Birth weight prediction from remote ultrasound examination

Abstract

From 259 cases studied, 245 were selected in which a live-born infant was delivered within 35 days of a complete fetal ultrasound evaluation. Multiple linear regression using the least-squares method enable us to generate an equation that incorporated lapse time (examination-to-birth interval) with the natural logarithm of head circumference, femur length, and abdominal circumference to estimate birth weight. With a lapse time mean of 16 +/- 11 days and a range of zero to 35 days, the generated equation accurately predicted birth weight (R2 = 0.84; P less than or equal to .0001). For all birth weights, the mean error was -15 +/- 306 g, the percent mean error was 0.51 +/- 10.2%, and the mean absolute error of the estimate was 82 g/kg birth weight. This accuracy was maintained across the full range of lapse time observed. For examinations performed within one week of delivery (N = 71), this formula more accurately predicted birth weight than five existing static formulas tested. The accuracy observed during model development was confirmed during testing upon 167 non-model cases. The accurate prediction of birth weight from remote ultrasound data is possible when lapse time is included in the predicting equation. The clinical value of this model is suggested when ultrasound is unavailable or unreliable.