Multifractal detrended fluctuation analysis of heart rate variability predicts short-term outcomes of patients with sepsis

Abstract

Intensive Care Unit (ICU) is a hospital unit that has an important role in restoring life-threatening patients condition. Predicting the patient's outcome after discharge from the ICU is necessary. The predicted outcome will determine the best medication for them. The paper will determine the ability of heart rate variability (HRV) parameters from multifractal detrended fluctuation analysis method to predict the short-term outcome of patient in ICU with sepsis condition. Data was obtained from MIMIC III Medical-ICU database. Only patients with sepsis condition is used in this study to reduce disease related influence in the study. Total 206 patients data were collected. It consists of 134 survivors and 72 non-survivors patients data. ECG recordings were obtained prior ICU disenrollment. First fifteen minutes recording were extracted into HRV time series. Then multifractal detrended fluctuation analysis were carried out to the time series of each patients. Along with the mentioned method, linear (time-frequency) and other nonlinear HRV measures were also calculated for comparison. Statistical significance is used to measure how the parameters perform to differentiate the outcome difference. The results shows that mean α from MF-DFA parameters can predict patients’ short-term outcome while the available linear and non-linear HRV analysis cannot.