The purpose of this study is to refine and validate an educational program for improving the accuracy of physicians' diagnostic judgments. The educational intervention will consist of three components: 1) microcomputer-based feedback of diagnostic accuracy on simulated patients; 2) a lecture; and 3) ongoing feedback of easily acquired information about diagnostic accuracy in actual patient care. The study is a controlled, preintervention-postintervention trial. Both the accuracy (calibration) and process of physicians' subjective estimates of disease probability will be measured. The program will be evaluated for two clinical problems: 1) streptococcal pharyngitis and 2) pulmonary embolus. In the pre- and postintervention periods, clinical data will be prospectively collected on eligible patients prior to performance of the relevant reference diagnostic test. For each patient, the physician will estimate the probability that the disease of interest is present. Diagnostic judgments will be compared with the results of the reference test. The effectiveness of the intervention will be determined by comparing the amount of improvement in diagnostic process and accuracy seen in the intervention group with that seen in the control group. In the intervention phase, interactive microcomputer-based feedback will compare physicians' probabilistic judgment for each case to probabilities estimated by a validated statistical model. At the end of a series of cases, graphic feedback will compare the diagnostic cues and weights assigned by the physicians to those assigned by the statistical model. Microcomputer-based feedback constitutes an inexpensive and generalizable decision aid in settings for which there are large enough data bases to develop statistical models. These methods should impact on the teaching of diagnosis and on research in the processes of diagnostic judgment.
