Improving the outlook through better management of alert fatigue
By Raj Gopalan, MD, MSIS, Vice President of Innovation and Clinical Informatics, Wolters Kluwer
When a viral outbreak threatens or claims hundreds of lives, it dominates the news cycle. Yet a more destructive, persistent, and deadly threat receives much less attention — medical errors.
Comparable to the recurring impact of an ongoing public health disaster, medical errors — defined as unintended acts, execution errors, and care planning mistakes — cause the deaths of nearly 550 people daily. In fact, research published in the British Medical Journal Medication named medical errors as the third leading cause of death in the U.S.
Adverse drug events (ADEs) — or injuries resulting from taking a medication — are one of the three most common and harmful categories of medical errors. Each year in the U.S. there are approximately two million ADEs that cause a staggering 100,000 deaths and increase healthcare costs by about $136 billion.
While industry awareness of the problem has existed for some time, efforts to reduce preventable medication errors have met their share of challenges. For instance, clinical decision support (CDS) and alerting mechanisms introduced with EHRs are designed to reduce the potential for ADEs by providing clinicians with potentially relevant warning information at the point-of-care. However, in many cases the benefits clinicians receive from these systems have been compromised because they generate too many alerts and cause alert fatigue — a psychological phenomenon that arises when clinicians are exposed too frequently, to too many alerts that are not relevant to their patient care issues. Once alert fatigue sets in, the effectiveness of warning information declines because a clinician is more inclined to override alerts without serious consideration.
Industry research points to a direct link between overrides and medication errors. For example, an analysis of medication errors reported through the Pennsylvania Patient Safety Reporting System identified 583 medication error events in one year where a clinician overrode an automated alert that could have helped the clinician notice and avoid the error. Simply put, too much “noise” at the point of care is drowning out real opportunities to enhance patient care by reducing medication errors.
Overrides caused by alert fatigue also contribute to therapeutic complications and higher costs. For example, if a clinician disregards a medication allergy alert when administering a medication and the patient goes into shock, the patient will require additional medications, extra provider time and a longer hospital stay. A provider may ultimately need to absorb these unnecessary costs that can negatively impact their bottom line.
Fortunately, technological advances, better collaboration, and a more holistic approach to system and content development are improving the quality and reducing the quantity of alerts. By combining functionality that considers contextual patient information with better filtering and user input in developing strategies to suppress irrelevant alerts, healthcare organizations are helping clinicians make better decisions and identify important patient safety issues at the point-of-care.
CDS Alerts: Opportunities And Challenges
Rapid adoption of EHRs in recent years has increased the healthcare industry’s awareness of the scope and effect of medication errors. Data housed in these systems has provided new information about errors that could not be identified by reviewing paper charts.
The insights from this new information have allowed medication-related alerting systems to address common sources of error including dosing, interaction, duplications, and allergies. These same areas are also the focus of new CDS regulatory requirements introduced in recent years. Examples of these types of insights include knowledge that: (a) drug dosing errors tend to be common in pediatric patients where body mass index varies greatly; (b) drug interactions commonly occur when a combination of drugs impacts the normal metabolism of particular medications, elevating drug levels and toxicity; and (c) because of duplication of active ingredients in combination medications, patients can potentially receive double or triple the safe dose.
CDS and embedded drug data solutions help address these issues by identifying potential drug errors and alerting physicians at the point of ordering. Historically, most systems worked in generalities that were not focused on unique patient characteristics, producing a large number of false positives alerts that clinicians (and particularly specialists) found irrelevant to their patients. With alerts designed around generalized drug data, important patient-specific factors, including the condition being treated and co-morbidities (e.g., diabetes, kidney failure, etc.) were not considered by the system before providing alerts. For instance, while a certain combination of drugs is deemed toxic for the average patients, complex conditions such as diabetes or kidney failure can warrant such choices.
The end result is somewhere between 40 and 90 percent of alerts are overridden according to industry data. This type of alert response is problematic since industry data also supports that 50 percent of alerts are valuable and relevant to patient care. Thus, the industry is challenged to advance CDS strategies such that clinicians pay attention to alerts and act on them.
A Holistic Approach To Alert Fatigue
Advanced drug data solutions are already addressing the need to screen drug alerts by patient, age, gender, diagnosis, lab results, and medications prescribed. That is an important first step to building context.
The next steps require teamwork, collaboration, and specific functionality. Clinical and IT teams must come together to identify ways to customize, filter, and suppress alerts based on clinical evidence and patient risk. This strategy begins with a basic understanding of how many alerts are firing, factors that contribute to high volumes of alerts, and why alerts are being overridden.
Advanced analytics infrastructures exist that support these efforts, providing a dashboard showing how clinicians are responding to alerts. By leveraging the right infrastructure, healthcare organizations can identify the top 20 adverse drug events that generate 80 percent of the alerts and uncover patterns such as patient demographics, disease states, provider specialty etc., that contribute to alerts that fall into the top tier. Often, specific patient profiles emerge, and clinical and IT teams are able to work together to suppress unneeded alerts or apply user customization to improve relevancy.
Nephrology patients, for example, tend to have complicated problems. Patients receive a lot of drugs with multiple compositions, often creating false positive or irrelevant alerts. As such, if a clinical team determines a particular alert is inappropriate for patients 40 to 60 years old who are located on the intensive care unit with kidney problems and on continuous monitoring, a trigger for that alert is no longer necessary. Ohio-based MetroHealth System conducted an analysis of it drug-dose alerts, leveraging its EHR and drug database system, to test multiple system-wide and drug specific strategies, ultimately identifying suppression techniques that decreased baseline drug-dose alerts by 80 percent and notably improving alert fatigue.
In terms of functionality that supports a holistic approach to alerts, healthcare organizations should consider systems that allow for user controls on organizational, departmental, patient profile, and specialty context levels. Functionality that supports the deployment of tiered alerts based on the above criteria as well as methods for updating clinical content on a regulator basis to ensure decision-making is based on the latest evidence.
The potential for CDS to impact patient care is significant. By implementing systems that support a holistic approach to alerts, healthcare organizations can make decision support more impactful and improve the outlook on medication errors.