Medical error is among the top ten causes of death in the United States, and while there are different forms and sources, diagnostic error is one of the most significant and consequential. Graber et al, 2002 classifies diagnostic error as one of 3 types:
- No-fault error
- System error
- Cognitive error
What is “cognitive error”?
Cognitive error encompasses all aspects of the clinician’s thought process that lead to diagnosis and intervention. With the inherent difficulty of reporting or tracking this form of error, it presents a great challenge for clinicians to understand and address. Also, factors that lead to cognitive errors are often subtle and ingrained in our daily schemata for patient assessment, thereby explaining the resistance encountered with attempts to create clinician awareness and generate solutions.
Cognitive error in medicine cannot be discussed without addressing heuristics. Heuristics refer to cognitive techniques for problem-solving that are based on an intuitive and subjective assessment of probability. They are extremely common, and arguably invaluable, in medicine as they lighten the cognitive load of decision making and improve clinical efficiency, especially in the setting of uncertainty.
Imagine for each patient complaint and presentation if you had to consider the prevalence (base-rate) of EVERY item on the differential diagnosis and form a numerical pre-test probability. For diagnoses that now meet the threshold to test, the next step would be to consider the performance characteristics of every lab test and imaging study ordered (read more about Testing Thresholds).The last step would then be to devise a post-test probability before management. This is often impractical and inefficient in medicine. Heuristics give us intuitive shortcuts and rules-of-thumb that work very well… most of the time.
Problems with heuristics
While heuristics are useful and cannot be avoided, we must be aware of potential biases and sources of error. Two common heuristics are Representativeness and Availability.
Representativeness refers to estimating the likelihood of a diagnosis based on how well the patient fits the prototype for that condition. For example, the likelihood of renal colic is deemed higher in the patient with sudden-onset intractable flank pain than in the patient with insidious mid-back pain. Given that typical presentations generally outnumber the atypical, this would work most of the time. One problem is that with Representativeness, there is no consideration of disease prevalence. Prototypes of rare diseases may be less accurate given the smaller sample size observed, and for very common diseases there may be larger numbers of atypical presentations. This would lead to underestimation of pre-test probability, perhaps to the extent that the threshold to test is not reached, or the wrong focused history is obtained.
Availability refers to estimating the probability of an event based on how readily it comes to mind. Considering that commonly occurring events are more likely to be recalled, the most common diagnoses will get prompt consideration. A potential pitfall is that higher-acuity or more publicized (M&M, Grand Rounds) diagnoses are easily recalled. This creates an overestimation of prevalence of disease and falsely elevated pre-test probability.
Heuristics and cognitive error are especially relevant to emergency physicians. The ED has the highest prevalence of diagnostic uncertainty, as well as the highest “decision density” in medicine (number of clinical decisions made per unit time). When limited time and resources are combined with practically unlimited patient volume, the emergency physician is constantly faced with the trying to balance efficiency versus accuracy.
Why should I care about heuristics and cognitive errors in diagnostic reasoning?
These topics are not new or ground-breaking subjects. Many experienced physicians have eventually intuitively developed an awareness of cognitive errors and developed strategies to combat them, partly through trial and error. However, it is a lengthy process with highly variable results when this unstructured approach is used. Given the high stakes environment that we work in, one can understand why a formal and systematic approach to studying cognitive error and developing strategies to combat it might be a better option.
Cognitive errors is one of the most preventable forms of error, not requiring lobbying, special committees, or targeted spending. All that is asked is that we begin to think about how we think.
Great articles by Dr. Croskerry
Much scholarly work already exists on cognitive error, perhaps most notably by Dr Pat Croskerry, an EM professor at Dalhousie University, Canada. See his list of Pubmed references.