Improving Buy-In for Medication Decision Support

By Matt Phillion

Alert fatigue, burnout, short staffing, and more: The pressure on providers grows in intensity and complexity every day. But a new technology enabling clinical decision support alert optimization is making strides toward reducing medication alert fatigue, cutting back on overrides, and offering greater avoidance of harmful medication use.

“I think a lot of organizations are dealing with a couple of issues. We’ve dealt with burnout, EMR fatigue, alert fatigue, all of which are driving some of this unaddressed problem,” says Patrick McGill, MD, executive vice president and chief transformation officer at Community Health Network, a nine-hospital health system in Central Indiana.

The system uses FDB Targeted Medication Warnings with FDB CDS Analytics, an FDB PatientFirst medication decision support solution, to identify and optimize 10 medication-related CDS alerts in an effort to make them more patient-specific, relevant to the immediate clinical context, and actionable.

“The other piece of what’s contributing is the EMR and tools people are using, with bad configurations or alerts to the wrong person, [or] alerts at a time and [place] in the workflow when you can’t do anything about it,” continues McGill.

And the third piece of the puzzle, he notes, is workforce turnover; it is estimated that healthcare organizations see 15%–25% turnover each year. “That’s a whole set of people who need to get used to your processes, procedures, and technologies,” he says. “That drives a smoldering problem that’s always there and raises its head at different times through adverse events or harm to patients.”

But it’s not an issue continually top of mind. “It’s not sexy, and sexy problems rise to the top—until you have a catastrophic event, and then everyone wants to engage,” McGill says. “It’s hard to keep organizations interested in pushing resources into it.”

It’s also hard to quantify the results when those resources work. “How do you measure harm avoided?” McGill says. “Executive leadership wants to see the return on investment. Harm avoided is a softer ROI, which makes it harder to get buy-in. However, we do have a number of other metrics we can measure before and after implementation, focused on patient safety and improved workflow for clinicians.”

Proving value and effectiveness

Dealing with the challenges inherent in clinical decision support requires a simple tool that can be built into the EMR and work alongside it, but doesn’t add to the workload or drain resources. To test the efficacy of FDB’s tool, Community Health Network looked at two factors.

“The first is ignore rate,” says McGill. “How many times did the provider not look at [an alert] or interact with it? Some of that is workflow and end-user education, and some of it has to do with where it fires off in the workflow: Is it firing for someone who can do something about it? Is it offering actionable advice? This goes back to good alert configurations.”

The other component Community Health Network measured was the follow rate, which indicates how often the provider follows the suggested guidance.

“You have to ask, did your ignore rate go down and your follow rate go up? Were they changing doses, doing something with this alert?” says McGill. “You want to see that.” If the ignore rate goes down without an increase in the follow rate, it doesn’t mean providers are following the actions in the alert, he explains.

Anna Dover, PharmD, with FDB, says some of its users have experienced a 60% increase in their acceptance rates, where providers either took action or selected a follow-up action they would do instead. “They took the time to say specifically what they were going to do instead of just buzzing through the alert.”

This is particularly impactful in areas like the ICU, says Dover, a former ICU pharmacist. “I learned that I couldn’t see anything that was fired at me—and everything in the ICU is technically wrong, but for that patient population it’s right,” she says, because almost every ICU case is reactive and involves out-of-the-ordinary circumstances. “It’s about understanding the context around it.”

When we think of medication alerts, it is generally in terms of drug/drug interactions or drug/diagnosis interactions. Those can seem significant on paper but may not represent a meaningful danger to the patient depending on the clinical situation. “We should only alert when it really matters. Is it an increased risk or an actual concern?” says Dover.

The wrong kinds of alerts drive physicians crazy, notes McGill. Drug-related potassium warnings, for example, were often dismissed at Community Health Network because patients were on two drugs that eliminated their risk of hypokalemia when taken concurrently.

“We put it into context, and that’s very hard to translate into decision support, and that’s why clinicians got frustrated,” he says. “It kept alerting, but lacked the context. So we needed to find a way to build context into our decision support tools, which is where FDB came in.”

Maximizing the point of impact

A tool intended for clinical decision support needs to consider both when the alert happens and what it involves.
“We get a lot of medication alerts that will fire—the patient has an allergy, or it’s the incorrect dosing for a child—but we’re really taking it to the next level,” says McGill.

An effective decision support tool needs to intervene at the point of prescribing, he says, to ensure that the clinician receiving the alert can decide to reduce the dosage or add a medication, for example.

“Where this is really important is for certain locations: Anytime you treat patients in the ICU or OR, often in the ED, there are lot of things that fall outside the boundaries of typical rules,” says McGill. “This gives you the flexibility to fire that alert in outpatient settings but suppress it in the ICU.”

One alert type that caused a lot of complaints prior to implementing decision support technology was the dosing frequency alert. “Say you typically can’t dose a patient more frequently than every 10 minutes,” says McGill. “In the OR, all bets are off. If you have a patient crashing or they need pain meds, it doesn’t matter how many minutes it’s been, you’ve got to control what’s going on.”

The ROI on prevention can be difficult to prove, but the technology has a powerful impact, McGill notes: happier staff, more time to interact with the patient, and, of course, harm avoided.

“Nobody wants to be in the OR. We run a 24/7/365 business where nobody wants to be there, so it’s all about having the highest-quality experience,” says McGill. “We always want people to practice at the top of their license, enabling physicians to do things only physicians can do, nurses to do things only nurses can do.” Alerts that are too numerous or aren’t relevant lessen that capability and force providers to spend nonproductive time on administrative tasks. This makes it imperative to ensure that the alerts providers get are meaningful.

“It’s easy to miss things, so we’re pulling together all the related information, not just drug A versus drug B, and aggregating it around the patient,” says Dover. “We’re trying to better interact with the physician for decision support in the way they look at all the factors involved with the patient. When you’re alert fatigued, it’s easy to miss something important, so we want to compile all this information and make it easier to assess.”

If there’s a challenge to implementing this type of tool (beyond the financial constraints all organizations are facing), it’s that a lot of traditional medication decision support is programmed too natively by the EMR, says Dover. “This is an add-on, which can be scary. But it takes a nominal commitment of IT team resources and budget,” she says.

Technology to cut down on alert fatigue will continue to advance and evolve. “The metrics are the next big adventure,” Dover says. “We can measure how often an alert fires, and we’re also evolving the decision support and all the different things we need to benchmark against.”

McGill says Community Health Network will continue exploring how the tool affects safety and outcomes.

“We need to continue to iterate and expand the functionality,” says McGill. “I would love to continue to measure outcomes to help people get that buy-in to move forward. There’s no person in healthcare who doesn’t want to provide exceptional care, and that’s a journey that never ends. We need to stay focused on that.”

Matt Phillion is a freelance writer covering healthcare, cybersecurity, and more. He can be reached at