Clinical decision support fails to reduce inappropriate imaging requests

A clinical decision support system failed to reduce the number of inappropriate imaging requests, according to the results of a randomized trial published Monday in JAMA.

Experts have touted CDS as a solution to reduce waste in the healthcare system—educating referrers about proper order selection while providing relief to overworked radiologists. European researchers sought to test this technology in a real-world setting, investigating its impact across 26 departments at three German university hospitals. 

However, participants saw no noteworthy difference in the amount of imaging requests, radiation exposure or costs, the randomized trial found.

“Notably, few physicians changed their initial imaging requests that were categorized as inappropriate, and, in rare cases, physicians even switched their order to less appropriate diagnostic tests,” lead author Stijntje W. Dijk, MD, with Erasmus MC University Medical Center, Rotterdam, the Netherlands, colleagues wrote Feb. 10. 

The study randomly assigned 13 departments to use the clinical decision support system while 13 more served as the control. Ordering physicians in the CDS group received information as to whether their requests were appropriate or unnecessary, with the study conducted between 2021 and 2024. A total of nearly 66,000 imaging requests were scored using the support system, the European Society of Radiology’s iGuide. 

Before implementing CDS, about 6.3% of imaging orders were classified as inappropriate in the control group versus 5.2% afterward. That’s compared to 7.6% in the study group before implementation and 6.4% afterward. Both groups saw similar decreases in inappropriate imaging requests, and there was a difference of about 1.3 percentage points, which the authors labeled as “not statistically significant.”

At the end of the study, all participating departments opted to discontinue use of the clinical decision support system. 

“Users expressed that the system lacked choice in indications, and the selection process was time-consuming and inefficient,” the study authors noted. “Despite the extensive coverage of the CDS, many items remained unscored. This aligns with a recent survey in which users found the system was not well-integrated, expressed that it did not help them choose the right imaging test, and reported not recommending the system to colleagues.”

Stijntje and co-authors speculated that the study setting may be one reason for its failure. They started with a low baseline percentage of inappropriate imaging requests (7% when they were expecting 20%), “leaving little room for reduction.” Also, the analysis did not attempt to ascertain why physicians made requests for imaging, against CDS’ suggestions. 

“Potential explanations include misclassification due to missing information, physician knowledge gaps, timing of feedback relative to the clinical process, timing of feedback relative to office hours, and the inability of the CDS to account for nuanced clinical scenarios and patient characteristics,” the authors added. “Qualitative research could provide deeper insights into these factors.”