ABSTRACT
Decision making is a critical aspect of medicine. The two most important decisions to make about patients concern their diagnosis and how to manage their illness. Unless the first is accurate, the second is likely to be ineffective. In the majority of cases, decisions about the patient’s diagnosis are sound and the management effective. However, multiple factors may influence decisions around a patient’s diagnosis, and error is inevitable. There is considerable variation in diagnostic failure rates across the various disciplines within medicine, but the overall rate is put at 10–15% (Berner & Graber, 2008) with a significant associated mortality. It is estimated that 40,000–80,000 patients die annually from preventable diagnostic failure in US hospitals (Leape, Berwick & Bates, 2002), but this may well be underestimating the problem as autopsy studies in intensive care units alone have estimated the annual rate at 40,000 (Winters et al., 2012). The vastness of the problem has attracted the interest of the US National Academies of Sciences, Engineering and Medicine, which recently published an extensive report covering its major aspects (National Academies of Science, Engineering and Medicine, 2015). While some diagnostic failure is due to system issues (patient follow-up failures, lost reports, laboratory error and many others), much of it is due to problems with human reasoning and decision making. Early workers in this field noted: ‘a physician’s judgment is human judgment and susceptible to human limitations’ (Detmer, Fryback & Gassner, 1978, p. 682), and more recent expert opinion has attributed the root cause of diagnostic failure to ‘misplaced dependence on the clinical judgments of expert physicians’ (Weed & Weed, 2014, p. 13). When James Reason, author of the classic Human Error (Reason, 2000), was asked, ‘How is it that so many errors are made in medicine, I mean, it’s not rocket science, is it?’, he replied ‘No, it isn’t, it’s more complicated’ (Croskerry, 2012). Some degree of the complexity is illustrated in the following real case (Croskerry, 2008).
