monitors (ICMs) allow for long-term AF surveillance and have been used successfully to screen for AF in selected patient populations.8,9 However, these devices are invasive, expensive, have a limited battery life, and provide no real-time feedback to the patient.10 Smartphone technologies have been developed that can assess heart rate and rhythm using either photoplethysmography or single-lead ECG but provide only a brief rhythm assessment without information on AF duration or burden.11 In contrast, KardiaBand (KB; AliveCor, Mountain View, CA) is a Food and Drug Administration–cleared smartwatch accessory that allows a patient to record a 30-second lead I rhythm strip (Figure 1). Coupled with an investigational application that provides continuous assessment of heart rate, heart rate variability, and activity along with automatic rhythm adjudication, the device has the capability of functioning as a continuous, wearable AF monitor with real-time patient notification that also provides data on AF duration. KB has been evaluated in the setting of electrical cardioversion, but no previous studies have evaluated the device in an ambulatory population with an AF history nor compared its accuracy with an ICM for detection of AF episodes and duration.12

    Figure 1.

    Figure 1. Atrial fibrillation (AF)–sensing watch. A, Application shows average heart rate (upper band) and activity level (lower chart) with an increase in heart rate (orange arrow) in the presence of physical inactivity. B, Patient notification to record ECG. C, ECG recording.

    Methods

    Study Design

    This study used 2 data sets: (1) an anonymous training data set of continuous heart rate, activity, and ECG data acquired from 7500 AliveCor users to train the AF-sensing watch (AFSW) algorithm and (2) a validation cohort of 26 patients from a single center (Northwestern Memorial Hospital, Chicago, IL) enrolled between May 2017 and September 2017. Patients with previously implanted ICMs (Reveal LINQ; Medtronic Inc, Minneapolis, MN) and a history of paroxysmal AF were eligible for enrollment. The AFSW consisted of a commercially available Apple Watch Series 2 (Apple Inc, Cupertino, CA) and KB with an earlier iteration of the proprietary SmartRhythm app (SmartRhythm 0.1). AliveCor provided the KBs and Apple Watches paired to smartphones (iPhone 6; Apple Inc, Cupertino, CA) for use in this study. Patients were asked to wear the watch during waking hours.

    The study was compliant with the Declaration of Helsinki and all patients gave written informed consent. The study was approved by the Northwestern University Institutional Review Board. AliveCor was not involved in the study design, implementation, or article preparation and provided patient data in a blinded manner only. AliveCor owns the intellectual property on the KB and all algorithms which operate on the heart rate, activity, and ECG data including the AF detection algorithms. Requests to access the data set from qualified researchers trained in human subject confidentiality protocols may be sent to the corresponding author.

    AFSW Algorithm and Training

    The deep convolutional neural network, SmartRhythm 2.0, operated on samples of heart rate and activity data from the Apple Watch, and output a probability from 0 to 1 indicating the likelihood that AF was detected in the time window spanned by the samples. The network consisted of 4 convolutional layers and 4 fully connected layers. Max pooling, batch normalization, and dropout were used. The network input was an array of 640 frames, with each frame containing a time-synchronized instantaneous heart rate sample, a motion index, and a timestamp. The heart rate sample and timestamp came directly from Apple HealthKit (Apple Inc, Cupertino, CA). The proprietary SmartRhythm application was used to configure the Apple Watch into workout mode, which enables high-frequency heart rate and activity sampling. The motion index was calculated from Apple HealthKit pedometer data and represented a moving integral of all steps detected in the 3 minutes before the corresponding heart rate sample. On average, Apple HealthKit produces a heart rate sample every 5 to 6 seconds in workout mode, so the 640-frame input to the neural network represented a real-world time period of ≈1 hour.

    The training data set consisted of continuous heart rate and activity data in the form described above collected from 7500 AliveCor users, along with ECGs taken at the users’ discretion. To train the network, the AliveCor automated ECG classification algorithm was used to detect AF in each recorded ECG. The algorithm has been previously validated.13,14 The 640 frames of heart rate / activity data before and after each recorded ECG were labeled according to whether the ECG algorithm detected AF. Heart rate / activity data with no ECG within ±640 frames were discarded. The network was then trained to detect the presence of AF in an arbitrary 640-frame window. The algorithm’s performance and convergence were evaluated against a validation set consisting of data from patients randomly excluded from the training set. For a goal of detecting the presence of at least 1 AF-classified ECG within a randomly sampled 1-hour window of heart rate/activity data that contained at least 1 ECG of any classification, the algorithm achieved a sensitivity of 74.8% at a fixed specificity of 90.0% against the validation set without correcting for ECG classification errors.

    Validation Cohort and Episode Classification

    The primary analysis was sensitivity of the neural network for detecting AF episodes ≥1 hour relative to the ICM (Figure 2). One hour was selected as a threshold given the higher false-positive (FP) rate for shorter episodes on the ICM.15 In this episode-based approach, all episodes were counted equally, irrespective of the patient who had the episode. If continuous AF developed during the study, this was considered to be a single episode. Only those ICM events that occurred while the smartwatch was simultaneously being worn were included in the primary analysis. Figure 3 shows how smartwatch detections were classified. Any ICM-detected AF episode ≥1 hour that overlapped either all or in part with an episode classified as AF on the AFSW was classified as an AFSW-detected episode. Partial overlap was accepted to allow for AFSW removal by the patient during an AF episode. Multiple AFSW detections that overlapped with a single ICM episode of AF were counted as a single AF episode. Subjects were excluded a priori if >50% of ICM-detected AF episodes were not due to AF (ie, >50% FPs) as determined by manual review of electrograms from the ICM. An episode of AF detected by the neural network with the arrhythmia not visible between the start time and end time on the ICM was counted as an FP episode. False-negative (FN) episodes were those seen on the ICM that had no overlapping episode detected on the AFSW. In contrast to true positive, FP, and FN episodes, the true-negative episodes could not be defined in the episode-based approach because the ICM was not programmed to collect electrograms in sinus rhythm. Therefore, only positive predictive value (PPV) and sensitivity could be determined, whereas negative predictive value (NPV), specificity, and accuracy could not be determined.

    Figure 2.

    Figure 2. Atrial fibrillation (AF) detection by an AF-sensing watch and an insertable cardiac monitor (ICM).A, An AF-sensing watch recording showing a one-day plot of heart rate (blue dots) and activity level (light blue vertical bars). ECGs in the validation cohort were classified as either sinus rhythm (green dots) or AF (red dots) with example tracings shown. B, An ICM printout showing duration of atrial tachycardia (AT)/AF corresponding to the same day (red box).

    Figure 3.

    Figure 3. Episode classification. AFSW indicates atrial fibrillation–sensing watch; ICM, insertable cardiac monitor; and TN, true negative.

    Secondary end points included a patient-based approach which assessed sensitivity, specificity, NPV, and PPV for detection of AF per subject. In this analysis, patients with at least one true-positive AF detection were considered a true-positive patient. A duration-based approach was also performed to assess the sensitivity, specificity, NPV, and PPV for identifying the total accumulated AF duration across all patients while the AFSW was worn. The protocol required synchronization of the ICM and AFSW clocks. Periods of AF seen by the ICM and AFSW were aligned timewise with the episode duration defined by the ICM. In cases where the AFSW was removed during an ICM-documented AF episode, the duration of the AF episode was truncated at the time of watch removal.

    Enrolled patients underwent reprogramming of their ICMs to study-specific settings that included detection on for AF only, sensitivity 0.035 V, balanced sensitivity for AF detection, and nominal ectopy rejection. Data were transmitted over the Medtronic CareLink Network to the study center.

    Statistical Analysis

    Statistical analysis was completed using SAS v. 9.4 (SAS Institute, Cary, NC). Descriptive statistics are reported as count and percentage for categorical variables and median, mean, SD, and interquartile range for continuous variables. Power calculations for sensitivity were performed under a variety of scenarios under the assumption that 75 to 100 AF episodes would occur during the monitoring period. We tested the null hypothesis that sensitivity for episodes would be <90%. Based on an exact test and exact estimation method, conservative estimates of statistical power exceeded 80%, at a type I error level of 5%, when the true sensitivity level exceeded 0.9.

    Results

    Study Population

    Baseline characteristics of the validation cohort are shown in the Table and in the Data Supplement. Of the 26 enrolled patients, 2 were excluded for demonstrating >50% of ICM-detected AF episodes that were not due to AF as adjudicated by available electrograms. Additionally, a complete list of AF episodes was not available for these 2 patients due to a high number of FP episodes that exceeded the memory of the ICM. The original indication for ICM monitoring was AF management in 76.9% of patients, cryptogenic stroke in 11.5%, and syncope in 11.5%. However, all patients had AF detected on their ICMs as an enrollment criterion.

    Table. Baseline Characteristics

    Characteristics% (n) or Mean (SD)
    Age, y72.1 (7.2) (median: 72.5)
    Female gender34.6% (9)
    Stroke15.4% (4)
    Transient ischemic attack7.7% (2)
    Congestive heart failure0.0% (0)
    Diabetes mellitus7.7% (2)
    Hypertension69.2% (18)
    Coronary artery disease15.4% (4)
    Prior myocardial infarction7.7% (2)
    CHADS2VASc risk score (categorical)
     0–17.7% (2)
     2–480.7% (21)
     >411.5% (3)
    CHADS2VASc risk score (quantitative)3.0 (1.3) (median: 3)
    Antiarrhythmic drugs34.6% (9)
    Oral anticoagulants84.6% (22)
    Rhythm control: previous ablation50.0% (13)
    ICM indication
     AF management76.9% (20)
     Palpitations0.0% (0)
     Cryptogenic stroke11.5% (3)
     Syncope11.5% (3)

    AF indicates atrial fibrillation; and ICM, insertable cardiac monitor.

    AF Detection

    In total, 63 518.8 hours (2646.6±857.9 hours per patient) of ICM monitoring time were included from 24 patients. During this time, there were 31 348.9 hours (1306.2±660.7 hours per patient) of AFSW monitoring corresponding to a mean (SD) of 11.3 (4.4) hours and median of 11.8 hours (interquartile range, 7.8–14.2 hours) of smartwatch monitoring per person per day. The mean (SD) number of days where the watch was worn per subject was 110 (35.7) and the median days worn was 123.8 (interquartile range, 107–129). Eighty-two episodes of AF ≥1 hour were detected on the ICM while the smartwatch was being worn, of which 80 episodes were detected by the AFSW (97.5% sensitivity per episode). Of these 82 episodes, the patient’s indication for ICM monitoring was syncope for 14 episodes, cryptogenic stroke for 3 episodes, and AF management for 65 episodes, though all patients had previously documented AF on their ICMs as an enrollment criterion. Two episodes involving 2 patients were not detected as AF by the AFSW. There were 499 AF-detected episodes on the AFSW, of which 199 overlapped with ICM-detected AF episode (PPV, 39.9%). The total duration of all AF episodes detected on the ICM during simultaneous AFSW monitoring was 1127.1 hours, of which 1101.1 hours were detected by the AFSW (duration sensitivity, 97.7%). The specificity, PPV, and NPV for detection of AF duration were 98.9%, 76.8%, and 99.9%, respectively. There were 18 patients in total who had any AF ≥1 hour on the ICM, of whom 15 were detected on the AFSW, resulting in an 83.3% sensitivity by subject. The 3 FN patients had AF ≥1 hour only overnight while the AFSW was not worn (mean AF duration, 3.1 hours). There were no FN patients among those who had AF ≥1 hour while wearing the AFSW (100% sensitivity among patients with AF ≥1 hour while wearing the AFSW). There were 5 true-negative patients with no AF on either the ICM nor the AFSW, and 1 patient with FP AF detection on AFSW, yielding a specificity of 83.3%, PPV of 93.8%, and NPV of 62.5% in the patient-based analysis. The mean heart rate during sinus rhythm and AF were strongly correlated between the smartwatch photoplethysmography sensor and the KB ECG (Figure 4).

    Figure 4.

    Figure 4. Comparison of mean heart rate (HR) measurements acquired using Apple Watch PPG (photoplethysmogram) and KardiaBand (KB). A, HR measurements during sinus rhythm. B, HR measurements during atrial fibrillation.

    During the 300 AFSW recordings that did not overlap with ICM-detected episodes, 118 ECGs were available. These were either prompted by the early generation SmartRhythm 0.1 algorithm present on the AFSW units worn by the subjects or were unprompted. These ECGs were manually reviewed as a sample of discordant episodes to calculate the proportion of ICM FNs and AFSW FPs. Eighty-nine (75%) of these were AFSW FPs, but 29 (25%) of the episodes were actually AF episodes missed by the ICM. The most common finding seen among the 89 AFSW FP ECGs was frequent ectopic beats (62; Figure 5), followed by normal sinus rhythm (19), supraventricular tachycardia (3), sinus tachycardia (1), sinus arrhythmia (3), and atrial flutter (1). The ECG findings were not considered in the end points. Therefore, the PPV of 39.9% appeared to underestimate the accuracy of the AFSW due to episodes missed by the ICM, and the true PPV may be closer to 50%.

    Figure 5.

    Figure 5. Example of a false positive atrial fibrillation (AF) episode. A, ECG obtained from a false-positive AF-sensing watch episode showing sinus rhythm with premature supraventricular complexes in bigeminy. Green dots represent sinus complexes, whereas blue dots represent premature supraventricular complexes. B, Plot of R-R intervals showing a pattern of premature complexes in atrial bigeminy.

    Discussion

    These results demonstrate that a commercially available smartwatch with a Food and Drug Administration–cleared ECG sensor, app, and investigational SmartRhythm algorithm is highly sensitive for the detection of AF episodes lasting ≥1 hour in an ambulatory population and for assessment of AF duration when compared with an ICM. Though ICMs have a high PPV for AF, particularly for episodes longer than one hour, cost and other infrastructural barriers limit the widespread use of this technology for detection and management of AF.15,16 The Kardia smartphone ECG adaptor and KB have previously been shown to accurately differentiate sinus rhythm from AF with sensitivity of 93% to 98% and specificity of 83% to 97%.12,13 However, the present study is the first to assess the accuracy of AFSW in an ambulatory population for continuous rhythm monitoring of AF and assessment of AF duration.12,13 Studies using implanted cardiac rhythm management devices have demonstrated that AF burden of ≥5.5 hours in a given day or episodes lasting >24 hours in duration are associated with increased thromboembolic risk, whereas this does not seem to be the case with very short episodes.17–20 As AF duration and burden are increasingly recognized as predictors for stroke risk,21,22 an accurate, wearable, and inexpensive AFSW could provide capability for long-term stroke-risk assessment and potential management of anticoagulation as has been previously demonstrated with ICMs and other implanted devices.9,23,24

    The AFSW equipped with KB that was tested in this study bears several differences to the Apple Watch Series 4 with ECG app and Irregular Rhythm Notification feature that is also commercially available. The AFSW acquired heart rate measurements every 5 seconds regardless of activity as opposed to the Apple Watch Series 4 Irregular Rhythm Notification feature which collects heart rate at a minimum interval of 15 minutes only when the individual is stationary. In addition, the Apple Watch Series 4 Irregular Rhythm Notification Feature is not Food and Drug Administration cleared for use in patients with a previous diagnosis of AF.

    There are several limitations to this study. First, the smartwatches used in this study had a battery life of ≈24 hours and required daily charging for 1 to 2 hours. Advances in battery technology are expected to improve longevity. Second, the mean wear time for the smartwatch was 11.3 hours daily, and the majority of individuals chose not to wear the watch while sleeping. Longer wear times may be expected in those individuals when using this technology in a clinical setting. Third, an AF threshold of ≥1 hour was used for inclusion of AF episodes in light of several studies suggesting that shorter episodes are not associated with stroke but are associated with high FP rates on ICMs.18,21,25,26 Indeed, when a 30-minute threshold is used in the present study instead of 1 hour, the sensitivity is similar (95.7%) but the PPV decreases to 29.9%. There are no differences in the patient-based analysis. Fourth, the study evaluated data in only 24 patients. Although the number of enrolled patients was smaller than other studies evaluating the accuracy of ICMs, the monitoring period was significantly longer and the number of AF episodes was similar to other studies that evaluated 2 monitoring techniques simultaneously.16,27,28 Fifth, the number of true-positive episodes recorded on the AFSW was higher than the number of AF episodes on the ICM. The AFSW could potentially interpret a single continuous episode of AF as several shorter contiguous episodes due to subjects removing the AFSW during an episode or due to intervening segments of more regular R-R intervals or slower atrioventricular conduction during AF. Sixth, AFSW accuracy was compared with the ICM as a gold standard, however, the present study demonstrated that not all AF episodes could be correctly detected by the ICM when verified against manual ECG review. Seventh, only patients with a prior history of paroxysmal AF were included, and the observed accuracy in a screening population may be different. Finally, this proof-of-concept study validated the AFSW using heart rate and activity data from subjects who had previously worn KBs with an earlier version of the SmartRhythm algorithm. Improvements in specificity and PPV would be expected if automated or expert-reviewed ECG adjudication were used in conjunction with heart rate data derived from photoplethysmography, and the iteration of AFSW tested in this study would likely still depend on provider review of ECGs recorded by the KB in clinical practice. Whether advances in machine learning and artificial intelligence will allow for AF diagnosis without the need for ECG documentation requires further study.

    Footnotes

    The Data Supplement is available at https://www.ahajournals.org/doi/suppl/10.1161/CIRCEP.118.006834.

    Rod Passman, MD, MSCE, Division of Cardiology, Northwestern University, Feinberg School of Medicine, 251 E Huron St, Suite 8-503, Chicago, IL 60611. Email

    References

    • 1. Lloyd-Jones DM, Wang TJ, Leip EP, Larson MG, Levy D, Vasan RS, D’Agostino RB, Massaro JM, Beiser A, Wolf PA, Benjamin EJ. Lifetime risk for development of atrial fibrillation: the Framingham Heart Study.Circulation. 2004; 110:1042–1046. doi: 10.1161/01.CIR.0000140263.20897.42LinkGoogle Scholar
    • 2. January CT, Wann LS, Alpert JS, Calkins H, Cigarroa JE, Cleveland JC, Conti JB, Ellinor PT, Ezekowitz MD, Field ME, Murray KT, Sacco RL, Stevenson WG, Tchou PJ, Tracy CM, Yancy CW; American College of Cardiology/American Heart Association Task Force on Practice Guidelines. 2014 AHA/ACC/HRS guideline for the management of patients with atrial fibrillation: a report of the American College of Cardiology/American Heart Association Task Force on Practice Guidelines and the Heart Rhythm Society.J Am Coll Cardiol. 2014; 64:e1–e76.CrossrefMedlineGoogle Scholar
    • 3. Page RL, Wilkinson WE, Clair WK, McCarthy EA, Pritchett EL. Asymptomatic arrhythmias in patients with symptomatic paroxysmal atrial fibrillation and paroxysmal supraventricular tachycardia.Circulation. 1994; 89:224–227.LinkGoogle Scholar
    • 4. Verma A, Champagne J, Sapp J, Essebag V, Novak P, Skanes A, Morillo CA, Khaykin Y, Birnie D. Discerning the incidence of symptomatic and asymptomatic episodes of atrial fibrillation before and after catheter ablation (DISCERN AF): a prospective, multicenter study.JAMA Intern Med. 2013; 173:149–156. doi: 10.1001/jamainternmed.2013.1561CrossrefMedlineGoogle Scholar
    • 5. Quirino G, Giammaria M, Corbucci G, Pistelli P, Turri E, Mazza A, Perucca A, Checchinato C, Dalmasso M, Barold SS. Diagnosis of paroxysmal atrial fibrillation in patients with implanted pacemakers: relationship to symptoms and other variables.Pacing Clin Electrophysiol. 2009; 32:91–98. doi: 10.1111/j.1540-8159.2009.02181.xCrossrefMedlineGoogle Scholar
    • 6. Charitos EI, Stierle U, Ziegler PD, Baldewig M, Robinson DR, Sievers HH, Hanke T. A comprehensive evaluation of rhythm monitoring strategies for the detection of atrial fibrillation recurrence: insights from 647 continuously monitored patients and implications for monitoring after therapeutic interventions.Circulation. 2012; 126:806–814. doi: 10.1161/CIRCULATIONAHA.112.098079LinkGoogle Scholar
    • 7. Mittal S, Movsowitz C, Steinberg JS. Ambulatory external electrocardiographic monitoring: focus on atrial fibrillation.J Am Coll Cardiol. 2011; 58:1741–1749. doi: 10.1016/j.jacc.2011.07.026CrossrefMedlineGoogle Scholar
    • 8. Reiffel JA, Verma A, Kowey PR, Halperin JL, Gersh BJ, Wachter R, Pouliot E, Ziegler PD; REVEAL AF Investigators. Incidence of previously undiagnosed atrial fibrillation using insertable cardiac monitors in a high-risk population: the REVEAL AF study.JAMA Cardiol. 2017; 2:1120–1127. doi: 10.1001/jamacardio.2017.3180CrossrefMedlineGoogle Scholar
    • 9. Sanna T, Diener HC, Passman RS; Crystal AF Steering Committee. Cryptogenic stroke and atrial fibrillation.N Engl J Med. 2014; 371:1261. doi: 10.1056/NEJMc1409495CrossrefMedlineGoogle Scholar
    • 10. Tomson TT, Passman R. Current and emerging uses of insertable cardiac monitors: evaluation of syncope and monitoring for atrial fibrillation.Cardiol Rev. 2017; 25:22–29. doi: 10.1097/CRD.0000000000000129CrossrefMedlineGoogle Scholar
    • 11. McManus DD, Lee J, Maitas O, Esa N, Pidikiti R, Carlucci A, Harrington J, Mick E, Chon KH. A novel application for the detection of an irregular pulse using an iPhone 4S in patients with atrial fibrillation.Heart Rhythm. 2013; 10:315–319. doi: 10.1016/j.hrthm.2012.12.001CrossrefMedlineGoogle Scholar
    • 12. Bumgarner JM, Lambert CT, Hussein AA, Cantillon DJ, Baranowski B, Wolski K, Lindsay BD, Wazni OM, Tarakji KG. Smartwatch algorithm for automated detection of atrial fibrillation.J Am Coll Cardiol. 2018; 71:2381–2388. doi: 10.1016/j.jacc.2018.03.003CrossrefMedlineGoogle Scholar
    • 13. Lau JK, Lowres N, Neubeck L, Brieger DB, Sy RW, Galloway CD, Albert DE, Freedman SB. iPhone ECG application for community screening to detect silent atrial fibrillation: a novel technology to prevent stroke.Int J Cardiol. 2013; 165:193–194. doi: 10.1016/j.ijcard.2013.01.220CrossrefMedlineGoogle Scholar
    • 14. William AD, Kanbour M, Callahan T, Bhargava M, Varma N, Rickard J, Saliba W, Wolski K, Hussein A, Lindsay BD, Wazni OM, Tarakji KG. Assessing the accuracy of an automated atrial fibrillation detection algorithm using smartphone technology: the iREAD study.Heart Rhythm. 2018; 15:1561–1565. doi: 10.1016/j.hrthm.2018.06.037CrossrefMedlineGoogle Scholar
    • 15. Mittal S, Rogers J, Sarkar S, Koehler J, Warman EN, Tomson TT, Passman RS. Real-world performance of an enhanced atrial fibrillation detection algorithm in an insertable cardiac monitor.Heart Rhythm. 2016; 13:1624–1630. doi: 10.1016/j.hrthm.2016.05.010CrossrefMedlineGoogle Scholar
    • 16. Hindricks G, Pokushalov E, Urban L, Taborsky M, Kuck KH, Lebedev D, Rieger G, Pürerfellner H; XPECT Trial Investigators. Performance of a new leadless implantable cardiac monitor in detecting and quantifying atrial fibrillation: results of the XPECT trial.Circ Arrhythm Electrophysiol. 2010; 3:141–147. doi: 10.1161/CIRCEP.109.877852LinkGoogle Scholar
    • 17. Glotzer TV, Daoud EG, Wyse DG, Singer DE, Ezekowitz MD, Hilker C, Miller C, Qi D, Ziegler PD. The relationship between daily atrial tachyarrhythmia burden from implantable device diagnostics and stroke risk: the TRENDS study.Circ Arrhythm Electrophysiol. 2009; 2:474–480. doi: 10.1161/CIRCEP.109.849638LinkGoogle Scholar
    • 18. Van Gelder IC, Healey JS, Crijns HJGM, Wang J, Hohnloser SH, Gold MR, Capucci A, Lau CP, Morillo CA, Hobbelt AH, Rienstra M, Connolly SJ. Duration of device-detected subclinical atrial fibrillation and occurrence of stroke in ASSERT.Eur Heart J. 2017; 38:1339–1344. doi: 10.1093/eurheartj/ehx042CrossrefMedlineGoogle Scholar
    • 19. Swiryn S, Orlov MV, Benditt DG, DiMarco JP, Lloyd-Jones DM, Karst E, Qu F, Slawsky MT, Turkel M, Waldo AL; RATE Registry Investigators. Clinical implications of brief device-detected atrial tachyarrhythmias in a cardiac rhythm management device population: results from the registry of atrial tachycardia and atrial fibrillation episodes.Circulation. 2016; 134:1130–1140. doi: 10.1161/CIRCULATIONAHA.115.020252LinkGoogle Scholar
    • 20. Wasserlauf J, Passman RS. When is device-detected atrial fibrillation actionable?Card Electrophysiol Clin. 2018; 10:75–85. doi: 10.1016/j.ccep.2017.11.007CrossrefMedlineGoogle Scholar
    • 21. Go AS, Reynolds K, Yang J, Gupta N, Lenane J, Sung SH, Harrison TN, Liu TI, Solomon MD. Association of burden of atrial fibrillation with risk of ischemic stroke in adults with paroxysmal atrial fibrillation: the KP-RHYTHM study.JAMA Cardiol. 2018; 3:601–608. doi: 10.1001/jamacardio.2018.1176CrossrefMedlineGoogle Scholar
    • 22. Chen LY, Chung MK, Allen LA, Ezekowitz M, Furie KL, McCabe P, Noseworthy PA, Perez MV, Turakhia MP; American Heart Association Council on Clinical Cardiology; Council on Cardiovascular and Stroke Nursing; Council on Quality of Care and Outcomes Research; and Stroke Council. Atrial fibrillation burden: moving beyond atrial fibrillation as a binary entity: a scientific statement from the American Heart Association.Circulation. 2018; 137:e623–e644. doi: 10.1161/CIR.0000000000000568LinkGoogle Scholar
    • 23. Passman R, Leong-Sit P, Andrei AC, Huskin A, Tomson TT, Bernstein R, Ellis E, Waks JW, Zimetbaum P. Targeted anticoagulation for atrial fibrillation guided by continuous rhythm assessment with an insertable cardiac monitor: The Rhythm Evaluation for Anticoagulation With Continuous Monitoring (REACT.COM) pilot study.J Cardiovasc Electrophysiol. 2016; 27:264–270. doi: 10.1111/jce.12864CrossrefMedlineGoogle Scholar
    • 24. Waks JW, Passman RS, Matos J, Reynolds M, Thosani A, Mela T, Pederson D, Glotzer TV, Zimetbaum P. Intermittent anticoagulation guided by continuous atrial fibrillation burden monitoring using dual-chamber pacemakers and implantable cardioverter-defibrillators: results from the Tailored Anticoagulation for Non-Continuous Atrial Fibrillation (TACTIC-AF) pilot study.Heart Rhythm. 2018; 15:1601–1607. doi: 10.1016/j.hrthm.2018.06.027CrossrefMedlineGoogle Scholar
    • 25. Glotzer TV, Hellkamp AS, Zimmerman J, Sweeney MO, Yee R, Marinchak R, Cook J, Paraschos A, Love J, Radoslovich G, Lee KL, Lamas GA; MOST Investigators. Atrial high rate episodes detected by pacemaker diagnostics predict death and stroke: report of the Atrial Diagnostics Ancillary Study of the MOde Selection Trial (MOST).Circulation. 2003; 107:1614–1619. doi: 10.1161/01.CIR.0000057981.70380.45LinkGoogle Scholar
    • 26. Turakhia MP, Ziegler PD, Schmitt SK, Chang Y, Fan J, Than CT, Keung EK, Singer DE. Atrial fibrillation burden and short-term risk of stroke: case-crossover analysis of continuously recorded heart rhythm from cardiac electronic implanted devices.Circ Arrhythm Electrophysiol. 2015; 8:1040–1047. doi: 10.1161/CIRCEP.114.003057LinkGoogle Scholar
    • 27. Lauschke J, Busch M, Haverkamp W, Bulava A, Schneider R, Andresen D, Nägele H, Israel C, Hindricks G, Bänsch D. New implantable cardiac monitor with three-lead ECG and active noise detection.Herz. 2017; 42:585–592. doi: 10.1007/s00059-016-4492-7CrossrefMedlineGoogle Scholar
    • 28. Nölker G, Mayer J, Boldt LH, Seidl K, VAN Driel V, Massa T, Kollum M, Brachmann J, Deneke T, Hindricks G, Jung W, Brunner KJ, Kraus S, Hümmer A, Lewalter T. Performance of an implantable cardiac monitor to detect atrial fibrillation: results of the DETECT AF study.J Cardiovasc Electrophysiol. 2016; 27:1403–1410. doi: 10.1111/jce.13089CrossrefMedlineGoogle Scholar

    eLetters(0)

    eLetters should relate to an article recently published in the journal and are not a forum for providing unpublished data. Comments are reviewed for appropriate use of tone and language. Comments are not peer-reviewed. Acceptable comments are posted to the journal website only. Comments are not published in an issue and are not indexed in PubMed. Comments should be no longer than 500 words and will only be posted online. References are limited to 10. Authors of the article cited in the comment will be invited to reply, as appropriate.

    Comments and feedback on AHA/ASA Scientific Statements and Guidelines should be directed to the AHA/ASA Manuscript Oversight Committee via its Correspondence page.