Neonatal anthropometric measurements to predict birth weight by ultrasound.

Abstract

To develop a more accurate ultrasound birth weight (BW) model using neonatal anthropometric measurements. Two hundred thirty-one newborns were evaluated. Measurements included weight; head, chest, and abdominal circumferences (umbilicus and liver), humerus, and femur lengths. Infants were randomly assigned into two groups (G(1) and G(2)). Anthropometric measurements that are obtainable by ultrasound were generated from G(1). Stepwise regression and a bootstrap analysis were used to create the prediction models. The models were validated using G(2). The final stepwise regression model included FL and circumferences of the head, chest, and abdomen. The correlations were: G(1): R(2)=0.91, p<0.001; G(2): R(2)=0.90 p<0.001. There was no difference between derived and actual BW in G(1) (p=0.42) or G(2) (p=0.28). The mean absolute percent error between the prediction model and actual BW was 3.8%. Neonatal anthropometric models are strongly predictive of actual BW. This model will be tested prospectively using ultrasound to predict fetal weight.