'Pseudodiagnosticity' in an idealized medical problem-solving environment

Sixty-five senior medical students chose symptom information that would allow them to assess which of two diagnoses was more appropriate for hypothetical patients. Although Bayes' theorem should have governed their data selection, 83 percent of the subjects did not choose the symptom information required for Bayesian computation. Instead, they showed an overwhelming tendency to seek data relevant to a single disease, while ignoring information related to an equally plausible alternative diagnosis. The tendency for subjects to select diagnostically irrelevant information in such tasks has been labeled "pseudodiagnosticity." The effect result from the difficulty of simultaneously evaluating the relevance of a single symptoms in relation to single diagnosis. Medical educators might incorporate classroom demonstrations of the pseudodiagnosticity effect in order to increase students' accuracy in differential diagnosis.