137-05: Accuracy and usability of handheld electrocardiogram recorders to detect atrial fibrillation in hospitalised patients

Purpose: The aim of this study was to evaluate the accuracy and usability of two handheld electrocardiogram (ECG) devices to detect atrial fibrillation (AF) in a hospital setting. Methods: In total, 503 patients hospitalised at the departments of cardiology (n = 344) and geriatrics (n = 159) received a 12-lead or 6-lead ECG recording. Immediately thereafter, patients were asked to consecutively hold two single-lead handheld ECG devices, i.e. MyDiagnostick (Applied Biomedical Systems, The Netherlands) for one minute and AliveCor (AliveCor Inc., USA) for 30 seconds. Two blinded experienced electrophysiologists reviewed each of the three ECG tracings independently from the others and classified them as AF, atrial flutter, sinus rhythm or uninterpretable. The 12-lead and 6-lead ECGs were used as a ‘gold standard’ to calculate the sensitivity and specificity of the automated algorithm of each device, as well as the performance after manual review of the tracings. Results: Handheld recordings were not possible in 11.5% of the patients because they were not able to hold the devices properly, mainly in the very elderly. In the remaining patients (n = 445), 2.8% and 3.4% of the MyDiagnostick and AliveCor recordings were classified as uninterpretable, respectively. Sixty patients (13.5%) had AF at the moment of the screening. Automated analysis of the MyDiagnostick had a sensitivity of 68.3% and a specificity of 94.0%. The automated algorithm of the AliveCor had a lower sensitivity of 50.0% and a higher specificity of 96.6%. After exclusion of patients with an implanted device (n = 67), sensitivity increased to 85.4% and 65.9% for MyDiagnostick and AliveCor respectively, without major impact on specificity (Table). Manual review of the ECG recordings was especially valuable for the AliveCor as it could increase the sensitivity to 92.7%. However, this was at the expense of a decreased specificity. Table: Performance of the MyDiagnostick and AliveCor for AF screening in hospitalised patients without an implanted device. NPV: negative predictive value, PPV: positive predictive value. Unreadable recordings are taken into account when calculating the sensitivity and specificity. Table: Performance of the MyDiagnostick and AliveCor for AF screening in hospitalised patients without an implanted device. NPV: negative predictive value, PPV: positive predictive value. Unreadable recordings are taken into account when calculating the sensitivity and specificity. Conclusion: Using AliveCor or MyDiagnostick handheld recorders for AF screening in a hospital setting seems challenging as not every patient is capable to hold the devices correctly and uninterpretable recordings and incorrect automatic analysis are still common. Exclusion of patients with an implanted device and manual review of the ECG tracings optimises the accuracy for both devices.