Key points are not available for this paper at this time.
Clinical predictive ability was assessed in 21 physicians and medical students. Each was provided a standard set of twenty case histories in random sequence and asked to predict tlje probability of five standard outcome events for each case. Predictions were scored against the actual known outcomes of the patients. For these same patients, computer predictions were derived from analysis of data bank experience of the most similar prior patients. Physicians and medical students were provided with the computer predictions after making their initial judgments and were allowed to make revised predictions incorporating this new information. Revised predictions after computer consultation were significantly (p < .05) improved from initial values. Computer predictions initially scored approximately midrange among specialty physicians. Groups of physicians scored better than single physicians. Subjects scoring less than the computer uniformly improved their predictive accuracy after computer consultation. All subjects receiving computer consultation information modified their predictive estimates toward those of the computer. Subsequent modification of computer procedures for selection of more similar patients improved computer predictive accuracy. Availability of prognostic information derived from computer data banks can modify physician behavior and, if accurate, can enhance prognostic accuracy of physicians.
Dannenberg et al. (Mon,) studied this question.