Predicting the length of stay for neonates using heart-rate Markov models

Recently, there has been an increase in preterm newborn's survival and as a result longer length of stay (LOS) in neonatal intensive care units (NICU) have been observed. Conseqyently, there has been an increased interest in accurate prediction of the LOS. Most of the existing techniques are based on physiological parameters i.e., scores such as (SNAP-II or SNAPPE-II). In this paper we propose to predict the length of stay using heart-rate measurements combined with physiological scores. We propose to model the heart-rate using Markov chain model and estimate transition probabilities using maximum likelihood estimator and the patient population from Neonatal Intensive Care Unit at McMaster Hospital. We then derive the maximum likelihood estimators of LOS using physiological measurements and transition probabilities. Training and test data sets were used to verify the proposed linear and nonlinear estimators.