Clinicians glean more from structured radiology reports

Several studies in the literature have found that clinicians would prefer to receive structured instead of unstructured radiology reports. But the structured format also yields valuable benefits such as better comprehension of the key information contained in the report, an Irish research team recently concluded.

After conducting an online survey of nearly 150 clinicians, a group led by Dr. Bryan W. Buckley of Mater Misericordiae University Hospital in Dublin found that the respondents had significantly better recall of critical and noncritical findings as well as the correct diagnosis after reviewing structured radiology reports than they did with unstructured radiology reports.

"[A structured report] is a safer and more efficient transfer of information, and seems more suited to complex studies [than unstructured reports]," he said. "And then there are of course the administrative benefits such as data mining, the low cost of implementation, billing, and research."

Buckley shared the results in a presentation at ECR 2017 in Vienna.

Assimilating information

Unstructured reports can have a free-text structure and no headings, while structured reports utilize a uniform, consistent template and itemized headings, Buckley said. A prior study in the literature found that 85% of clinicians prefer structured reports and 50% of the clinicians believed the radiologist had not looked at any organ that wasn't mentioned in the report.

Structured reporting has become commonly used in surgical and pathology reports, but there hasn't been much research on how structured reporting in radiology affects how referring clinicians assimilate information in reports, Buckley said. As a result, the team sought to quantify the recall of radiology reports by healthcare workers from both unstructured and structured reports.

They created hypothetical unstructured reports for a coronary CT angiogram and abdominal CT study, as well as hypothetical structured reports for a brain MRI and a thorax CT using the RSNA's open-source structured reporting templates. The four radiology reports included one to two critical findings, as well as two to three noncritical positive findings.

Using Survey Monkey, the researchers created an online survey and included a link in email newsletters to the membership of the Royal College of Physicians of Ireland and the Royal College of Surgeons in Ireland -- encompassing 6,405 physicians, surgeons, and medical students. Radiologists were intentionally excluded from the survey.

There was no time limit for the survey. Respondents read each report and then were presented with a multiple-choice question on some aspect of the report's findings. Multiple answers could be selected. They were unable, however, to return to the report once presented with the question, Buckley said.

Of the 207 who clicked on the link in the newsletter, 148 (71.4%) completed the survey and took an average of 21 minutes to complete it. The researchers used the Wilcoxon signed rank test and an independent Kruskal-Wallis test for data analysis.

Recall rate of findings by report type
	Unstructured radiology report	Structured radiology report
Critical findings	65.1%	82.7%
Correct diagnoses	81.9%	93.8%

"This meant that critical findings were missed in almost 35% of unstructured reports, compared with over 17% in structured reports," Buckley said. "Respondents incorrectly selected the diagnosis in 6.2% of the structured reports, compared with 18.1% of the unstructured reports."

All of these differences were statistically significant.

Worst-performing report

The unstructured report for coronary CT angiograms produced the worst performance of the four hypothetical reports in the study. Only 51% of respondents correctly identified both critical findings: a right upper lobe nodule and left ventricular apex infarct. In addition, 63.1% correctly recalled noncritical findings and 23.6% of responses were incorrect.

"[That] would be in agreement with some of the research that is out there already on unstructured CT coronary angiograms," Buckley said. "It seems that the more complex the study, the greater the benefit that you can yield from having a structured report."

Multivariate analysis showed the demographics of the response had no effect on the results.

"Structured reporting lead to superior recall of all findings, superior recall of critical findings, and less incorrect diagnoses," he said.