Turning down the noise: Hospital Sisters Health System reduces alert fatigue
When our hospital system implemented clinical decision support (CDS) in the computerized provider order entry (CPOE) system in a few years ago, leaders and clinicians were expecting to experience an improved CDS environment. However, we quickly realized that medication alert fatigue could affect the entire clinical quality improvement initiative at Hospital Sisters Health System (HSHS).
When we started using CPOE in our electronic medical record, our clinicians were bothered by too many alerts and that was preventing them from fully utilizing the system. Even though these alerts called attention to important information, the large volume was simply wearing our clinicians down. The result: Clinicians began to simply ignore the alerts, and became frustrated since they were not seeing the value they expected to see from the CDS.
To adequately get utilization to the level where it would have a positive impact on patient care and also meet Meaningful Use requirements, we needed to first address the alert fatigue issue. Our goal was to make every interaction with the system meaningful for our providers and pharmacists. To accomplish this, we wanted to find a way to get that “just-right” level of relevant alerts and then, perhaps more importantly, a means to implement a system that optimally leverages CDS at the point of care.
We set out to substantially bring down the number of alert fatigue events in hopes of providing our clinicians with a system that truly supports patient care. More specifically, we set out to increase alert specificity, avoid inappropriate alerts, avert unclear information, and improve usability.
A complicated challenge
The challenge turned out to be more difficult than it seemed. For example, because patients are extraordinarily complex, an alert that might be clinically significant for one may not be significant for another. Similarly, an alert that is relevant in the ambulatory environment might not be appropriate in the acute care setting.
To further add complexity, our HSHS hospital system is comprised of 14 hospitals throughout Illinois and Wisconsin, with 10 institutions sharing the same clinical information. With the vastly different patient populations represented by each institution – ranging from critical access to academic center – the issue of alert fatigue cannot be solved without consensus and cooperation.
To kick off the alert fatigue reduction initiative, our pharmacy department staff identified alert override patterns, with the hope that they could eventually zero in on potential changes that could help reduce alert fatigue.
Drug interaction alerts bubbled to the top of the remediation list as many of these alerts lack specificity, are missing necessary data, and do not provide clear action steps. In addition, drug interactions are often influenced by multiple variables and may warrant additional fine tuning. The pharmacists’ investigation revealed that the hospital system generated 243,803 drug interaction alerts annually. This resulted in 668 overridden alerts per day.
Even though our pharmacists could identify the alerts that were causing the fatigue, they could not start whittling down the alerts without seeking input from other clinicians. As a result, we had to put processes in place that made it possible for a wide spectrum of colleagues to evaluate each alert. Eventually, we expanded the scope of the initiative to include drug allergy, dose range checking, duplicate therapy, and drug-disease interactions as well.
To aid us in this process, we started working with FDB AlertSpace, a web-based solution from First Databank that enables clinicians to collaboratively build on their own localized experience in fine-tuning medication alerts, as the right solution to take our efforts to the next level of specificity and customization. This system enables us to keep track of customizations, create an audit record, and load the results of modifications directly into the decision support system for immediate use in the workflow.
As the hospital system implemented this alert management tool, we realized that we could not alter alerts without first ensuring that any recommended changes would be warranted and appropriate. So, we determined that a comprehensive alert approval process was in order.
The process we chose to adopt includes these nine steps:
- An informatics pharmacist runs a report against their data repository that includes all medication alerts triggered in the CPOE system for a specific time. These reports are then compiled into a spreadsheet.
- The informatics pharmacist reviews the medication alerts spreadsheet and selects candidates for modification or inactivation. Priority for analysis is given to alerts that fire more than 250 times per quarter.
- The informatics pharmacist sends an email outlining the alerts for review to the members of a Medication Alerts Workgroup for approval.
- The decision from the Medication Alerts Workgroup is submitted to a Medical Informatics Committee for final approval.
- Education on alert changes is presented to the Pharmacy IT Workgroup.
- The informatics pharmacist publishes the changes in FDB AlertSpace. Changes are included in the next data download from First Databank.
- The First Databank data download is performed in the test environment by a pharmacy analyst.
- Alert changes are confirmed and quality control checks are performed in the test environment.
- Alert changes are moved to the live environment in the next scheduled First Databank data load.
Realizing results
With this medication alert solution and process in place, we have reduced alert fatigue in the following categories:
- Drug-drug interaction. Alerts related to drug-drug interactions have been reduced system wide by 110,904 alerts per year. This represents a 45.5 percent reduction in annual drug-drug interaction alerts.
- Dose range checking. Based on data from one institution, we have turned off 134 dose range checking alerts (5.9 percent) and deceased monthly alerts by 550 (60 percent). Additionally, 23 dose range checking alerts were fine-tuned.
- Drug-allergy interaction. We eliminated one allergy cross sensitivity alert, which accounted for 51 percent of our allergy alert fatigue.
- CPOE effectiveness. Eliminating alerts made it possible to leverage the CPOE system to eliminate more than 700 preventable errors monthly and realize a cost avoidance of $300,000 per month.
By decreasing our alert fatigue at HSHS, we can now better focus in on clinical decision support and making meaningful clinical interventions for our patients.