Automated versus interactive spectral analysis of heart rate variability from 24-hour Holter recordings in heart failure patients

Accurate spectral analysis of 24-hour heart rate variability (HRV) requires time consuming interactive control of the overall processing, limiting its feasibility in large scale clinical trials. Fully automatic procedures allow a dramatic reduction of analysis time but the level of accuracy loss is unknown. In this study we compared automatic and interactive measurements of HRV indexes in a sample of 246 Holter recordings from chronic heart failure patients. We measured the total power of HRV and the power in the very low (0.01-0.04 Hz), low (0.04-0.15 Hz) and high (0.15-0.4 Hz) frequency bands. The comparison between the two methods was performed by computing the normalized difference between automatic and interactive measurements and estimating the bias and 95% limits of agreement. Automatic measurements showed a small bias (<±8%) compared to interactive measurements, while the limits of agreement were ≤ ±23% in all spectral indexes. Our results indicate that the systematic and random difference between automatic and interactive measurements of spectral indexes of HRV are acceptable for clinical studies.