Assessing the accuracy of an automated atrial fibrillation detection algorithm using smartphone technology: The iREAD Study

• Published in: Heart rhythm Vol. 15; no. 10; pp. 1561 - 1565
• Main Authors: William, Amila D., Kanbour, Majd, Callahan, Thomas, Bhargava, Mandeep, Varma, Niraj, Rickard, John, Saliba, Walid, Wolski, Kathy, Hussein, Ayman, Lindsay, Bruce D., Wazni, Oussama M., Tarakji, Khaldoun G.
• Format: Journal Article
• Language: English
• Published: United States Elsevier Inc 01.10.2018
• Subjects: Adult; Aged; Aged, 80 and over; Algorithms; Atrial fibrillation; Atrial Fibrillation - diagnosis; Atrial Fibrillation - physiopathology; Cardiac rhythm monitoring; Digital health; Electrocardiography - methods; Female; Heart Conduction System - physiopathology; Humans; Male; Middle Aged; Mobile Applications; Mobile health; Predictive Value of Tests; Reproducibility of Results; Signal Processing, Computer-Assisted; Single-Blind Method; Smartphone; Telemetry - instrumentation; Digital health; Smartphone; Atrial fibrillation; Mobile health; Cardiac rhythm monitoring
• ISSN: 1547-5271; 1556-3871; 1556-3871
• DOI: 10.1016/j.hrthm.2018.06.037

The Kardia Mobile Cardiac Monitor (KMCM) detects atrial fibrillation (AF) via a handheld cardiac rhythm recorder and AF detection algorithm. The algorithm operates within predefined parameters to provide a “normal” or “possible atrial fibrillation detected” interpretation; outside of these parameters, an “unclassified” rhythm is reported. The system has been increasingly used, but its performance has not been independently tested. The objective of this study was to evaluate whether the KMCM system can accurately detect AF. A single-center, adjudicator-blinded case series of 52 consecutive patients with AF admitted for antiarrhythmic drug initiation were enrolled. Serial 12-lead electrocardiograms (ECGs) and nearly simultaneously acquired KMCM recordings were obtained. There were 225 nearly simultaneously acquired KMCM and ECG recordings across 52 enrolled patients (mean age 68 years; 67% male). After exclusion of unclassified recordings, the KMCM automated algorithm interpretation had 96.6% sensitivity and 94.1% specificity for AF detection as compared with physician-interpreted ECGs, with a κ coefficient of 0.89. Physician-interpreted KMCM recordings had 100% sensitivity and 89.2% specificity for AF detection as compared with physician-interpreted ECGs, with a κ coefficient of 0.85. Sixty-two recordings (27.6%) were unclassified by the KMCM algorithm. In these instances, physician interpretation of KMCM recordings had 100% sensitivity and 79.5% specificity for AF detection as compared with 12-lead ECG interpretation, with a κ coefficient of 0.71. The KMCM system provides sensitive and specific AF detection relative to 12-lead ECGs when an automated interpretation is provided. Direct physician review of KMCM recordings can enhance diagnostic yield, especially for unclassified recordings.