Research Shows Physicians Commonly Override Medication Alerts

A new study published in the Feb. 9 issue of Archives of Internal Medicine concludes that physicians across all medical specialties override most medication alerts that are designed and built into electronic health record, or EHR, systems specifically to improve patient safety.

The authors of the study, "Overrides of Medication Alerts in Ambulatory Care," (abstract available free; nonsubscribers pay $15 for full text of article) said, "Given the high override rate of all alerts, it appears that the utility of electronic medication alerts in outpatient practice is grossly inadequate." The authors speculated that for active clinicians, "most alerts may be more of a nuisance than an asset."

Researchers looked at more than 3 million electronic prescriptions initiated by 2,872 clinicians from 862 practices across the states of Massachusetts, New Jersey and Pennsylvania.

All clinicians involved in the study used e-prescribing software made by the same software vendor. Collectively, those systems generated 233,537 medication safety alerts during the course of the study, which ran from Jan. 1- Sept. 30, 2006.

Researchers found that overall, clinicians accepted 9.2 percent of drug interaction alerts and 23 percent of allergy alerts.

Drug interaction alerts were divided into three groups designated as high, moderate or low severity. Although more than 60 percent of the alerts received were classified as "high severity," the clinician acceptance rate was only slightly higher for those alerts (10.4 percent) than it was for moderate- or low-severity alerts (7.3 percent and 7.1 percent, respectively).

Lead study author Thomas Isaac, M.D., M.B.A., M.P.H., an internist in the Division of General Medicine and Primary Care at Beth Israel Deaconess Medical Center in Boston, told AAFP News Now that his personal experience with medication alerts mirrors the study findings. "I override the majority of the alerts that I encounter -- about 90 percent," he said.

Most of the data gleaned from the study aligned with Isaac's initial hypothesis that clinicians in ambulatory settings were, like their colleagues with practices affiliated with academic medical centers, overriding a majority of alerts in their e-prescribing software. However, one unexpected outcome of the study was that clinicians routinely overrode even the major alerts.

"I was expecting a little bit greater differentiation between minor, moderate and major alerts in terms of the frequency with which clinicians overrode them," said Isaac.

Isaac said a major issue he sees is that physicians "encounter a lot of alerts which have a theoretical basis of harm but very infrequently cause any problems." In other words, "there are a lot of alerts that are of questionable utility to clinicians."

Electronic prescribing is second nature to an early EHR adopter like FP Susan Andrews, M.D., of Murfreesboro, Tenn. But she said it's not unusual to see alerts that don't make sense. For example, the majority of patients with diabetes should be on an ACE inhibitor, but a common medication alert warns of an increased risk for hyperkalemia in patients with diabetes who take ACE inhibitors. "That warning should usually be overridden," said Andrews.

"Warnings do not take the place of good clinical judgment," she added.

Capt. Robert Marshall, M.D., M.P.H., U.S. Navy Medical Corps, is the director of clinical informatics at the U.S. Navy Bureau of Medicine and Surgery in Washington. He also has a master's degree in information systems management and serves on the AAFP's Commission on Quality and Practice. Marshall said physicians should be cautious when overriding medication alerts. "Some alerts are frivolous, and some are quite important. It still takes current medical and medication knowledge to discern which is which," he said.

Marshall added that overriding alerts without certain knowledge that the alert is erroneous or is overstating the risks could lead to patient safety issues. "And just like any medical misadventure, overriding a legitimate alert and having an adverse outcome can potentially result in litigation and damages," he said.

According to Marshall, he overrides medical alerts in his e-prescribing software if he knows for certain "that the alert is not legitimate and/or simply represents a too sensitive alert system." He, like Andrews, questions the need for an alert on an ACE inhibitor for a patient with diabetes; he also commonly overrides an alert for a nausea/vomiting "allergy" to codeine.

However, Marshall plays it safe and teaches residents at the Dewitt Army Community Hospital Family Medicine Residency in Fort Belvoir, Va., to do the same. "If I have any uncertainty about an alert, I will look it up, ask the patient or ask the pharmacist directly. I will then document it in our EHR for future reference," he said.

The study authors noted that their findings should help software designers improve medication safety alerts. In particular, the authors suggested that designers

The authors concluded that, "unless designers take steps to improve prescription alert systems, the potential benefits of electronic prescribing may not be realized."

But just how do software designers keep abreast of the changes physicians want? Steven Waldren, M.D., M.P.H., director of the Academy's Center for Health IT, said most vendors keep a finger on the pulse of current research.

"Vendors also have user groups and forums and annual meetings where they get feedback from members about products and services," said Waldren. Many software designers also have what they call "innovation sites" where they get regular feedback on changes they're considering in their products, he added.

However, sometimes changes take additional teamwork from other players. For example, in the case of medication alerts, the underlying drug database might also need updating.

So how soon can Drs. Isaac, Andrews and Marshall expect to see changes that will make their medication alert software more useful? Waldren said 18 months to two years would be a reasonable time frame to make the kinds of tweaks called for in this study.