A mobile one-lead ECG device incorporated in a symptom-driven remote arrhythmia monitoring program. The first 5,982 Hartwacht ECGs

• Published in: Netherlands heart journal Vol. 27; no. 1; pp. 38 - 45
• Main Authors: Selder, J. L., Breukel, L., Blok, S., van Rossum, A. C., Tulevski, I. I., Allaart, C. P.
• Format: Journal Article
• Language: English
• Published: Houten Bohn Stafleu van Loghum 01-01-2019; Springer Nature B.V
• Subjects: Accuracy; Algorithms; Cardiac arrhythmia; Cardiology; Cardiovascular disease; Classification; Health care; Heart; Medical Education; Medicine; Medicine & Public Health; Original; Original Article; Outpatient care facilities; Patients; Sinuses; Smartphones; Netherlands; AliveCor; arrhythmia; Kardia; ECG; remote monitoring; eHealth; Hartwacht
• ISSN: 1568-5888; 1876-6250
• DOI: 10.1007/s12471-018-1203-4

Background
In recent years many mobile devices able to record health-related data in ambulatory patients have emerged. However, well-organised programs to incorporate these devices are sparse. Hartwacht Arrhythmia (HA) is such a program, focusing on remote arrhythmia detection using the AliveCor Kardia Mobile (KM) and its algorithm.
Objectives
The aim of this study was to assess the benefit of the KM device and its algorithm in detecting cardiac arrhythmias in a real-world cohort of ambulatory patients.
Methods
All KM ECGs recorded in the HA program between January 2017 and March 2018 were included. Classification by the KM algorithm was compared with that of the Hartwacht team led by a cardiologist. Statistical analyses were performed with respect to detection of sinus rhythm (SR), atrial fibrillation (AF) and other arrhythmias.
Results
5,982 KM ECGs were received from 233 patients (mean age 58 years, 52% male). The KM algorithm categorised 59% as SR, 22% as possible AF, 17% as unclassified and 2% as unreadable. According to the Hartwacht team, 498 (8%) ECGs were uninterpretable. Negative predictive value for detection of AF was 98%. However, positive predictive value as well as detection of other arrhythmias was poor. In 81% of the unclassified ECGs, the Hartwacht team was able to provide a diagnosis.
Conclusions
This study reports on the first symptom-driven remote arrhythmia monitoring program in the Netherlands. Less than 10% of the ECGs were uninterpretable. However, the current performance of the KM algorithm makes the device inadequate as a stand-alone application, supporting the need for manual ECG analysis in HA and similar programs.