Table 2 Types of cognitive errors. Adapted from Croskerry [23] unless cited otherwise

Satisfaction of search

Perhaps the most significant cause of diagnostic error, once a diagnostic finding is met with, the search stops, with the potential of missing a second finding which might be even more significant than the first.

Availability bias

Recent experience will modify the threshold of diagnosis for that condition, if a certain condition has been seen recently; the tendency is to think of that in a new patient. Even more importantly, if a condition has been missed and brought to the notice of the physician, the next or next few patients will certainly be assessed on those lines. Similarly, if a similar finding has not been encountered for a long time it might not be considered as easily.

Capture

A more frequently used schema captures or takes over from a similar but less familiar one [19]; for example, if an imaging routine involves looking at the left flank after the right flank, and the patient points to a mass in the mid abdomen that is seen after looking at the right flank, the left flank can be missed. These are also called post-completion errors and are most frequent when the interruption occurs just before the step that needed to be completed [45].

Gambler’s fallacy

Thinking that if a series of patients of the same kind have been seen sequentially, the chances of the next patient having the same condition diminished, something like imagining that if on a coin flip you get ten heads in a row, the chances of an 11th head is reduced.

Aggregate bias

Thinking that an individual physician’s patients are somehow unique and do not display the common features of a particular process, this can lead to false diagnoses and unnecessary procedures. A tendency to neglect or acknowledge the base rate or the local prevalence of that condition distorts Bayesian reasoning, although sometimes increasing the previous probability of a condition that might enable diagnosis or a rare or rarely encountered condition.

Ascertainment bias

The patient has non-medical attributes that touch upon the physician’s own prejudices, biasing him in a certain direction; overweight people [60], women [61], minorities [62] may all pay the price of a visceral bias held by the physician.

Anchoring

Making an impression very early in the diagnostic process and then refusing to change it as new evidence becomes available. This leads to confirming when the evidence that supports the initial opinion is acknowledged while that to the contrary is ignored. The diagnosis made gains its own momentum and it becomes more and more difficult to think of alternative possibilities. Finally, closure takes place and further thinking about a possible alternative diagnosis stops. One of the reasons might be the mental investment already made in this diagnosis and the reluctance to see the work go to waste.

Alliterative errors

A previous report of another radiologist or even the same reader will influence the current reading. If a lesion has been ascribed benign findings previously, it will be similarly judged, and if a significance has been assigned so will it on subsequent readings [63].

Overconfidence

Tendency to believe that one knows more than one really does, prompting action on incomplete information, intuition or hunches.

Framing bias

The patient’s diagnostic possibilities are restricted by the referral situation, or the question that is asked. For example, a referral from a gastroenterology service might cause a focus on the liver and gut but ignore the other viscera; or a differential diagnosis of a finding might be limited to only or mostly gastroenterology.

Pressure to report

There is a “need” to find something wrong with the patient, so findings, often insignificant or even “invisible” are reported in a language that is ambiguous but might be misinterpreted as something significant (local data).

Misdirection

Similar to the framing bias; commonest where the patient interacts with the imaging physician and points to the wrong site or emphasises a minor symptom leading to a less detailed evaluation of the region where the significant pathological condition might actually lie. An example is a woman who will not be forthcoming about a gynaecological symptom during an ultrasound examination and will insist that her presenting the complaint is elsewhere. Walk-in patients and those without a proper referral requisition are most prone to creating this bias (local data).