Abstract
If the objective of clinical medicine is defined as the maximization of the expected utility of the patient, a computer cannot be used in the problems of medicine until some logical structure is first developed. In such a structure, the doctor is seen as a decision maker moving through a decision tree while the patient's symptoms, signs, and findings can be represented as a set of elements of evidence which assigns him with a certain probability to a disease class. To make this allocation, a method of inference is necessary and if the costs of eliciting evidence are to be included, the method of inference has to proceed sequentially. Management decisions can only be formalized if the probabilities of their outcomes are known and the associated utilities of states of health can be estimated.