Graber ML. — RTI International
Hess BJ, Lipner RS. — American Board of Internal Medicine
Thompson V. — University of Saskatchewan
Presented: Diagnostic Error in Medicine Conference, November 2012
According to the dual-process paradigm of clinical reasoning, medical diagnosis reflects some combination of intuitive (System I) and deliberate (System II) consideration. Experts, and probably most experienced clinicians, rely largely on the intuitive system because most new problems they see closely resemble problems they've seen before, and the vast majority of these diagnoses will be correct. It is known, however, that System I is to some extent error-prone, so the question becomes: Should physicians trust their intuition, or should they systematically reconsider their initial impressions?
To address this question, the keystroke response data were studied from a random sample of 500 residents who took the ABIM 2010 Internal Medicine Certification Examination. The exam was comprised of realistic clinical vignettes with multiple-choice single-best answers. The 80 questions regarding diagnosis were of focus. Level of cognitive skill (ability) was measured by selecting the bottom and top quartile performers on the overall exam. Complexity was assessed from the percentage of examinees who correctly answered each question, and the questions in the most difficult quartile (<60% correct answers) were compared to the questions in the easiest quartile (>80% correct). Multiple regression analysis was used to examine relationships between time to initial response and proportion of correct diagnosis made on the initial response, controlling for residents' ability.
Overall, residents changed their answers on 12% of all diagnosis questions. With the exception of low-ability residents answering complex questions, changing an answer from wrong to right was almost twice as likely as changing an answer from right to wrong. By examining residents' average time spent making their initial response to questions, results showed that easy questions were answered faster than difficult ones, and when residents spent more time on initial responses to more difficult questions, they tended to make more correct diagnoses (β= 0.14, p<0.001).
These results show that reflection is generally beneficial for both simple and complex vignettes. The findings contradict the traditional belief held by many test-takers to “just trust your intuition,” and support the hypothesis that routine reflection in clinical decision-making could reduce the likelihood of diagnostic error. Follow-up studies should explore how reflective reasoning can be fostered and invoked appropriately so that trainees can improve clinical decision-making, increase efficiency and ultimately decrease errors and waste in medicine.
For more information about this presentation, please contact Research@abim.org