Guided Reflection Interventions Show No Effect on Diagnostic Accuracy in Medical Students

Background: Guided reflection interventions, in an effort to reduce diagnostic error, encourage diagnosticians to generate alternative diagnostic hypotheses and gather confirming and disconfirming evidence before making a final diagnosis. This method has been found to significantly improve diagnostic accuracy in recent studies; however, it requires a significant investment of time, and psychological theory suggests the possibility for unintended consequences owing to cognitive bias. This study compared a short and long version of a guided reflection task on improvements in diagnostic accuracy, change in diagnostic confidence, and rates of corrected diagnoses. Methods: One hundred and eighty-six fourth- and fifth-year medical students diagnosed a series of fictional clinical cases, by first impressions (control condition) or by using a short or long guided reflection process, and rated their confidence in their initial diagnostic hypothesis at intervals throughout the process. In the “short” condition, participants were asked to generate two alternatives to their initial diagnostic hypothesis; in the “long” condition, six alternatives were required. Results: The reflective intervention did not elicit more accurate final diagnoses than diagnosis based on first impressions only. Participants who completed a short version of the task performed similarly to those who completed a long version. Neither the short nor long form elicited significant changes in diagnostic confidence from the beginning to the end of the diagnostic process, nor did the conditions differ on the rate of corrected diagnoses. Conclusions: This study finds no evidence to support the use of the guided reflection method as a diagnostic aid for novice diagnosticians, who may already use an analytical approach to diagnosis and therefore derive less benefit from this intervention than their more experienced colleagues. The results indicate some support for a shorter, less demanding version of the process, and further study is now required to identify the most efficient process to recommend to doctors.