Accuracy of physicans' predictions of cholecystography results

In the decision theory model of medical diagnosis and treatment, optimal choice of diagnostic tests requires accurate estimation of the probability that a given test will be positive. We assessed the ability of physicians to estimate the probability that a specific test (cholecystography) would be positive. For 102 patients, the predicted number of gallstone cases, 35.8, was significantly greater than the observed number, 15 (p less than 0.001), even though 13 of the 15 observed cases were patients with previous radiographic evidence of gallstones. The overestimation of probability of positive tests casts doubt on the correctness of decisions regarding selection of diagnostic tests when these decisions are derived from probability estimates based on intuition or expert opinion.