Leveraging alert logic parameters to minimize noise at the point of care
Alert fatigue remains one of the greatest hindrances to optimal use of clinical decision support (CDS) at the point of care. An ongoing battle of information overload, characterized by electronic health records (EHRs) and other clinical systems firing off a whole host of warnings that physicians believe are not relevant to a patient’s care, is perhaps the greatest impediment to fully leveraging the potential of CDS technology.
The industry at large recognizes the potential of CDS to impact care delivery in a positive way, especially because at least one recent study found that medical errors could be the third leading cause of death in the United States. These errors are attributed to 250,000 deaths annually and include medication errors associated with dosing mistakes, drug interactions, drug duplications or drug allergies.
By providing point-of-care warnings to physicians, the hope is that effective CDS systems can help reduce the potential for errors to go undetected. Unfortunately, when there is a high volume of alerts and a high number of false positives, a phenomenon known as alert fatigue can desensitize clinicians to potentially useful information when a clinically significant intervention opportunity presents itself.
The good news is that the industry has made progress in recent years to advance strategies for addressing alert fatigue through a more holistic approach. Emerging models address the need to factor in both the content that drives these alerts as well as sociotechnical factors such as design and workflow considerations of how and when these alerts are displayed. Improving the specificity of alerts relies on being able to make meaningful connections between the medication knowledge base and the EHR, thus producing smarter, more relevant alerts that consider the context of patient and physician preferences.
Finding the Right Balance
Finding the right balance of the volume of alerts to display has been an elusive pursuit for EMR and CDS vendors. Balancing patient safety and the clinical significance of alerts with a clinician’s ability to consume information is subjective and complex. One approach to tackling the problem of alert fatigue that has shown promising results is a vendor-assisted option that leverages evidence-based classifications and guidance for filtering and suppression settings.
At Group Health Cooperative (GHC) of South Central Wisconsin, the use of filtering mechanisms to improve alert logic proved to be an effective strategy. The organization is addressing alert fatigue through a filtering system built on content and evidence-based guidelines from Wolters Kluwer. GHC’s drug reference solution, Facts & Comparisons, classifies interactions by severity, assigning a significance score to each alert between one and five based on industry evidence. In this case, designations of one and two are backed by credible evidence pointing to an interaction that is “probable” or “suspected.”
GHC then utilized this data within Medi-Span to set filters to trigger only those alerts with a significance score of one or two, reducing the number of alerts firing from 87% to 27%. At the beginning of the alert fatigue initiative, GHC clinicians were receiving approximately 143 alerts per one hundred orders. Fourteen days after go-live of the filtering initiative, that number was reduced to 52 per 100. Physician override rates decreased from approximately 95 to 100 percent to 92.2% at the 60-day mark.
Ohio-based MetroHealth System embarked on a similar filtering initiative to address alert fatigue by focusing on drug-dose alerts. A preliminary analysis indicated that alerts would fire on approximately 13% of all medication orders, resulting in avoidable alert fatigue caused by a preponderance of low-risk alerts. To optimize low-impact drug-dose alerts, the team leveraged its Epic EHR and Medi-Span knowledge base to test multiple system-wide and drug specific strategies to filter out alerts that physicians deemed clinically insignificant. By focusing on patient safety related drug-dose alerts that could potentially cause significant patient harm, baseline drug-dose alerts initially decreased by nearly 80%. Primary system approaches decreased alerting to 5% of orders, while secondary drug-specific approaches dropped them to 3%. As end user feedback was incorporated into the process, alerts declined to under the 3% threshold.
These examples show that customized evidence and experience based solutions may work better than out-of-the box solutions to help users optimize benefits from CDS alerts. There are already available solutions that provide customization functionality that can help end users interact with and customize medication-relation decision support knowledge bases in ways that better suit their needs.
Improving Alert Logic
Healthcare organizations can take similar steps to address the challenges of alert fatigue, but the key to any attempt to change alert logic starts with an acknowledgement that clinical environments are different and priorities regarding the relevance of alerts will differ depending on the setting. There is no “one size fits all” approach. Some high-level considerations for implementing holistic alert fatigue strategies include:
- EHR technology that allows for user controls, whether that be at an organizational, departmental or end-user level
- Employing design strategies and consideration of human factors to guide the presentation of alerts
- Ongoing analyses of patient populations and clinical workflows need to be taken into account when customizing alerting systems
- Identification and deployment of contextual or tiered alerts (using patient data such as age, weight, gender, renal function, etc.)
- Ongoing maintenance and updating of clinical content to deliver the most current evidence at the point of care
CDS alerts are a big benefit that clinicians can get from implementing EHR technology. By understanding the many facets that need to be taken into account to optimize CDS alerts, we can improve patient safety by harnessing the full potential of alerts to help clinicians provide better care.