Scaling profiles: A novel diagnostic approach for enhanced characterization of heart rate variability in cardiac pathologies

Time series analysis methods, such as Detrended Fluctuation Analysis (DFA) and Detrending Moving-Average Analysis (DMA), estimate scaling exponents and characterize long-range temporal correlations by identifying broad linear regions in log-log plots of fluctuation magnitude and scale. However, this approach has limitations when applied to complex physiological signals that exhibit transitions from scaling to non-scaling behavior, lack clear scaling properties, or display multiple scaling regimes. Previous research has indicated that long-range temporal correlations in heart rate variability (HRV) differ between healthy individuals and those with cardiovascular conditions such as congestive heart failure and atrial fibrillation. Despite these findings, the full diagnostic performance of scale-dependent variations in these exponents remains unexplored. To address this gap, we introduce a novel “scaling profile” methodology that systematically maps local scaling behavior across various scales and window widths. We preferred DMA for all analyses because DFA yields highly unstable local scaling exponents. We establish the validity of this methodology using simulated processes with known long-range temporal correlations. Applying this method to HRV data from healthy individuals and patients with congestive heart failure and atrial fibrillation demonstrated remarkable diagnostic performance. Specifically, we quantified the diagnostic performance of scaling profiles using receiver operating characteristic (ROC) analysis, revealing patterns of discriminatory power that align with pathophysiological mechanisms underlying different cardiac conditions. Congestive heart failure exhibited good discrimination (maximum area under the curve, AUCmax = 0.889) at smaller window widths and shorter scales, while atrial fibrillation showed excellent discriminatory power with AUCmax ∼ 1, reflecting condition-specific patterns of autonomic modulation. ROC analysis revealed local scaling exponents could effectively discriminate between survivor and nonsurvivor CHF patients ( AUCmax = 0.63) and between AF patients with and without ischemic stroke (AUCmax = 0.64). Kaplan-Meier survival analyses demonstrated significant stratification of highand low-risk CHF patients for mortality (χ2 = 7.07, p = 0.008) and highand low-risk AF patients for ischemic stroke (χ2 = 8.48, p = 0.004) using optimal cutoff values of local scaling exponents, confirming their robust prognostic value across both cardiovascular conditions. Our approach reveals rich dynamical signatures that more comprehensively characterize cardiac pathologies, advancing the theoretical framework and clinical utility of scaling analysis in physiological time series.