A Cognitive Autopsy Approach Towards Explaining Diagnostic Failure

Diagnostic failure has emerged as one of the most significant threats to patient safety. It is important to understand the antecedents of such failures both for clinicians in practice as well is those in training. A consensus has developed in the literature that the majority of failures are due to individual or system factors or some combination of the two. A major source of variance in individual clinical performance is cognitive and affective biases; however, their role in clinical decision making has been difficult to assess partly because they are difficult to investigate experimentally. A significant drawback has been that experimental manipulations appear to confound the assessment of the context surrounding the diagnostic process itself. We conducted an exercise on selected actual cases of diagnostic errors to explore the effect of biases in the ‘real world’ emergency medicine (EM) context. Thirty anonymized EM cases were analysed in depth through a process of root cause analysis that included an assessment of error-producing conditions (EPCs), knowledge-based errors, and how clinicians were thinking and deciding during each case. A prominent feature of the exercise was the identification of the occurrence of and interaction between specific cognitive and affective biases, through a process called cognitive autopsy. The cases covered a broad range of diagnoses across a wide variety of disciplines. A total of 24 discrete cognitive and affective biases that contributed to misdiagnosis were identified and their incidence recorded. Five to six biases were detected per case, and observed on 168 occasions across the 30 cases. Thirteen EPCs were identified. Knowledge-based errors were rare, occurring in only five definite instances. The ordinal position in which biases appeared in the diagnostic process was recorded. This experiment provides a baseline for investigating and understanding the critical role that biases play in clinical decision making as well as providing a credible explanation for why diagnoses fail.