Impact of Electrocardiogram Length on Diagnostic Properties of Heart Rate Variability Indices in Atrial Fibrillation Detection

Abstract

Heart rate variability (HRV) measures various prop-erties of RR intervals from electrocardiograms (ECGs) with sinus rhythm (SR). Recently, many HRV parameters have been employed to distinguish ECGs with SR from those with atrial fibrillation (AF). However, studies on the diagnostic properties of HRV for AF detection differ both in the tested parameters and in the length of studied ECGs. We aimed to systematically test the diagnostic properties of selected HRV indices for differentiation between AF and SR in ECGs of a wide range of lengths between 5 to 300 s. We demonstrate that the diagnostic performance of HRV for the AF detection strongly depends on the length of the ECG, usually improving with longer recordings. The best diagnostic properties of all HRV indices have been found for pRR50 with the area under curve from 0.881 for 5 s to 0.980 for 300 s, and diagnostic odds ratio from 44 (5 s) to 421 (300 s). We conclude that the length of ECG recording and the selection of HRV parameters are crucial for the proper differentiation of AF from SR. Clinical Relevance- This study demonstrates that the ECG length determines the diagnostic properties and the cutoff values of HRV parameters for AF detection. Additionally, if RR intervals time series analysis is applied to differentiate AF from SR, pRR50 outperforms other HRV parameters used, regardless of the ECG length between 5 and 300 seconds