Digital tool could decrease reporting variance among radiologists

A computer-based reporting tool could be reducing report variation among radiologists, according to data published this month in the Journal of the American College of Radiology.

Interreader variability—the idea that radiologists interpret and report test results differently—is being addressed in the medical space, but many aspects of the issue remain untouched, first author Benjamin Wang, MD, and colleagues wrote. Though clinicians have advocated to reduce variability before, largely through the implementation of systems with uniforum nomenclatures, the need for time and reader training has limited the success of those ideas.

“Improving radiology report quality in our current age of PACS and speech recognition has been an area of active research,” Wang and co-authors said. “Interreader variability in reporting degenerative findings identified on lumbar spine MRI is a well-recognized issue in clinical practice and is a topic of much investigation in the literature.”

Lumbar spine injuries like degenerative disc disease are particularly pertinent when it comes to reporting variability, the authors wrote, because complex images can be hard to link to concrete symptoms and radiologists differ in what they choose to omit or report at each spinal level. In an effort to reduce that variability, Wang et al. recruited five fellowship-trained musculoskeletal radiologists to examine 30 lumbar spine MRIs using both a traditional free dictation approach and a computer-assisted reporting (CAR) tool developed by the researchers. The blind evaluations were situated a month apart.

The authors’ CAR tool supported the use of standard nomenclature, the study stated, and included galleries of representative MRI examples. The highly graphic method was designed to translate spoken user input into a physical report without the clinician having to do anything.

Wang and colleagues found a statistically significant decrease among the five radiologists in percentage disagreement for neural foraminal stenosis and facet joint osteoarthritis in reports created by free dictation compared with those generated with the CAR tool. Other aspects of spinal imaging— assessment of central canal stenosis, lateral recess effacement, disc herniation, disc bulge or herniation location—saw no statistically significant differences between experiments.

The authors wrote that decreased reader variability among experts made sense in the cases of neural foraminal stenosis and facet joint osteoarthritis.

“These were the two imaging features with the largest interreader variability with conventional free dictation, so it is fitting that CAR had the greatest impact on reducing the interreader variability for these two features,” they said.

But even with decreased interreader variability, the CAR system isn’t perfect. More uniform reports come at the cost of increased reporting time, Wang and co-authors wrote, since CAR reporting times averaged 3.4 minutes per lumbar level, and editing the final transcript would likely take longer due to dictation errors.

“Although decreasing interreader variability is a potential benefit of a CAR tool, the true clinical impact of improving interreader variability remains unknown,” the authors said. “Future directions of investigation include testing the CAR tool on a larger consecutive sample of lumbar spine MRI examinations and determining the impact of improved interreader variability on patient outcomes and symptom correlation.”