Analysis of heart rate variability influence on heart rate turbulence using boosted regression trees in heart failure patients

Heart Rate Turbulence (HRT) is a physiological phenomenon used as cardiac risk stratification criterion. The relationship between Heart Rate Variability (HRV) and HRT has been documented in the literature. However, the influence of HRV on HRT using individual tachograms has not been addressed. Our aim was to propose a nonparametric model, based on Boosted Regression Trees (BRT), of turbulence slope (TS) as a function of coupling interval (CI), Age, Sex, and HRV time-domain indices. We used data sets of myocardial infarction (MI) and heart failure (HF) patients. HRV was assessed on 3-min NN interval segments just before to individual ventricular premature complex (VPC) tachograms. We proposed to model TS as a function HRV indices using BRT, which is an ensemble approach to build regression models using several small trees. We segmented data into high risk and low risk according to HRT cut-off values of TS. Variables related to HRV were the most important explaining the HRT in low risk patients, while in patients with high risk, CI and heart rate just before the VPC played an important role explaining the HRT response.