5 cognitive biases common to radiology—and how to beat them back

Twitter icon
Facebook icon
LinkedIn icon
e-mail icon
Google icon
 - MindGears

Cognitive bias accounts for as much as three-quarters of malpractice suits filed against radiologists, according to findings published in 2013. A new paper lays out some ways rads can leverage quality-improvement (QI) projects to help steer clear of such mental shortcuts before they lead to missed or inaccurate diagnoses.

“Cognitive biases are best dealt with by raising awareness of potential pitfalls, which will enable radiologists to recognize and override their biases when appropriate,” write Cindy Lee, MD, of UC-San Francisco and colleagues.

Their paper, published online Jan. 28 in the Journal of the American College of Radiology, recommends QI interventions that can cut errors caused by five common cognitive biases.

1. Framing bias

Clinical context is essential in informing image interpretation. However, it can paradoxically prompt the radiologist to make a wrong call. Lee and colleagues note that such misleading “framing bias” is more likely when a patient’s medical history is incomplete.

“The availability of clinical histories can be improved by having technologists collect information directly from patients to supplement referring clinicians’ histories,” the authors write. “Other QI ideas include creating a display of patient history customized for radiologists that is aggregated from the electronic medical record and PACS.”

2. Availability bias

This the authors define as the preference given to diagnoses that were more recent, more memorable or observed personally. They offer as an example a scenario in which a radiologist who missed a breast tumor, was sued for the error and now recalls more patients from screening mammography for additional workup.

“Objective incidence rates for each diagnosis could help prevent the over- or undercalling of diagnoses,” Lee and colleagues write. “Similarly, local performance could be compared with national performance benchmarks.” Meanwhile, participation in clinical data registries can provide “ongoing, standardized performance metrics to … member facilities, including recall rate, cancer detection rate and biopsy cancer yield.”

3. Satisfaction of search

It’s not unheard of that a radiologist catches an abnormality and assumes the search is over, causing the missing of additional or incidental abnormalities.

QI projects targeting such satisfaction-of-search errors “may include adopting a systematic approach or checklists for image interpretation,” the authors write, adding that the use of structured reporting, for example, can serve as a checklist and improve consistency. “Morbidity and mortality meetings may reveal common misses, and this information could be captured and used to design specific QI interventions to reduce these errors,” they write.

4. Premature closure

When a radiologist synthesizes all available information and makes a case-closed diagnosis—and in the process is remiss in considering additional differentiating possibilities—he or she runs the risk of prematurely concluding, for example, that a new lung mass in a patient with colon cancer is metastatic rather than primary lung cancer.

“QI projects to minimize premature closure may include radiology decision tools that will generate a list of probable differential diagnoses based entirely on lesion descriptors," the authors write. “Another QI project that can help decrease premature closure is a feedback loop from pathology to the interpreting radiologist. … This idea was conceived as early as 1978 but has yet to gain in popularity and be assessed scientifically in terms of its learning value for radiologists.”

5. Anchoring bias

Many if not most humans assessing a situation tend to rely heavily on the first piece of information to come before them—the “anchor” info—and then stick with it even when new information arrives to challenge it.

“QI projects could focus on avoiding early guesses, obtaining second opinions when appropriate, reconsidering the first diagnosis when the patient’s symptoms worsen or fail to respond to treatment as expected … and resisting the assumption that the most common diagnosis is always correct,” the authors write.

Lee et al. conclude by pointing out that diagnostic errors and their contributory cognitive biases frequently go overlooked by QI projects, “probably because it is difficult to obtain reliable and objective measurements of their effects on clinical outcomes.”

“The field of cognitive science is rapidly growing as we begin to understand the human interpretation process,”