Purpose: We have shown previously that reduced integer heart rate variability (HRVi) predicts death in trauma patients. We hypothesized that heart rate multiscale entropy (MSE), a potential measurement of physiologic complexity, would predict death more robustly than HRVi.
Materials and methods: Two hundred eighty-five patients had heart rate data meeting completeness and density criteria (>12 hours, >/=0.4 Hz) available in the first 24 hours after admission. Missing data points were interpolated, and a publicly available algorithm (MSE of Costa et al; Phys Rev E Stat Nonlin Soft Matter Phys. 2005;71[2 Pt 1]) was applied (www.physionet.org, m = 2, r = 0.15). Integer heart rate variability was computed using methods described previously (percentage of 5-minute intervals having heart rate SD between 0.3 and 0.6). Sample entropy was compared between survivors and nonsurvivors at each scale factor using Wilcoxon rank sum test. Logistic regression was used to assess risk of death based on HRVi, MSE, and/or covariates (age, sex, injury severity).
Results: Decreased HRVi and MSE each predicted hospital mortality (median day of death, 3; mean, 7.1). Multiscale entropy-based risk stratification (area under the receiver operating characteristic curve [AUC] = 0.76, scale 15) was superior to HRVi (AUC = 0.70), but this difference in AUC was not statistically significant. Multiscale entropy stratified patients by mortality at every scale factor (P < .001).
Conclusions: Multiscale entropy and HRVi measured within the first 24 hours each identify trauma patients at increased risk of subsequent hospital death.