Preprocessing heart rate variability data improves risk stratification in congestive heart failure

Ambulatory 24-hour heart rate variability (HRV) data were used to stratify thirty congestive heart failure (CHF) patients according to their risk of death. Eight of these patients subsequently died. A simple nonstationarity index (NSI) was used to sort the data from least to most stationary. Beginning with the most stationary segments, periodogram spectra were cumulatively estimated, and confidence intervals were also computed. Spectra were computed using only those segments which yielded narrower confidence intervals, and also using all the data segments. Spectral measures extracted included total spectral power and low to high frequency power ratio. A comparison between patients alive and deceased revealed that using the NSI improved the statistical significance of spectral measures. Thus preprocessing of HRV data prior to spectral analysis might provide better means of risk stratification among cardiac patients with subtle differences in spectral characteristics.