When doctors meet numbers.

A Statistical Skills Self-Assessment Questionnaire (SAQ) was developed using hypothetical clinical questions to explore respondents' mastery of vocabulary and rules of inference that seem relevant to the use of quantitative information. The SAQ was administered to 281 subjects, including 36 medical students, 45 interns and residents, 49 physicians engaged in research and 151 physicians in full-time practice. All groups of subjects showed frequent lack of consensus on the meaning of terms in common use (e.g., "false-positive rate" and "p values") and unfamiliarity with some important principles in quantitative inference (e.g., the Central Limit Theorem and Regression to the Mean). Subjects often seemed willing to draw conclusions unsupported by available data. Performance on the SAQ was inversely correlated with length of time since graduation from medical school, and practicing physicians tended to err more frequently than the other three groups.