Low Birth weight prediction based on maternal and fetal characteristics

Newborn size is an important indicator of infant survival and childhood morbidity [1] and appears to be related to subsequent risk of type 2 diabetes, hypertension, cardiovascular disease, and other disorders [2], [3]. Therefore, many studies have attempted to identify sources of variation in newborn size. The purpose of this study is to determine whether accurate prediction of term birth weight is possible based on maternal and fetal characteristics routinely measured remote from term. Multiple linear regressions is deployed to define which combinations of these variables are significant using real data collected in a maternity clinical and birth weight prediction equations are developed. The models are then used to predict the delivery weight for the Low Birth Weight (LBW) babies. The efficacy of the predication models are assessed and compared based on their mean and standard error of the predicted weights. The paper proposes two regression models based on some measurable characteristics of both mother and fetal. The models can explain 62.9% and 59.4% of the delivery weight variation for the low birth weight babies. The proposed models were then used to estimate the recorded weights together with their corresponding 95% confidence and predication intervals for the LBW babies. The results indicate that the most significant factors for the reduced regression model are head circumference, gestation age, and fetal length. The reduced model can explain 59.4% of the delivery weight variation for the low birth weight babies. While the regression model based on the above predictors as well as mother hemoglobin level, chest circumference and mother height and BMI can explain 62.9% of the delivery weight variation for the low birth weight babies.