The management of a department of cardiology has to plan the capacity of both elective and non-elective patients. Heart failure (HF) patients are admitted to the hospital in a non-elective way. The precision with which the capacity needed for non-elective patients can be predicted determines the degree of flexibility in planning the admission of elective patients. In this study we want to determine how accurately we can predict the bed occupancy of HF patients using a simulation model. Data of the year 2000 were used to obtain the necessary probability distribution functions. Data from the year 2001 were used for determining the prediction accuracy. The results show that the arrival of new HF patients can be adequately predicted. However, the bed occupancy by new and especially current patients is predicted less accurately. Still in 70% (90%) of the days of a 5-day-prediction interval the error is at most one (two) bed(s). The results may improve if the cardiologist is asked to predict the length of stay of the current patients.