Cognitive bias and medical imaging: 3 things every radiologist should know

Heuristics, or mental shortcuts based on past experiences, can help decisionmakers get things done at a rapid rate—but they can also lead to cognitive bias and significant mistakes. A new study published in the American Journal of Roentgenology examined how heuristics and cognitive bias impact image interpretation in radiology.

These are three key takeaways from that study:

1. Heuristics can be helpful in radiology—until they’re not.

“Physicians rely on these shortcuts (heuristics) in reasoning to minimize delay, cost, and anxiety in our clinical decision making,” wrote authors Jason N. Itri, MD, PhD, Wake Forest Baptist Medical Center in Winston Salem, North Carolina, and Sohil H. Patel, MD, University of Virginia Health System in Charlottesville. “Heuristics used in the interpretation of imaging studies are generally helpful but can sometimes result in cognitive biases that lead to errors.”

In an example provided by the authors, an abdominal radiologist sees “enlargement of the pancreatic head with biliary and pancreatic duct dilation in an older patient with right upper quadrant pain and jaundice” and, based on previous patients, concludes that there may be pancreatic adenocarcinoma. Biopsies and a Whipple procedure, however, revealed the patient had autoimmune pancreatitis.

“Although the conclusion of pancreatic head cancer may have been correct for most patients, failed heuristics led the radiologist to overlook the possibility of the less common diagnosis of autoimmune pancreatitis, in which the morbidity and mortality associated with the Whipple procedure could have been avoided,” the authors wrote.

2. There are more than one type of cognitive bias—and they can all have a negative impact on interpretations.

The authors explored several types of cognitive bias in their research, highlighting the true complexity of the issue.

Availability bias, for instance, is when an individual is “unduly influenced by easily recalled experiences.” By focusing more on peer-reviewed publications and conferring with colleagues, Itri and Patel explained, radiologists can work to avoid availability bias.

Reading a colleague’s radiology report can also result in bias, known as alliterative bias. If that other radiologist made an error, alliterative bias could lead to that same error being made by another radiologist. “Radiologists should not become overly reliant on prior reports and keep an open mind to diagnostic possibilities other than those that have already been suggested,” the authors wrote.

In addition, Attribution bias is when characteristics are attributed to something because it belongs to a certain class or category. “Consider the interpretation of CT and MRI of the abdomen in patients with cirrhosis and hepatocellular carcinoma,” the authors wrote. “Commonly described routes of extrahepatic spread of hepatocellular carcinoma include the lung, abdominal lymph nodes, bones, diaphragm, and adrenal glands. Radiologists who interpret these examinations may not know that hepatocellular carcinoma can disseminate into the peritoneum and can subsequently miss peritoneal implants because this cancer is not typically associated with intraperitoneal spread, as opposed to cancers of the ovary, pancreas, and stomach. Reviewing imaging studies without knowing the clinical indication or patient demographics may help address attribution bias.”

Structured reporting templates, Itri and Patel added, can also help reduce the chance of attribution bias taking place. If a known blind spot exists, the report could automatically mention that blind spot when the radiologists reaches a certain point in his or her report.

The authors covered other types of cognitive bias as well, including blind spot bias and regret bias. As radiologists learn about each type of bias, it gives imaging providers a better chance of producing low error rates and delivering a significant amount of value to patients.

3. Imaging groups should develop a strategy to reduce biases and learn from mistakes.

Leaders looking to reduce the influence of cognitive bias need to implement a group-wide plan, according to Itri and Patel.

“The key components of this strategy are processes to identify errors, analysis of errors for systematic causes and biases, using what is learned from these errors to develop educational content and systematic improvements, and developing a culture of quality that promotes reporting and learning from errors,” the authors wrote. “Implementing an educational curriculum for trainees, radiologists, and staff is a foundational element of this strategy, because awareness is arguably one of the most effective tools we have to reduce errors.”