Background. Intraventricular hemorrhage (IVH) incidence is used to assess neonatal therapy, and to make interhospital quality assessments. Unbiased assessment is complicated by the amount of confounding factors .Subjects. We investigated if an artificial neural network (ANN) is able to forecast a severe IVH accurately at an early stage. Methods. Data sets (38 first-day-of-lifestandard items) were available from 872 preterm neonates (ga < 32 wks, bw < 1500 g). Patients were randomly assigned to a training or validation set (50%/50%). Using training set data, multiple logistic regression analysis helped identify significant predictor variables. An ANN was trained on that variables of the training set data. Using the validation set input data, both models delivered individual estimates of the probability for severe IVH to occur. Results. Seven independent predictor variables were identified. The area under the Receiver-operating-characteristic curve was 0.86 for the logistic regression model and 0.90 for the ANN (p=0.02). Adjusted for 90, 85, 80 and 75% specificity, the sensitivity of an ANN-based prognosis was significantly greater than that of the logistic regression model (p<0.05).Conclusion. Due to its ability to give an accurate prognosis based solely on first-day-of life standard items, a trained ANN qualities as a tool for local and regional quality control.