Résumé : Introduction: In most European countries, hospital funding systems are moving towards prospective payment systems based on Diagnosis related groups (DRGs). This forfaitarisation is still very partial in Belgium. In those systems, tariffs are determined on the basis of actual costs (with some adjustments). Objectives: The objective of this study is to identify a limited number of parameters, which would allow for the calculation of nursing costs per stay, with a better method than a method based on the length of stay (LOS). Method: Digital pens were chosen to record activity information. The 23 RIM (Belgian nursing minimum dataset) factors have been valorised in minutes. For each patient, a number of minutes were calculated on the basis of 23 RIM factors. This time was used as cost driver to calculate nursing costs per stay. Different simulations to select the most representative variables (and weight) were done. Results: A patient hospitalized in cardiology consumes 1,209 minutes of nursing services, for a length of stay of 5,43 days, it represents 223 nursing minutes per day (3.71 hours/day). The nursing cost is € 551.14. Among the "financial" variables routinely available, the best combination is an allocation on the basis of the length of stay at 83%, drug charges at 1% and total charges at 16%. This combination gives an mean square error (RMS) of 66.28%. This combination is better than an allocation based on the length of stay only (RMS of 75.16%). 4 parameters based on RIM (nursing minimum dataset) can reduce the RMS by 62%. Conclusion: The use of 3 financial variables in addition to the length of stay gives a better result than the length of stay only. It is still possible to easily improve the calculation of nursing costs. To do so, it would be necessary to record per day the first four dependence RIM items.