A method to Quantify Memory Utilization of Heartbeat Dynamics in Continuous Time

Abstract

We present an approach to quantify the predictive capacity of point processes analyzed in continuous time. An index of the so-called memory utilization rate (MUR) is defined as a model-free measure of the amount of predictive information stored in a point process. A strategy for the data-efficient, model-free estimation of MUR is proposed and validated in simulations of independent and history-dependent heartbeat intervals. Our results document the ability of the proposed MUR estimator to detect the absence and quantify the extent of predictable information in heart rate variability signals.