Validity and reliability of the ultra-short-term heart rate variability features in predicting ventricular tachyarrhythmia

Ultra-short-term heart rate variability analysis refers to the analysis of variability in time intervals between successive heartbeats over recordings shorter than 5 min. Recently, several studies have conducted HRV analysis on the ECG recording shorter than 5 min for predicting the onset of ventricular tachyarrhythmia. However, these studies employed the ultra-short-term HRV features without questioning its validity and reliability as a surrogate of the short-term HRV features, which served as the gold standard. Most of them applied only statistical tests, such as the student's T-test, to rank and select the HRV features based on their significance differences between VTA and control groups. To the best of the authors’ knowledge, none of the existing work has rigorously investigated the validity and reliability of the ultra-short-term HRV features extracted from recordings shorter than 5 min in predicting the ventricular tachyarrhythmia. In this study, a total of 30 HRV features, extracted from the time domain, frequency domain, and nonlinear analysis were thoroughly analysed with the corresponding short-term HRV features using the proposed inter-group and intra-group assessments based on statistical significance and correlation analysis. From the experimental findings, only 9 ultra-short-term HRV features successfully passed both the inter-group and intra-group assessments and were selected as the optimal feature subset for predicting the onset of ventricular tachyarrhythmia. With the optimal feature subset, optimistic performance was achieved with an accuracy of up to 86.39% in predicting the onset of ventricular tachyarrhythmia 2 minutes prior to its occurrence using machine learning algorithms. © 2025 Elsevier Ltd