An Algorithm for Predicting Neonatal Mortality in Threatened Very Preterm Birth

Abstract Objective To develop a prediction model for neonatal mortality using information readily available in the antenatal period. Methods A multiple logistic regression model of a complete population-based geographically defined cohort of very preterm infants of 23 + 0 to 30 + 6 weeks' gestation was used to identify antenatal factors which were predictive of mortality in this population. Infants < 23 weeks and those with major anomalies were excluded. Results Between 1996 and 2012, 1240 live born infants < 31 weeks' gestation were born to women residing in Nova Scotia Decreasing gestational age strongly predicted an increased mortality rate. Other factors significantly contributing to increased mortality included classification as small for gestational age, oligohydramnios, maternal psychiatric disorders, antenatal antibiotic therapy, and monochorionic twins. Reduced neonatal mortality was associated with antenatal use of antihypertensive agents and use of corticosteroids of any duration of therapy given at least 24 hours before delivery. An algorithm was developed to estimate the risk of mortality without the need for a calculator. Conclusion Prediction of the probability of neonatal mortality is influenced by maternal and fetal factors. An algorithm to estimate the risk of mortality facilitates counselling and informs shared decision making regarding obstetric management.