EXECUTIVE SUMMARY:
Across the U.S., as healthcare providers implement computerized physician order entry (CPOE) systems, they find themselves dealing with the growing issue of clinical alert fatigue. With patient care alerts proliferating within clinical decision support (CDS) systems, physicians have often come to ignore all alerts. Healthcare IT leaders are working to resolve this important issue to everyone’s benefit, increasingly implementing systems that put out only effective alerts or apply asynchronous alerting strategies.
What happens when something designed for patient safety ends up having the exact opposite effect? In what can best be described as a “boy who cried wolf”-type scenario, this is exactly what is happening in some healthcare communities with CPOE systems.
Systems are implemented with patient-safety alerts, a CDS tool that helps physicians recognize when a physician needs to be made aware of any of a variety of possible situations, such as when a patient shouldn’t take a prescription order for reasons such as drug interaction, drug allergies or dose-range checking. However, the alerts can often become excessive to the point where physicians will simply override them as not to disrupt their workflow.
The phenomenon is called “alert fatigue,” and it’s become a significant issue in hospitals that have implemented CDS systems. Studies, like a 2009 report from the Boston-based Beth Israel Deaconess Medical Center (BIDMC) and the Dana-Farber Cancer Institute, document the seriousness and scope of the issue. The researchers looked at the safety alerts generated by 2,872 clinicians through 3.5 million electronic prescriptions over a nine-month period. Of the 233,537 alerts, 98 percent were drug-drug interaction issues, more than 90 percent of which were overridden. Clinicians overrode more than 77 percent of the allergy alerts as well.
“It would be easy to think that more alerts equals more safety, but alert fatigue—fatigue is probably too generous of a word, I’ve seen wholesale ignoring in some cases—presents the doctor with trying to weed out the meaningful alerts from the meaningless ones, and I’ve seen articles quoting 98 percent alerts that weren’t acted upon,” explains Mark Van Kooy, M.D., director of informatics, Aspen Advisors (Denver, Colo.). “That means if you have 100 alerts, you have to go through 98 until you find two that are justifiable. At some point, you just start missing alerts. That’s a worst-case scenario, but it’s real-world.”
Acting on two percent of alerts isn’t doing anyone any favors, and analysts like Van Kooy say hospitals need to figure out a situation where the alerts that are coming up, are acted on approximately four out of five times. In all likelihood, this strategy would mean cutting down on the number of alerts that come up in current CPOE systems.
AHS' STORY
Many hospitals have begun to work on this sort of thing already. One such institution is the Altamonte, Fla.-based Adventist Health System (AHS), a faith-based hospital system with 44 hospital campuses across 10 states. AHS' vice president and CMIO, Phillip A. Smith, M.D., says the organization rolled out its CPOE across 26 states and saw issues with alert fatigue shortly thereafter. Despite a conscious effort to be “more effective” with its alerts, AHS found out physicians were getting 80 alerts per 100 medications.
Phillip A. Smith, M.D.
Smith said the organization immediately recognized this was far too many to avoid alert fatigue. “We knew what our target was, about 10 alerts for every 100 medications ordered,” he says. “That’s where doctors actually change their behavior.” After getting it down to 34 alerts per 100 medications and then 22 alerts per 100 medications, where there was a 50-50 chance the doctor would ignore the alert, the group worked with its CPOE vendor Cerner (Kansas City, Mo.) to reach its target level.
With Cerner’s help, AHS implemented multi-functional tools called MCDS, which refines the alerts and allows the organization to go after and reduce “nuisance” alerts. Thanks to this tool, AHS was able to get itself down to 14 alerts per every 100 medications, with an average of 10 ignored and four overridden. The tools, which Smith says will be available shortly from Cerner for customers who upgrade their systems, were able to eliminate certain duplicates and other unnecessary alerts.
ASYNCHRONOUS VS. SYNCHRONOUS
In California, leaders at the 311-bed Lucile Packard Children’s Hospital (LPCH) in Palo Alto, Calif. have recognized the issue of alert fatigue, citing studies that appeared in the Journal of American Medical Informatics Association, which had physicians overriding numbers of allergy alerts occurring approximately 60-95 percent of the time. Natalie Pageler, M.D., medical director of clinical informatics at LPCH, and Christopher Longhurst, M.D., CMIO at LPCH, say clinical alerts can be divided into either synchronous or asynchronous decision support, the latter of which they say is a better solution for fixing alert fatigue.
Christopher Longhurst, M.D.
“When we talk about alerts, most people are thinking of synchronous alerts, which happen when someone is actually placing an order. You place a medication order, get a pop-up that says the patient is allergic to this medication,” Dr. Pageler says. “But the fact is, there’s a lot of information you’d like to get to the provider that happens at times when they’re not actually entering an order. Like for instance, you enter an order based on a particular level of renal function, and then the renal function changes three days later. How do you get that information to the provider, who isn’t at the computer making the order? There’s lots of ways to do this type of asynchronous alert.”
LPCH has tested asynchronous alerting through various methods including the development of a “highly elaborate tab,” according to Pageler. The tab includes information on the patient’s vitals, medications, care providers names, and other critical information. On the tab, Pageler says, are alerts, but they aren’t interruptive. “So if the kidney function isn’t normal, for instance, it will be highlighted in red on the tab,” she says. “It’s an alert, it’s highlighted in red, but it doesn’t interrupt their workflow.”
Natalie Pageler, M.D.
In addition, LPCH has developed a patient care and quality dashboard, which has an enhanced healthcare related checklist included in the EMR. The non-interruptive checklist provides information on vital information, both general and specific to the patient. The dashboard, which was sponsored by Hewlett-Packard (Palo Alto, Calif.), is currently in pilot.
COMPLEX SOLUTION
Even with their focus on asynchronous alerts, both Pageler and Longhurst say the focus on alerts comes down to improving clinical outcomes and not the process. Longhurst was quick to point out a study conducted by LPCH that found pop-up messages, which are synchronous, built into an EMR could prevent physicians from ordering unnecessary treatment, in this case blood transfusions. In the study, the pop-up alert saved LPCH from conducting 460 unnecessary red blood cell transfusions—$165,000 in one year.
The moral of the story, as Aspen’s Van Kooy says, is that there is no simple solution. “The solution requires thought, effort, analysis, and engaging all the stakeholders—pharmacy, nursing, physicians—as key success factors. There are systems that are doing this well and they (healthcare providers) should keep their eyes on the literature to keep informed, and they should look for continuing development of evidence based guidelines on how to approach this problem,” he says.
