An algorithm for risk stratification of preterm infants

Abstract

Preterm infants have a higher incidence of life-threatening events including apnea (cessation of breathing), bradycardia (slowing of heart rate) and hypoxemia (oxygen de-saturation). In Neonatal Intensive Care Units, clinicians face a demanding task of assessing the risk of these infants based on their physiological signals. In this study, we propose an algorithm of heart rate dynamics that could potentially be employed for risk stratification of preterm infants. We collected and analysed heart rate (HR) measures in beats per minute (bpm) in 18 preterm infants for 24 hours during oxygen therapy. We investigated whether the HR fluctuations in the first one hour could predict the number of bradycardia events N (i.e. HR below 100 bpm) in the subsequent 23 hours. Since RR intervals estimated from HR (i.e. RR = 60/HR) in seconds follow a lognormal distribution, we employed an algorithm based on a point process modelling framework to capture HR fluctuations. We found that the instantaneous variance σ2(t) calculated by the point process model for the first 1-hour correlates significantly with N. We also found that σ2(t) correlates with number of hypoxemia in the subsequent 23 hours. Thus, we conclude that the fluctuations in the HR data captured using a point process model can be used to predict life threatening events.