- Electronic health record (EHR) alerts play an important role in patient safety by notifying providers of potentially dangerous and hazardous clinical events, including inappropriate prescribing and medication administration errors.
However, these notifications can also be the source of physician frustration, causing more clinical harm than good. Providers are often bombarded with unnecessary and inconsequential alerts that lead to overrides, alarm fatigue, and potential patient harm.
At St. Jude Children’s Research Hospital, a team of researchers recently set out to optimize EHR alerts and reduce fatigue among clinicians.
“St. Jude was relatively early in adopting the EHR and computerized physician order entry (CPOE),” James Hoffman, PharmD, MS, Chief Patient Safety Officer at St. Jude, told HealthITAnalytics.com.
“We knew alert fatigue was an issue, so when we finished CPOE for medications in May 2010, we did a lot to try to manage that. But around 2013, we started seeing an increase in fatigue, and we started to hear more concerns about excess alerts. So even though we had worked on it early on, we realized it was still a challenge that we had to manage.”
To minimize the potential for alarm fatigue, the team analyzed the organization’s alert data. Researchers found that the most common interruptive alerts were drug-drug interaction (DDI) notifications designed to notify providers when prescribed medications may lead to contraindication.
The group observed that in some instances, clinicians were receiving a constant stream of DDI alerts, many of which weren’t applicable to pediatric care.
Researchers used three different methods to review and modify DDI alerts, such as forming a multidisciplinary alert advisory group which met consistently over the study period to review and improve alert refinements. Group members included physicians, pharmacists, nurse practitioners, and other EHR users across the hospital.
“The three methods were really important because they created some synergy,” Hoffman said.
“The advisory group gave us a broader perspective and allowed us to gather different people’s expertise. We as a team could have just looked at all the data, but if we did that, we wouldn’t have gotten as much input.”
The team also conducted in-person and email surveys with providers who had encountered interruptive DDI notifications.
“The surveys were helpful because we really got to understand clinicians' experiences with the alerts,” Hoffman said.
“People have different preferences: What one clinician sees as alert fatigue another clinician may see as acceptable and beneficial. The survey process also engaged providers. We collected a lot of the surveys by going to different service meetings, where we were able to ask providers how things were going. That prompted other valuable feedback.”
In addition to the advisory group and the surveys, researchers analyzed EHR data to determine which alerts were most ignored or overridden.
Using these three methods, the team made 26 changes that affected 47 percent of alerts. Changes included eliminating unnecessary alerts, as well as providing context to alerts to make them more valuable to providers.
These modifications helped fix some of the organization’s most significant alarm issues, such as alerts related to potassium and drugs that increase potassium levels.
Researchers found that nearly 20 percent of all the alerts generated in a three-year period were for potential increases in potassium. However, potassium levels were already being monitored regularly.
The team was able to turn off the vendor-supplied potassium alarm, and they developed a new alert that would notify providers only when previous potassium levels were increased. These changes helped reduce alert fatigue and target patient safety issues, and Hoffman said.
“One of the things that was unique and new to us was that we didn’t just suppress the alerts; we actually made some of the alerts smarter,” he explained.
“Alerts should be altered to fit the context of a clinical situation. If it’s a drug interaction where potassium is important, then we should be able to add in more detail about a patient's most recent potassium levels.”
After making changes to DDI notifications, the researchers found that alert overrides decreased by 40 percent.
Gathering input and recommendations from different individuals at St. Jude contributed significantly to the study’s success, Hoffman said.
“Collaboration across the organization was critical for optimizing alerts,” he said.
“We had the patient safety perspective, as well as the perspectives of pharmacists, nurse practitioners, physician assistants, and physicians. It was also critical that we had the input of the information services team, because they know the EHR and understand what the vendor is doing. It really was a collaboration and partnership across a lot of people at the hospital.”
Hoffman also pointed out that the methods used in this study can be applied to other institutions as well.
“Even though St. Jude is a highly specialized hospital that focuses on children with catastrophic diseases, we think a lot of our experiences would translate to other settings,” he said.
“A certain level of local adaptation and refinement is necessary to make alerts the best they can be. Because there are different patient populations, and there are different preferences among providers, organizations have to refine alerts at the local level.”
Putting the suggestions of providers and other EHR users into action can help improve EHR alerts and patient outcomes.
“I hope that EHRs continue to enable collaboration and partnership, which we were able to incorporate into this project,” Hoffman concluded.
“Hearing input from key players is very valuable, and something that is transferable to other organizations. If alerts can get smarter and more sophisticated, there will be less chance for fatigue and alerts will be better equipped to increase safety.”