Medical decision making and the retrieval of medical databases

Searching for a diagnosis D in a group of patients with the help of a diagnostic test T can be compared with searching this diagnosis out of a patient-oriented medical database. The quality index 'precision' is an estimator of the Bayesian postretrieval probability P(D/T) and is dependent on the unconditional probability P(D), which can be interpreted as a priori precision. The precision value by itself is unimportant, only the difference P(D/T)-P(D) matters. Such a difference represents an increase in knowledge and can also be expressed by the information theory. Further analyses show a bijective conjunction between the Bayesian and information-theoretical approaches and lead to an improvement of retrieval strategies similar to the optimizing of diagnostic tests.<>