Risk prediction model for 24-hour mortality in preterm infants using lactate and blood gas analysis: A machine learning approach and retrospective cohort study

Background: This study aimed to evaluate the performance of machine learning algorithms using lactate and arterial blood gas parameters predict imminent risk death in extremely low birth weight infants. Methods: A retrospective cohort analyzing preterm infants with less than 1000 grams a single-center tertiary neonatal intensive care unit São Paulo, Brazil, between 2012 2017 was carried out. We included all at least one analysis paired serum lactate. To assess 24-hour mortality risk, we conducted three (Logistic Regression, Extreme Gradient Boosting, AutoML Tables). Results: analyzed 1932 samples matched measurements. Our population had median gestational age 27.1 (26 – 29.1) weeks 746 (600 880) grams. The Boosting model achieved highest area under receiver operating characteristic (AUROC) 0.898. Base excess, lactate, pH were, order importance, most important features associated mortality. Conclusions: Incorporating into real-time predictive models may aid identify those higher death.