Computerized diagnosis: implications for clinical education.

The advent of the small computer as a basic clinical tool will have a significant impact on clinical practice and medical training. The application of probability theory to clinical diagnosis has led to the development of several practical diagnostic programs which run on small computers. Expert systems--interactive programs which function as 'electronic consultants'--have now been successfully developed for a number of clinical situations. Experience with two of these, INTERNIST/CADUCEUS and MYCIN, has provided insight into problems and prospects for expert systems in medicine. Less complex programs, particularly those employing clinical prediction rules, and expert system shells, seem well suited for clinical environments. Although computerized medical diagnosis holds great promise as an aid to clinicians, its success will largely be determined by the quality of the information that clinicians provide for analysis. A brief review of the status of bedside diagnosis reveals that data-gathering strategies and techniques must be better understood. In order to take full advantage of computer programs for diagnosis, basic diagnostic skills must be more heavily emphasized in clinical training.