Two things are likely familiar to any physician working in an electronic health record (EHR): Seeing clinical decision support (CDS) alerts and overriding those alerts.

Such alerts remind clinicians about everything from a patient's drug allergies, to possible drug interactions, to dosing guidelines, to lab testing guidance. Clinicians can either follow the alerts' recommendations, override them, or ignore them.

"They're frequently overridden, and often these are done appropriately," said Adrian Wong, PharmD, MPH, an outcomes research and pharmacy informatics fellow in the division of general internal medicine and primary care at Brigham and Women's Hospital in Boston.

Overriding alerts that are clinically relevant can potentially lead to harm, according to a study in BMJ Quality & Safety, by Wong and colleagues.

The prospective observational study examined medication-related CDS overrides among adults admitted to any of six Brigham and Women's ICUs from July 2016 to April 2017.

The study included 2,448 overridden alerts from 712 unique patient encounters. Although 81.6% of the overrides were appropriate, the "inappropriate over-rides were six times as likely to be associated with potential and definite ADEs [adverse drug events], compared with appropriate over-rides," according to Wong's group.

For example, in two cases, physicians ordered certain medications even though there was a "reaction that was documented" and "there were other alternative options available," the authors stated. In both cases, the patients had a similar adverse drug event to what was documented.

However, the doctors "might not even have realized that they had encountered this alert at all because of the amount [of alerts] that they encounter," he said.

Wong said that ICU patients have a higher risk of ADEs because they're generally sicker, in the hospital longer, and receive more medications than other hospital patients.

Wong's past research over a 3-year period at a 793-bed tertiary-care teaching institution demonstrated that almost three-quarters of alerts were overridden, and 40% of the overrides were not appropriate. Also, appropriateness of medication-related CDS overrides in the ICU varied by the type of alert.

However, researchers "haven't really evaluated why [we] should care that they're overridden" until now, Wong said.

Although other research has shown that clinicians find alerts to be "excessive," such a huge increase in the risk for ADEs indicate that some clinical alerts convey critical information. For example, one study found that the "use of CPOE [computerized physician order entry] and isolated CDSSs [clinical decision support systems] can substantially reduce medication error rates."

That research conducted systematic reviews of studies that evaluated the effect of CPOE and CDSSs systems on medication safety, and found that "Of the CPOE studies, 2 demonstrated a marked decrease in the serious medication error rate, 1 an improvement in corollary orders, 1 an improvement in 5 prescribing behaviors, and 1 an improvement in nephrotoxic drug dose and frequency. Of the 7 studies evaluating isolated CDSSs, 3 demonstrated statistically significant improvements in antibiotic-associated medication errors or adverse drug events and 1 an improvement in theophylline-associated medication errors."

That leads to the possibility that too many -- and not the right kind -- of alerts are to blame. Even though Wong's group did not delve into the "why" of CDS overrides, he said it is likely due to alert fatigue, which "may lead to providers [to not trust] the system because of all the clinically irrelevant alerts they find (and potentially alerts that they had expected to see but did not encounter)," he wrote in an email.

"More alerts are happening and added to the system without looking at the clinical relevance," he said.

Does the answer to the problem reside in technology or clinicians? "It's everyone's solution," Wong said, meaning that IT teams, vendors, and providers need to work together to gain buy-in, be able to tailor alerts to institutional needs, and critically evaluate what alerts are necessary.

For example, at his institution, Wong's group is making suggestions for improvements, such as eliminating alerts for medication orders of electrolytes, a change that would reduce the number of renal alerts significantly.

But simply reducing the number of alerts isn't always the answer.

"There are studies that evaluate just reducing the number of alerts in an institution with varying success," Wong said. "You should evaluate these alerts and see what you can do to improve or reduce the alerts that shouldn't be there or that are clinically irrelevant."

He advised that healthcare organizations work with clinicians and CDS teams to evaluate which alerts are firing and whether they're right for that institution's unique needs.

"These clinicians are often not incorporated from the beginning from these decisions," Wong said. "Really paying attention to the frontline clinicians is really important to identifying areas of improvement."