Decision support probably isn’t to blame for much ‘imaging leakage’

It’s understandable that hospital systems might balk before setting up ordering providers with order-entry clinical decision support (CDS) tools that, in theory, stand to drive imaging referrals outside the system. After all, referrers have been known to get around CDS recommendations they don’t like by simply sending patients elsewhere.

The fear of such business “leakage” isn’t entirely unfounded but is largely overblown, according to a study running in the June edition of the Journal of the American College of Radiology.

Harvard radiologist Anand Prabhakar, MD, and colleagues at Massachusetts General Hospital queried their home institution’s outpatient CDS system over a three-year period (2011 to 2013) for studies that received a low CDS appropriateness score and were canceled by the ordering physician.

Focusing on patients who met these criteria and were participating in risk-sharing contracts, the team cross-referenced imaging utilization reports in a risk-contract insurance payment database.

The goal was to figure out if, contra the CDS recommendation, the patients received outpatient imaging within 60 days of the order and, if so, where they received it.

The risk-shared insurance database contained an average of 63,378 patients who had 18,008 MRIs and 18,014 CTs, according to the study report.

Overall, 111 imaging exams received a low appropriateness score in the risk-shared patient population and were performed within 60 days despite the low score.

Of these exams, 106 (95.5 percent) were performed within MGH’s parent system, Partners HealthCare (104 at MGH, two at affiliated institutions).

Only 5 of the 111 (4.5 percent) were performed outside of Partners.

Meanwhile total leakage was 31.2 percent (11,234 of 36,022 CT and MR exams), with 9.8 percent of patients going outside MGH but remaining within Partners.

A not-insubstantial 21.4 percent did indeed go to unaffiliated imaging facilities.

However, over the three-year period, the relative frequency of imaging leakage was consistent with other-than-imaging leakage—which has been shown in previous research to range in hospitals from 20 percent to 30 percent.

Plus the team recorded little variability in annual leakage over the study period, “suggesting that consistent factors may be contributing to leakage,” the authors write in their discussion.

Among the limitations in their study’s design, Prabhakar et al. acknowledge their single-system vantage point and their review only of cases that had low CDS scores.

“[I]f a general aversion to order-entry CDS drove imaging leakage, this fact would not be detected reliably, given the study design,” they write.

Still, the key takeaway is that decision-support tools for ordering providers “do not seem to drive out-of-network leakage of imaging studies,” the authors conclude.

Although leakage of imaging exams “was a significant occurrence among population health management patients, order-entry CDS systems did not seem to contribute significantly to this leakage,” they write. “Therefore, institutions can consider implementing order-entry CDS for both Medicare and non-Medicare populations, with less fear of driving substantial imaging revenue outside of their institution.”

The authors close with a call for further research to figure out what factors contribute to—and how best to plug the holes causing—imaging leakage.