Home  |   Subscribe  |   Resources  |   Reprints  |   Writers' Guidelines

November-December 2020

Fine-Tune the Alert Volume
By Lisa A. Eramo, MA
For The Record
Vol. 32 No. 6 P. 14

Too much noise in the EHR can drown out the sound of serious medication errors. How can organizations get the mix just right?

When it comes to medication safety, EHRs haven’t exactly lived up to the hype. That’s the finding of a recent study conducted by scientists at the University of Utah Health, Harvard University, and Brigham & Women’s Hospital in Boston. Specifically, researchers found that, on average, EHRs fail to detect 33% of medication errors that lead to serious harm or death.

In some cases, EHRs failed to detect 90% of errors or more, says David C. Classen, MD, MS, corresponding author of the study and a professor of internal medicine at the University of Utah. “Hospitals were told by EHR vendors that all of these problems would be solved,” Classen says. “Despite an enormous investment from the federal government and private sector, problems still exist. That’s what’s concerning.”

Julia Adler-Milstein, PhD, a professor of medicine and director of the Center for Clinical Informatics and Improvement Research at the University of California, San Francisco, agrees. “I do think there was an expectation that if anything was going to get better, it was going to be medication safety. We would have thought there would be more stellar results at this point,” she says.

However, the research indicates otherwise. To arrive at their conclusions, Classen and colleagues analyzed a decade’s worth of data from 2009 to 2018 collected through the Leapfrog computerized physician order (CPOE) entry EHR test. (Leapfrog is an annual web-based voluntary survey that collects safety, quality, and resource use information.)

The study’s goal was to evaluate the performance of seven leading EHR vendors. The survey feeds various clinical scenarios into EHRs to help organizations determine whether their systems correctly issue warnings or alerts about potential medication problems.

Almost all of the Leapfrog scenarios used in Classen’s study were based on real-world adverse drug events that harmed or killed patients. Cumulatively, his research included more than 8,000 hospital observations from more than 2,000 hospitals nationwide.

The range of medication error detection (ie, under 20% to almost 100%) is perhaps the study’s most striking finding, Classen says. “Any time you see that much variability, you know there’s a problem,” he notes.

Tackling a Complex Challenge
EHRs are a complex technology implemented in a complex environment. EHR vendors provide a technology framework, and medication data vendors provide relevant content about drug-drug interactions, drug-allergy interactions, drug-dose recommendations, drug-disease interactions, therapeutic duplicates, and more.

However, organizations are responsible for the final version of the EHR that goes live, including the decision to turn certain alerts on or off. This means that a serious drug interaction that would trigger an EHR warning at one hospital might not at another.

Although it’s easy to point the finger at EHR vendors and blame them for failing to detect medication errors, hospitals need to recognize their critical role in the safety of their systems, Adler-Milstein says. “It would be great if it were as easy as vendors building better clinical decision support from the start, but I think frankly, a lot of it depends on the hospitals,” she says. “They are making a lot of the decisions about how the EHR works in practice. It’s a question of how many resources they can—and choose—to devote to this issue.”

Classen says hospitals may not take action until there are regulations from The Joint Commission or Office of the National Coordinator of Technology (ONC) requiring them to prove the safety of the operational clinical decision support they use in a live environment. Currently, federal regulators inspect only factory specifications, not how the technology is altered after installation.

“ONC has not made safety a priority,” Classen says. “It’s great to have a totally interoperable system, but is it automatically a safe system as well? Absolutely not.”

Interestingly, the hospitals included in Classen’s research are those that voluntarily used the Leapfrog survey. The majority of hospitals don’t actually take the survey. Classen suspects that EHRs in those organizations perform far worse—likely failing to detect 50% or more of medication errors.

When compared with other industries, health care has a lot of room for improvement when it comes to safety, Classen says. “Other industries have learned that software in a high-risk industry can cause a huge disaster,” he says. “With airlines, they have routine checks of the flight management system. Health care doesn’t do this.”

Experts agree there are several steps organizations can take to improve medication safety when using an EHR.

Form an Interdisciplinary Decision Support Team
A data governance team that includes pharmacists, prescribers, HIM, IT, and other any relevant clinical staff (eg, nurses and medical assistants) are advised to take the following measures:

• Establish clinical protocols for adjusting medication doses based on organ function and other parameters. This requires clinician input and consensus.

• Ensure certain structured fields in the EHR are populated consistently and without error.

• Identify enterprisewide metrics to monitor alert performance over time.

Organizations put a lot of focus on medication safety during the first year or 18 months of rollout, but then it falls to the back burner, says Dana Bensinger, MSN, RN-BC, a client solution executive at Computer Task Group. “Organizations need continuous quality improvement and evaluation,” he explains.

Create Clinically Relevant Alerts
Continuous quality improvement means focusing on ways to avoid alert fatigue, says Adam Tallinger, RPh, MHA, CPHIMS, vice president at Impact Advisors. “You need to pick the right level of alerting to improve safety, but you also don’t want to hinder safety by causing people to blindly bypass alerts,” he says.

The more relevant the alert to the individual patient, the better. “We are all unique in our reactions to drugs,” Tallinger says. “The health of our organs, our genomics, our ethnicity, and our exposures to environmental factors all produce variability in medication effectiveness, dosage needs, and adverse effects.”

There are subjective sensitivities that are completely unrelated to the reaction experienced by a patient, Tallinger says. For example, consider a patient who gets an upset stomach whenever they receive penicillin. In most EHRs, an allergy to penicillin would be flagged as severe. However, the alert doesn’t take into consideration that this specific patient’s reaction is mild compared with those who experience life-threatening anaphylaxis and that it may be okay in certain circumstances to prescribe it.

Tallinger says there are plenty of other examples that demonstrate how clinical circumstances should determine whether an alert actually fires. For example, there are many cases where the medical community–accepted use of a drug exceeds FDA-approved dosing. In addition, there are various ways in which drugs may interact. Simply knowing there is an interaction isn’t enough.

“They may cause toxicity, synergy, or they may decrease effectiveness of one or the other,” Tallinger says. “This can also be time dependent either based on just needing to space out their administration or a reaction that happens only after long-term use. These levels of interaction all highlight the need for a level of alerting but allowing a physician’s knowledge of the patient to drive a decision on the drug’s use.”

Scott Nelson, PharmD, MS, an assistant professor in the department of biomedical informatics at Vanderbilt University Medical Center in Nashville, agrees. “A lot of the baseline alerts are way too overly simplistic. The EHR doesn’t do a good job of looking at all of the contextual factors,” he says.

At Vanderbilt, a clinical decision support team creates alerts that take lab values, age, kidney function, medication, patient-specific adverse outcomes, and more into consideration. “We bring in a lot of context to decide whether it’s an alert that the provider even needs to see,” Nelson says. “It takes a lot of work to maintain this.”

To complicate matters, as new medications enter the market and others are removed, the team must update existing alerts. The same is true when the health system acquires new hospitals, merges two departments, or uses new lab tests, all of which can break the decision support, according to Nelson, who adds that if hospitals can’t afford to create this type of customization, they may need to contract it out to a third party.

Teach Providers Not to Rely Too Much on the EHR
Experts agree that in an age of automation, some providers may become overreliant on the EHR. “Providers assume that if they don’t see an alert, it must be safe to give the medication, which may or may not be the case,” Nelson says.

“Those safety nets degrade because everyone starts to trust the system,” Classen says. “When you automate something, people come to rely on the automation.”

Organizations need to do a better job of emphasizing that the EHR is only one of several safeguards. Using an EHR doesn’t absolve physicians from using their clinical expertise to ensure safety at the time of ordering, Tallinger says. It also doesn’t absolve pharmacists from using their expertise to ensure safety at the time of fulfilling the order or nurses from ensuring safety at the time of actually dispensing the medication to the patient.

Each of these steps is vital. In theory, it’s a series of checks and balances that relies on all players to do their part at all times—particularly during transitions of care, Tallinger says.

Nelson uses a Swiss cheese analogy. “Every layer has holes in it that will allow errors to go through, but if you have enough layers, then hopefully those holes don’t all line up to let the error go through,” he says.

Providers need to be reminded to question the validity of the information in the EHR and always do what’s best for the patient, Bensinger says. “We need to remember that the tool is an accessory to care. It’s not the be all, end all,” he says. “Sound clinical decision making still needs to come into play.”

Test Alerts Frequently
Testing is a critical element of ensuring patient safety. At a minimum, organizations should test high-risk medication orders annually and any time they perform a minor or major upgrade to their EHR, according to Bensinger.

An easy way to do this is to complete the Leapfrog CPOE EHR survey, says Classen, who is in the process of creating an outpatient version of the survey for physician practices and ambulatory clinics. The outpatient survey it set to release at the end of 2021.

Tallinger says hospitals can also use their adverse event/safety reporting system to capture and test a subset of the previous year’s reported events. “This makes the annual testing more dynamic and relevant to actual events that made it through all layers of the Swiss cheese rather than theoretical scenarios that might be more likely to be caught by the physician, pharmacist, or nurse,” he explains.

Nelson also recommends asking users for feedback. At Vanderbilt, users can rate alerts and provide input on why a particular alert isn’t relevant. “Then we look into it to find ways to fine-tune the alert even more,” Nelson says.

Consider Bar Code Medication Administration
Bar code medication administration ensures that the right patient receives the right medication (as dispensed) at the right time and at the right dose. “It’s a step on the path to reduce errors, but it’s only a first step,” says Fern FitzHenry, RN, PhD, a research assistant professor in the department of biomedical informatics at Vanderbilt University Medical Center.

However, the technology has limits. For example, it validates only what formulation was profiled to be dispensed to the patient—not what the physician ordered, FitzHenry says. In addition, bar code medication administration isn’t ideal for drug administration that requires interpretation (eg, if the patient’s glucose is between these two numbers, give this dose of insulin), she says.

Another limitation? Some nurses administer the medication and then scan the barcode. “That defeats the whole purpose of using the technology,” Nelson says.

The Role of EHR Vendors
Although organizations are responsible for making critical decisions that affect medication safety, EHR vendors aren’t completely off the hook.

Adler-Milstein says one way in which vendors can be a part of the solution is to provide standardized reports on how well certain alerts perform within a given hospital. They could also share best practices across clients and help achieve economies of scale, she adds.

“I think the more feedback they can give to hospitals, the better. Each hospital shouldn’t have to build reports itself. Vendors should be able to do at least some of it at scale,” Adler-Milstein says. “A little more feedback and insight could go a long way toward helping hospitals that have limited resources.”

Vendors should also be involved in annual medication safety testing, Bensinger says. “That information is also incredibly valuable to the vendor. Allowing them to be an active partner enables them to fix things,” he notes.

Finally, EHR vendors need to set realistic expectations, according to Bensinger. “They need to make it clear that they’re delivering a foundational piece of software and that there are maintenance steps the organization will need to take to ensure medication safety,” he says.

Doing What’s Right
When it comes to medication safety, patients’ lives are ultimately on the line, and so is a hospital’s reputation, Classen says. “Would you get on an airline if the pilot said we’ll get you there safely two-thirds of the time? You wouldn’t,” he says. “Likewise, we shouldn’t accept the medication safety system at a hospital if it’s accurate only two-thirds of the time.”

— Lisa A. Eramo, MA, is a freelance writer and editor in Cranston, Rhode Island, who specializes in HIM, medical coding, and health care regulatory topics.


EHRs fail to detect medication errors every day. However, pointing fingers at EHR vendors isn’t the answer. Experts say the following organization-level strategies are critical:

• Devote resources to creating clinically relevant alerts that reduce alert fatigue.

• Encourage providers to have a degree of healthy skepticism of EHR data.

• Form an interdisciplinary committee to address critical questions and ensure enterprisewide buy-in.

• Prioritize medication reconciliation during transitions of care.

• Test alerts frequently to determine efficacy and ensure patient safety.