In this paper, we describe an approach to make case-based reasoning methods appropriate for medical problems. From the class of therapeutic problems we have chosen calculated antibiocs therapy advice for patients in an intensive care unit who have developed an infection as an additional complication. As advice is needed quickly and the pathogen is not yet known, we use an expected pathogen spectrum based on medical background knowledge and known resistances, which both will be adapted to the results of the laboratory. Case-based reasoning retrieval methods provide the advice for similar previous patients. The previous solutions are adapted to be applicable to the new medical situation of the current patient. Because of the large and continously increasing number of cases, we use prototypes as a structural aid. We present some experimental results of our studies on the performance of our prototype design.