Combatting Alert Fatigue: Holistically Reducing Noise at the Point of Care

Pharmacy Practice in Focus: OncologyAugust 2015
Volume 2
Issue 3

Adopting and implementing clinical decision support (CDS) technology is a critical element of efforts to advance national quality initiatives and evidence-based practices at the point of care.

How to leverage proven industry methods for combatting alert fatigue.

Adopting and implementing clinical decision support (CDS) technology is a critical element of efforts to advance national quality initiatives and evidence-based practices at the point of care. Unfortunately, alert fatigue remains a primary hindrance to fully leveraging these tools to positively impact patient safety and outcomes.

Alert fatigue occurs when too many red flags are triggered across work flow applications, including electronic medical records, computerized physician order-entry systems, pharmacy systems, or other applications in which a patient order is entered. Although these alerts may not actually create a noise, clinicians can become overwhelmed and desensitized due to the amount of irrelevant data being presented. Ultimately, this phenomenon minimizes the positive impact of CDS technology on care delivery as clinicians ignore or deem alerts unimportant, increasing the possibility that a clinically significant intervention opportunity will be missed.

The good news is that, through a more holistic approach, the industry has made progress in recent years to advance strategies for addressing alert fatigue. Whereas the challenge has historically been identifying the appropriate number of alerts, research suggests that a one-size-fits-all solution is not a realistic course of action due to the differences between various specialties and clinical settings.

Results from the study, Inter-Rater Agreement Among Physicians on the Clinical Significance of Drug—Drug Interactions, revealed that there was little agreement, even between physicians of the same specialty, as to which alerts for drug–drug interactions were clinically significant.1 Leveraging a Fleiss Kappa method to measure the rate of agreement, 16 general physicians were divided into 2 equal groups and given 100 drug—drug interactions randomly selected from the Medi-Span Drug Therapy Monitoring System database to review. In the end, there were only 3 interactions out of the list of 100 that all physicians agreed were clinically significant.

Thus, the first step in addressing alert fatigue is to acknowledge that every clinical environment is different and will have distinct priorities when it comes to the relevancy of alerts. The next step is laying a foundation of technology and governance that can address the unique needs of an identified clinical environment.

To holistically address alert fatigue, consider the following elements:

  • Technology that allows for user controls
  • Ongoing analysis of patient populations and clinical pathway needs in order to customize alert systems for relevancy
  • Identifying and deploying contextual or tiered alerts (including patient data such as age, weight, gender, and renal function) designed to help professionals respond to a patient or situation
  • Ongoing maintenance and updating clinical content to deliver the most current evidence at the point of care

To address ongoing alert fatigue issues, Group Health Cooperative of South Central Wisconsin (GHC-SCW) undertook an initiative in 2015 to minimize the impact of nuisance alerts and reduce the clinician override rate, which had reached 95%. The initiative included a holistic strategy that leveraged industry-accepted metrics and input from clinician staff.

Ultimately, GHC-SCW was able to implement a filtering strategy, based on the latest industry evidence, that reduced the number of alerts firing, resulting in more relevant alerts delivered to providers. Within 60 days, providers were taking action on nearly twice the number of alerts as were documented before the filtering strategy. The organization also improved the distribution of alerts by moving two-thirds of the load to pharmacy, significantly reducing the burden on physicians.

Another example of successfully leveraging a holistic approach to alert fatigue is MetroHealth’s recent effort to implement drug-dose CDS to minimize the potential for dosing errors. Recognizing that dosing errors account for as much as 60% of prescribing errors,2 the health system encompassing 400 primary and specialty physicians filtered alerts by optimizing sensitivity and specificity. MetroHealth was able to reduce the incidence of alerts to 3% of orders, achieving an 80% drop in the number of alerts while maintaining those that were clinically significant.

CDS solutions are a critical aspect of any strategy to improve care delivery and meet quality expectations now and in the future (see the Sidebar “Clinical Decision Support Systems.”) As the industry continues to make significant strides to deliver smarter alerting systems, health care providers and organizations will need to consider how to holistically implement these systems in a way that minimizes the impact of alert fatigue and more fully leverages the potential of CDS.

Chuck Presti, MD, is a clinical product manager with Wolters Kluwer Clinical Drug Information.


  • Strasberg HR, Chan A, Sklar SJ. Inter-rater agreement among physicians on the clinical significance of drug—drug interactions. AMIA Annu Symp Proc. 2013;2013:1325-1328.
  • Kaelber DC, Bates DW. (2007). Health information exchange and patient safety. J Biomed Inf. 2007;40(suppl 6): S40-S45.

Sidebar: Clinical Decision Support Systems

Keith Streckenbach, MBA

At the heart of the alert-fatigue conundrum are the negative consequences of interrupting work flow without changing the course of clinical decision making for the better. One might ask, how can clinical decision support (CDS) systems be designed to interrupt only when they will change the course of the clinical decision for the better?

Provide the decision instead of information, of course.

Let data-driven computer programs do more of the thinking and provide an actual decision instead of information to assist in the decision. Doing so, one could apply machine learning to the big data generated, based on clinical and financial outcomes.

Current pharmacy systems mine data, apply rules to identify a problem, and notify the provider, and may even present best-practice information to facilitate the pharmacist’s decision. Action is not taken until a human intervenes and makes a decision. In a new model, computer programs would make the decision, enact the intervention (eg, fill a lower dose), and retrospectively feed the outcomes (eg, clinical, lab, financial) associated with the decision to the system to allow the machine to “learn” and continuously and iteratively improve its decisions. Employees, in this case pharmacy staff, could randomly spot-check decisions made by the machine, just as they do in automated assembly lines.

By providing information, clinical decision—support systems are intended to make it easier for a provider to make a well-informed decision. CDS systems are not designed to go further and make an actual decision, perhaps in an attempt to avoid the legal costs and innovation speed trap of the FDA regulatory approval process of a medical device. The FDA’s current position on regulating clinical decision support is ambiguous, which leads clinical decision–support developers and entrepreneurs to innovate on the side of avoiding regulatory approval. Interestingly, in its FDA Safety Innovation Act Health Information Technology Report, the FDA states that because CDS risks “are generally low compared to the potential benefits, FDA does not intend to focus its oversight on most clinical decision support. FDA, instead, intends to focus its oversight on a limited set of software functionalities that provide clinical decision support and pose higher risks to patients, such as computer-aided detection/diagnostic software. In the FDA Center for Devices and Radiological Health Fiscal Year 2015 (FY 2015) Proposed Guidance Development, however, draft guidance on medical device decision support software made the “A” list of priorities.

Most of the discussion of this topic has focused on one-to-one alert fatigue, meaning 1 system to 1 provider. The problem is only going to get more complicated and taxing as health care continues its momentum toward population-based health. Advanced decision-support systems are now guiding the actions of not only one provider, but an entire community of care providers and ancillary staff.

In my opinion, the closer decision-support systems can get to actually making the decision, the better we can use technology to measure the value of the decision, accelerate learning of best practices to make better decisions, and greatly reduce alert fatigue.

Keith Streckenbach, MBA, is the CEO of HighFive Healthcare. He formerly served as CEO and vice president of strategy with Pharmacy OneSource.

Related Videos
Practice Pearl #1 Active Surveillance vs Treatment in Patients with NETs
© 2024 MJH Life Sciences

All rights reserved.