Studying the Influence of Maternal Nutritional Status on Offspring Low Birth Weight Prediction Using Machine Learning

Birthweight (BW) is one of the most extensively used markers of postnatal clinical illness and risk of neonatal mortality, and neonates with low birthweight (LBW) are at higher risk. Accurate and early prediction of LBW in the newborn can positively influence the newborn’s safety and health. Various machine learning approaches have recently been implemented for LBW classification and prediction with different maternal features, mainly based on physiological and ultrasound characteristics. However, maternal nutritional attributes such as vitamin B12, iron and folate status have not been widely exploited for BW prediction. This study aims to demonstrate how clinical nutritional attributes associated with LBW (<2500 g) and preterm birth, such as plasma vitamin B12, folate, and iron deficiency, influence the prediction of BW by leveraging advanced ML models. This model will use clinical nutritional attributes to predict the class of BW during the second trimester of pregnancy. The developed models are assessed in a private dataset, including 729 cases from a Spanish birth cohort. The outcomes showed that the ANN with different feature sets outperformed the other classification models by achieving an accuracy of 94.48% and an AUC of 0.94. The findings of this study indicate that maternal vitamin B12 and plasma folate status during pregnancy are the most significant influencing attributes for predicting BW.