Clinical decision support malfunctions are widespread and persistent, AMIA says
Clinical decision support misfires are commonplace but often hard to detect, according to a close examination of CDS systems at Brigham and Women's hospital in Boston published in the most recent Journal of the American Medical Informatics Association.
In addition, a separate poll of chief medical information officers found that the vast majority – 93 percent – had experienced at least one CDS malfunction, according to researchers. Two-thirds experienced system failures on an annual basis.
The study spotlighted four decision support glitches at Brigham and Women's.
In the first, an alert for monitoring thyroid function stopped working when an internal identifier for the drug amiodarone was changed in another system. In the second, an alert for lead screening in children malfunctioned when the rule to prompt it was edited inadvertently.
The third instance occurred when an EHR software upgrade "caused numerous spurious alerts to fire." The fourth, a mistake in an external drug classification system, led to inappropriate suggestions that antiplatelet drugs be given to patients already taking one, such as aspirin.
"Significant and mounting evidence suggests that clinical decision support, when used effectively, can improve healthcare quality, safety, and effectiveness," the study's authors wrote. "A large part of the electronic health record’s promise to improve patient care rests upon CDS. Because of this, implementing CDS has been an increasingly substantial part of the meaningful use program."
But as Gregory Paulson, deputy director of programs and operations at Trenton Health Team, told Healthcare IT News in 2015, meaningful use's technology-centric focus on CDS may have led to some of the challenges we see today.
"I think one of the big errors of meaningful use and the adoption of health IT to date has been the focus on technology, without looking at the processes that were implemented and the people working with those processes," Paulson said. "This work is about people, processes and technology – in that order.”
The importance of human oversight – and the ability to notice and respond to CDS errors – is borne out by the JAMIA study, which finds that malfunctions of decision support systems "are widespread and often persist for long periods."
Indeed, "the failure of alerts to fire is particularly difficult to detect," researchers write, pointing to a range of causes, including changes in codes and fields, software upgrades, inadvertent disabling or editing of rules, and malfunctions of external systems."
The verdict? "Current approaches for preventing and detecting such malfunctions are inadequate."
In its survey of CMIOs, the study asked which types of CDS are most prevalent in their hospitals. Drug-drug interaction alerts and allergy alerts were far and away the most common (97 and 93 percent, respectively), followed by screening/preventive care reminders (72 percent), renal dose adjustments (55 percent), alerts about abnormal test results (55 percent), drug-pregnancy alerts (38 percent) and reminders to patients (17 percent).
Asked how often has they'd experienced CDS malfunctions, 38 percent of CMIOs said four or more times a year, 28 percent each said one-to-six times a year and less than once a year. Just 7 percent said they'd never experienced a system misfire.
As for the causes, 62 percent of respondents cited an upgrade to their EHR software; another 62 percent pointed to changes to underlying codes or data fields. Forty-one percent found that inadvertent disabling or enabling of a rule caused malfunctions. Rounding out the list were upgrade of another clinical information system 34 percent) and database corruption or another system malfunction (24 percent).
Those findings "mirror Brigham & Women's experience," according to the researchers. "Eighteen of the survey respondents reported CDSS malfunctions occurring at the time of an upgrade of their EHR software, and 18 reported issues in connection with changes in data codes or data fields (both of which are known high-risk events for CDSS malfunctions)."