This paper deals with a prediction model for the nursing care demand in a surgical department. The model is based on a description of disease dynamics by semi-Markov stochastic processes and on the identification of typical care requirements associated with each state of disease. The prediction model has been implemented on a decentralized information system using a small desk computer, and the methodology has been tested in both a simulated and real environments.

The results show a good predictive capability on an horizon of several days. This should permit the nursing care managers to improve their resource allocation.