Predicción del síndrome de dificultad respiratoria neonatal en recién nacidos de pretérmino tardío

Late preterm births (34 + 0-36 + 6 gestational weeks) comprise 75% of preterm deliveries and are at increased risk for respiratory distress syndrome (RDS), a leading cause of neonatal morbidity. We aimed to develop a predictive model for RDS in late preterm neonates based on maternal and prenatal clinical variables available before the onset of labor.

A retrospective cohort study was conducted in a tertiary medical center between May 2016 and June 2023. The institutional healthcare database was searched for all women who gave birth to a singleton neonate at 34 + 0-36 + 6 gestational weeks. Three models were applied to predict the likelihood of development of neonatal RDS: multivariable logistic regression, conditional inference trees (CIT), and extreme gradient boosting (XGBoost).

Out of 37,205 deliveries during the study period, 2,444 singleton neonates were included, of whom 137 (5.6%) were diagnosed with RDS. Multivariable logistic regression analysis identified five significant predictors of RDS: gestational age (aOR 0.35, 95%CI: 0.32-0.38), male sex (aOR 1.68, 95%CI: 1.47-1.92), cesarean delivery (CD) (aOR 1.28, 95%CI: 1.20-1.37), nulliparity (aOR 0.47, 95%CI: 0.40-0.54) and antenatal corticosteroids (ACS) (aOR 0.54, 95%CI: 0.45-0.64 for administration before 34 weeks; aOR 0.60, 95%CI: 0.49-0.72 for administration after 34 weeks). These findings represent predictive associations rather than causal treatment effects. Among the predictive models, the logistic regression demonstrated the highest discriminatory performance with an area under the curve (AUC) of 0.75 (95% CI: 0.70-0.79), followed by XGBoost algorithm with an AUC of 0.70 (95% CI: 0.66-0.75), and CIT with an AUC of 0.68 (95% CI: 0.64-0.73).

Models based on prenatal maternal and fetal characteristics demonstrated modest discrimination but strong negative predictive value for identifying RDS in late preterm infants. Such tools may assist with antenatal counseling and support decisions regarding the appropriate level-of-care setting for delivery. Further studies are needed to validate these findings and to explore whether incorporating additional clinical variables can enhance predictive performance.