Do Birthweight-For-Gestational Age Centiles Predict Serious Neonatal Morbidity and Neonatal Mortality?

Studies show that foetal and birthweight-for-gestational age centiles are poor predictors of serious neonatal morbidity and neonatal mortality (SNMM) in univariable models. We assessed the predictive performance of multivariable SNMM models based on maternal/pregnancy characteristics, with and without birthweight centiles. The study was based on all live births in the United States, 2019-2021, with data obtained from the period live birth-infant death files of the National Center for Health Statistics. SNMM was defined as any one or more of the following: 5-minute Apgar score 30 or neonatal death. SNMM was modelled by log-linear regression on maternal/pregnancy characteristics as predictors, with and without birthweight centiles. Models were developed for live births at 24-42 weeks' and 39 weeks' gestation to all women and those with hypertensive disorders or pre-existing diabetes. Model performance was assessed using area under the curve (AUC). The study population included 10,487,243 live births and 221,728 SNMM cases (2.1 per 100 live births). The models with all live births at 24-42 weeks' gestation had AUCs of 0.83 (95% confidence interval CI 0.82, 0.83) based on maternal/pregnancy characteristics and 0.83 (95% CI 0.83, 0.84) based on maternal/pregnancy characteristics and birthweight centiles. However, AUCs of models based on all live births at 39 weeks' gestation were 0.66 (95% CI 0.64, 0.68) with maternal/pregnancy characteristics and 0.69 (95% CI 0.68, 0.71) with maternal/pregnancy characteristics and birthweight centiles. AUCs of the models with live births at 39 weeks' gestation to women with pre-existing diabetes were 0.69 (95% CI 0.66, 0.72) based on maternal/pregnancy characteristics, and 0.77 (95% CI 0.74, 0.79) with the addition of birthweight centiles. Birthweight centiles improve multivariable SNMM predictive performance in specific subpopulations, although evaluation of decision thresholds is required to determine the clinical importance of improvement in predictive ability.