Banner Health is one of the largest nonprofit health systems in the United States, serving more than 8 million patients across six states in the Southwest. It has countless providers, all of whom are racking up normal levels of stress, combined with pandemic levels of stress.
More than ever, providers are at risk for burnout, which can impact patient care and the lives of caregivers. To focus on provider burnout, Banner Health launched a multiyear strategy to support provider wellbeing called Cultivating Happiness in Medicine, or CHIM.
The strategy includes a comprehensive approach that addresses burnout and improves wellbeing from operational, leadership, organizational, individual wellness, community and second-victim perspectives.
“One of the first priorities for the CHIM initiative was addressing the pebbles created by the electronic health record, and, more specifically, alert fatigue,” explained Denise Erickson, PharmD, senior director, clinical pharmacy services, at Banner Health. “Our clinicians receive a staggering number of decision-support alerts, many of which are related to medications.”
Prior to working with its EHR vendor Cerner, each month Banner Health clinicians received more than 1.4 million medication clinical-decision-support alerts, which consisted of drug-drug, drug-allergy and drug-duplicate alerts. The mCDS alert rate prior to the CHIM project was 0.79 (79 alerts per 100 orders) with 6.2 alerts per alerted day for providers and 128.5 alerts per alerted day for pharmacists.
This alert rate stood out as an opportunity for improvement due to the high alert rate, coupled with a high 90% override rate, she said.
One of the common frustrations Banner Health hears from its clinicians throughout the organization is that there are too many onscreen interruptive alerts.
“Excessive alerting leads to alert fatigue, which ultimately desensitizes clinicians and decreases the utility of alerts,” Erickson noted. “Similar to other organizations, we’ve used several strategies, with varying degrees of success, to reduce nuisance alerts and minimize the effects of alert fatigue. Targeting drug-drug, drug-allergy and drug-duplicate alerts has historically been a challenge, since it is resource intensive, a manual process that is difficult to sustain, and there have been functional limitations of EHRs.”
“Some of the filtering scenarios that had the greatest impact include the ability to suppress alerts within a specific order set, filtering alerts based on order details, and filtering alerts based on the venue of care.”
Denise Erickson, PharmD, Banner Health
When working with Cerner, Banner Health identified an opportunity to reduce the number of alerts using new technology that addresses some of the challenges the health system faced in the past.
The new Cerner technology was a key component of this project. As an organization, Banner Health has spent considerable effort targeting alert fatigue, but historically there have been EHR limitations that prevented the organization from seeing the full benefit of this work.
“One example is when there was an interaction between medication A and medication B,” Erickson explained. “Prior to technology updates, the clinician would see an alert that medication A interacts with medication B. The clinician would then see a second alert notifying them that medication B interacts with medication A. Addressing this issue is a quick way to reduce the number of alerts and to condense the information displayed in the alerts to what is clinically relevant.”
In addition, there were some inconsistencies with regard to which alerts displayed to the various users within the organization, such as alerts that only displayed to pharmacists. This created inconsistency and led to over-alerting for some positions and not enough alerts for others.
“As we reviewed the information and workflow in each clinical area – inpatient, ED, clinic – we decided to adjust alert preferences to ensure the right information is displayed, at the right time, to the right end user,” Erickson said.
“In some cases, this meant removing unnecessary alerts for pharmacists, but in other cases it meant displaying new alerts to providers at the time of ordering. While adding new alerts seems counterintuitive, we felt it would save providers and pharmacists time in the end by reducing the number of phone calls from pharmacists to providers regarding drug interactions.”
By taking a team approach and using data-driven dashboards for mCDS alerts, Banner Health was able to identify areas for optimization. The ability to use the data-driven dashboards with internal Banner data, paired with Cerner technology to eliminate these alerts via Cerner tooling, allowed the organization to focus not only on reducing the overall number of alerts, but also on increasing overall efficiency and provider satisfaction.
“All of this supported our goal of reducing the number of alerts, improving the alert acceptance rate and reducing provider burnout,” Erickson stated.
There are many vendors with electronic health records systems on the health IT market today, including Allscripts, athenahealth, Cerner, DrChrono, eClinicalWorks, Epic, Greenway Health, HCS, Meditech and NextGen Healthcare.
MEETING THE CHALLENGE
There were a few technology components to the Cerner solution that were important to this project. The first was creating an interactive dashboard that described the characteristics of mCDS alerts.
“Numerous Cerner reports were used to gather background information regarding mCDS alerts, such as the number and frequency of alerts, which roles/positions were receiving alerts, when alerts were displayed, and how often alerts were overridden,” said Erickson.
“The data from these reports were integrated into the dashboard, which enabled our team to predict the impact of the Cerner technology on the number of alerts. The dashboard was an import tool for the team in making decisions regarding how and to what extent the new technology would be implemented.”
The other main component of this project was implementing new tools within the EHR to filter unnecessary alerts for end users. Historically, Banner Health’s ability to modify drug-drug interactions was limited to adjusting the interaction severity, and it had a very limited ability to adjust drug-duplicate interactions.
“The Cerner solution provided us with 14 new ways to suppress nuisance alerts,” she explained. “Some of the filtering scenarios that had the greatest impact include the ability to suppress alerts within a specific order set, filtering alerts based on order details, and filtering alerts based on the venue of care. For example, our PACU order set by design avoids duplication in therapy. However, it was generating many duplicate alerts due to limitations in how drug-duplicate alerts have historically functioned.”
Staff was able to eliminate more than 88,000 nuisance alerts by suppressing alerts within this order set. In another example, staff was seeing many unnecessary duplicate alerts when there were orders for two pain medications with different indications (for example, one order for moderate pain and one for severe pain).
The Cerner technology enabled staff to remove alerts when the orders had unique indications while keeping alerts when there are true duplicates.
“Even though we had strong technical and analytical support, there was quite a bit of manual work and review with this project,” Erickson noted. “The team reviewed each of the new filtering scenarios in detail and used the dashboard to ensure we were getting the most value out of each filtering scenario. This included granular discussions ranging from reviewing our order sets to discussion around allergy cross reactivity.”
Staff also evaluated the clinical significance for several drug interactions and gained clinical consensus on recommended changes through their P&T Clinical Consensus Group. While they have done this extensively in the past, the dashboard was very helpful for prioritizing and targeting their efforts. As with any major EHR change, this project also involved a lot of testing and validation.
Through the implementation of various mCDS filtering scenarios, Banner Health was able to demonstrate the following reductions in alert firings:
- Reduced mCDS alert rate from 0.79 to 0.54 (a 32% reduction).
- Reduced pharmacists mCDS alerts per alerted day from 128.5 to 85.2 (a 34% reduction).
- Reduced physician mCDS alerts per alerted day from 6.2 to 4.7 (a 24% reduction).
- Reduced 500,000 alerts over the course of a month, which was sustained over a year with a decrease of over 6.2 million alerts.
ADVICE FOR OTHERS
“First and foremost, understand the problem you are trying to solve,” Erickson advised. “There is a difference between simply reducing the number of alerts and streamlining clinical efficiency and patient care. Obviously, there is overlap between those two, but make sure that the entire team is focused on the same end goal.”
The same goes for data review, she added.
“It’s easy to get stuck in ‘analysis paralysis’ and piecemeal the data to the point where the task seems so overwhelming,” said Erickson. “With the end goal of improving patient care and provider efficiency, look at the data for trends, or for the low-hanging fruit. Pick a starting point and take some steps forward. Even if the first step is looking at one medication interaction, take that step and improve that process.”
Then take another step, she said.
“We have many alerts that have been in place for many years, and it would’ve been very easy for us to look at an abundance of data and determine we didn’t have the time or the resources to work on such a large project,” she concluded.
“However, with the focus on improving efficiency, every little bit helps, and it was much easier to take a step forward with small improvements, rather than try to determine every step of the way before we even got started.”