Computer prediction of neonatal outcome and comparison with assessments by physicians and midwives

A simple statistical model for the prediction of 1' and 5' Apgar scores is implemented by computer and tested prospectively in 1000 deliveries. Data consist of model predictions in early labor and prior to delivery. Independent predictions of neonatal outcome by physicians and midwives are also included. While the predictive value of the original model is 81.6%, the overall predictive value of the computer model and of the physicians' assessments is only 67%. This finding is consistent irrespective of the time of prediction. Several factors are analyzed for possible association with the decrease in the model's predictive value. It is concluded that the development of a simple yet highly predictive model cannot be achieved. While measures for assessing the usefulness of the model need to take the effect of medical management into account, model optimization and the use of additional variables are indicated in order to improve sensitivity and specificity.<
>