Predicting the transition to acute heart failure by refined multiscale entropy analysis of heart rate variability in chronic heart failure patients

Abstract

Introduction: Several nonlinear heart rate variability (HRV) algorithms were developed to stratify various risks of cardiac patients, but the comparison of the methods showed different results. Telemedicine ECG monitoring is an approach to the early evaluation of changes in physiological parameters that lead to acute decompensate heart failure and would be valuable to reduce rehospitalization rate and reduce the costs. Objectives: Todetermine the risk of death predicting by theRefinedMultiScale Entropy (RMSE) analysis of HRV in chronic heart failure patients (CHF). Methods: Refined Multiscale Entropy (RMSE) is based on three steps: 1) progressive elimination of the fast time scales; 2) coarse graining procedure necessary to assess entropy rate; and 3) calculation of the entropy rate. Sixtysix CHF patients (NYHA III-IV, EF b 35%, age: 68.4 ± 1.3 years, M/F: 37:29) were enrolled in the first phase of our telemedicine study. Patients were followed for 18 months; a 24-hour Holter registration (HeartRhythmGuard telemedicine system) was performed in every week. The ECG signals were managed by the store-and-forward technique and the supervision of the automated non-linear analysis was performed in the telemedicine center. Group_A contains 23 (dead), Group_B contains 43 (alive) patients. In the second telemedicine study, the RMSE parameters were used as an indicator of CHF worsening: a multivariate discriminant score was calculated for each patient (input: SE values at 20 scales, output: Group_A or _B). Shifting the score to the Group_A value means high risk, therefore immediate telemedicine consultation was performed. Thirty-nine patients (T_Group) were enrolled in the next 18-month study (NYHA III-IV, EF b 35%, age: 63.7 ± 1.8 years, M/F: 16:23), and the data of survival, hospitalization rate and costs were compared with the non-telemedicine (nT_Group: 36 CHF patients NYHA IIIIV, EF b 35%, age: 64.4 ± 1.6 years, M/F: 18:21) patients. Results: All the RMSE curves exhibited a minimum at short time scale followed by an exponential increment with t-scale. The largest statistical difference was obtained at the time scale t = 7 with a p-value = 0.0019 (Group_A: SE = 1.113 ± 0.021, Group_B: SE = 1.477 ± 0.019). The risk of death increased continuouslywith each quartilewith an adjusted relative risk of 4.2 (95% CI 1.7–6.9, p b 0.0001) was observed. Using this cutoff value at scale_t: 7, the unadjustedOR for death was 4.4 (p b 0.0001). The results of the second study: hospitalization rate (in 18 months) in T_Group: 2.1 ± 0.21 and in nT_Group: 3.9 ± 0.24; duration of hospitalization in T_Group 764 days, and in nT_Group 1668 days, the cost savings was 68466 US$/year.