The Bayesian approach to evaluation of diagnostic data.

Following review of the basic principles of Bayes' theorem, its application to diagnostic procedures is discussed. It is asserted that Bayes' theorem can be applied to clinical diagnostics, provided that disease prior probabilities can be calculated and it can be assumed that a particular patient and those included in the calculation of prior probabilities represent two samples of the same homogeneous population. Problems associated with both these requirements are briefly reviewed. With the aid of some clinical examples, the usefulness of Bayes' theorem is detailed in: (a) evaluating the information obtained through the results of diagnostic studies, (b) evaluating the level of diagnostic reliability achieved with obtained data and (c) a priori evaluation, i.e. before knowing the result, of what kind of information a test can supply. It is concluded that Bayes' theorem is a powerful tool in the rational control and improvement of diagnostic procedures, if not relevant in the formulation of the diagnosis of a single patient.