Background. In neonatology, the early prediction of length of stay (LOS) may help in decision making. Subjects. We retrospectively studied the accuracy of model based prediction. 2144 preterm neonates were randomly assigned to a training (75%), or a validation patient set (25%). 38 first-day-of-life items (input data) and the date of discharge (output data) were routinely available. Training set data were used i) to identify input items with impact on LOS (input variables), ii) to perform multiple linear regression analysis (MR); iii) to establish a MR prediction model: and iiii) to train an artificial neural network (ANN) on those selected variables .Results. At discharge postconceptional age ranged from 33.7 to 74.8 weeks, and the body weight from 1850 to 6280 g. The correlation coefficient between birth weight and LOS was -0.77. Fed with validation set data, predicted LOS obtained from MR and ANN was compared individually with actual LOS. Both correlated highly (for MR, r = 0.85-0.90; for ANN, r = 0.87-0.92).Conclusion. By allowing for matching resources to predicted requirements, and for prospectively monitoring differences in predicted and achieved date of discharge, ANN and MR predictions might become powerful tools in the fields of clinic management and quality control.