From Clinfowiki
Jump to: navigation, search

Alerts may be broken down in to Medication Safety Rules and Non-medication Safety Rules.

Medication safety rules

Non-medication safety rules

  • Diagnosis-Order Rules Drug and level. Postop order sets, disease specific order sets. Suggested dose. Suggested alternate medication for shortage or formulary. Guided dose algorithims for complex orders sucha s those required with insulin and heparin infusions in which nurses are given parameters with which to adjust dose on a regular basis.

Alert fatigue

Alert fatigue is a commonly perceived occurrence with the recent implementation of electronic medical records and specifically clinical decision support.

Alert fatigue is similar to click frustration.


Given that medical errors receive much press in reality many of the errors are secondary to a provider's difficulty with knowledge management. Clearly, the volume of information an average ambulatory provider must remember is too much. The volume increases daily and in order to keep current a provider needs help. Decision support is one type of help that has evolved. As stated by Dr Eric Rose, "where human brains fail, computers excel."

One form of tools to aid the provider is alerts. Alerts can be in the form of "pop-ups," contact-dependent (during access of patient's record), and/or contact-independent (alert "delivered" to provider). The alerts, while found to be beneficial in some cases, can result in a type of "fatigue" whereby the provider, after receiving too many alerts, begins to ignore and/or override the alerts. Receiving too many alerts can result in slowing the provider down rendering the alert useless.

A recent review stated that safety alerts are overridden by clinicians 49-96% of the time (1). For example, in Portland, Oregon the Multnomah County Health Department, which recently implemented an EMR, decided to significantly reduce the number of drug-drug interactions providers were seeing during order entry. The providers felt in order to "get through their daily work," they were forced to override several of the drug-drug interactions.

Likewise, as studied in ambulatory settings alert overrides were secondary to poor specificity and CPOEs need to suppress alerts for renewals of medication combinations that patients currently tolerate (2). By changing the severity level of drug-drug interactions Multnomah County Health Department providers have commented positively on the drug-drug interaction alerts. Some suggestions to avoid alert fatigue are alerts should be not overused, not repeated several times a day, alert gives enough time to make a decision, and creating selectively targeted alerts.

Similarly, Shah NR, et al found, in a 6-month study, that by changing the alert setting to critical/high severity (i.e. high specificity) led to 71% of the alerts being non-interruptive (3). This study and others show the need for the distinction between appropriate and useful alerts. A recent example of a useful alert was the acceptance and highly successful alert of cancelling a medication order when the creatinine clearance of a patient made the medication order not safe (4).

While studies on the cognitive processes ["fatigue"] playing a role in overriding drug safety alerts are lacking, an in depth analysis of the practice/provider's needs may result in a significant "buy-in" resulting in an effective alert with improved outcomes (1). In summary, overriding of alerts is a common practice, but whether "alert fatigue" is a reality remains to be proven. Future studies to address the cognitive effects may elucidate the root of overrides and possibly reveal the perceived "fatigue."

In order to understand click frustration, a basic understanding of alert fatigue is necessary. Alert warnings, reminders, and recommendations have been well defined in CDS [1][2]. Alert fatigue is related to the barrage of message provided by a CDS which can overwhelm a provider and cause them to ignore messages. Stelle et al showed that providers will adhere to some alerts, which can be used to improve patient care and ensure that proper corollary orders are also input [3]. Corollary orders have been shown an improvement of provider adherence to guidelines and a decrease in errors of omission [4]. While disabling drug-drug interactions alerts is one potential method of dealing with the problem of alert fatigue, there seems to be no consensus on which alerts are needed and how to safely disable the alerts [5]. Other studies show positive effects of alerts [6]. There are other studies that show alert threshold can affect provider attitudes [7]. Regardless, CDS with alerts and reminders can improve patient safety and the quality of care [8].

Alert fatigue deals with actions that are needed by the user, responses and messages are generated by the decision support system. Users must act on those alerts and reminders by taking action, whether that action be to implement the support recommendations or to ignore them, a user must react to the system. Click frustration also incorporates user input. Alert fatigue is an active response to CDS and system function. Click frustration is a passive medium from an event driven approach to the EMR. Unlike the active support provided by the CDS, all electronic records are based on the event driven programming paradigm. This paradigm requires users to take an active approach to information management. The user must make decision for system interactions, namely, use the mouse and keyboard for input and decision making tasks. While in this course of events, the mouse click becomes an event which drives the system. While this concept is well known to all who use any type of windows based user interface, the problem is multiplied by the complexity of the electronic medical record.

Amit Shah, MD [As001]

Alert Fatigue Reduction Roadmap

An article from Children’s Hospital in Omaha, Nebraska (Health Data Management 1 Oct 2009: 42) suggests that business intelligence software can fine tune and reduce alert fatigue. The software from Swedish firm Qliikrech reviews frequently triggered alerts and allows a refined breakdown based on user, clinical situation, and action taken on the alert. Clinicians were tasked to determine appropriateness of frequent alerts. In this case only 6.6% of orders triggered alerts after the new review process.

Intermountain Healthcare in Utah provides a more comprehensive roadmap for medication alert reduction (American Journal Health System Pharmacy 1 Dec 2009: 2098-2101). Their goal was to reduce alerts and overrides that were ubiquitous between pharmacist and providers. Strategies included: 1-suppression of alerts based on user or clinical location, 2-evaluate moderate and severe alerts to determine if aligned with local practice, 3-resolve alerts in order sets, and 4-present drug interaction alternatives.

A multidisciplinary group at Partners Healthcare in Boston, MA (Health Affairs Dec 2011 (12):2310-2317) proposed that alerts can be reduced without an increase in liability for EHR vendors or health systems. Vendors have been traditionally loath to remove alerts prior to delivery due to legal concerns and do not provide easy mechanisms for end user customization. EHR vendor sites share the liability concern. However, as of 2011 there were no confirmed product liability cases based on EHR decision support. Lack of FDA support of regulatory requirements has potential to increase liability if systems do not met requirements. However, vendors should be able to reduce alerts as they tend to be held to hold-harmless clauses, providers are considered more responsible, and a negligence model reduces risk relative to a strict liability model as the software is considered a service. Too many alerts and overrides can conversely increase risk. The best way to reduce risk for vendors is to focus on core functionality and for clinical sites to focus on provider competence in a more general manner. International guidelines for alert reduction would reduce risk the most.

The following year an overlapping research group at Partners evaluated the benefits of making drug-drug alerts non-interruptive to reduce alert fatigue (JAMIA 2013;20 489-493). They noted that drug-drug alerts are among the worst offenders with override rates of 90%. They assembled an expert panel of health system, governmental, and vendor representatives to create 33 categories of DDI that can be downgraded to non-interruptive. They suggest that this list be shared nationally to reduce alerts in a vendor neutral manner.

The above articles provide a roadmap to the worthy cause of alert fatigue reduction. They emphasize that alert fatigue causes worthwhile and dangerous medication alerts to be overridden. Therefore reducing alerts is a patient safety imperative. Strategies to consider include: (1) Using specialized software to tailor alerts, (2) Make alerts more sophisticated and focused based on the end user and location and provide medication alternatives (3) Understand mechanisms that mitigate legal risk within the alert reduction process and (4) Consider making some DDI alerts non interruptive based on expert consensus or local expertise.

Noah FInkel MD

Related Articles


  1. Rose E. "Life after Go-Live, Part 4: Preventing Error in an EMR." Journal of Healthcare Information Management. Vol 17, No.4
  2. Krall M. "Clinicians' Assessments of Outpatient Electronic Medical Alert and Reminder Usability and Usefulness Requirements: A Qualitative Study." May 2002.
  3. Heleen van der Sijs, et al. "Overridding of Drug Safety Alerts in Computer Physician Order Entry." J Am Med Inform Assoc. 2006;13:138-147.
  4. Weingart, et al. "Physicians' decisions to override computerized drug alerts in Primary Care." Arch Intern Med. 2003 Nov 24;163(21):2625-31.
  5. Shah NR, et al. "Improving Acceptance of Computerized Prescribing alerts in Ambulatory Care." J Am Med Inform Assoc. 2006 Jan-Feb;13(1):5-11.
  6. Galanter, et al. "A trial of Automated Decision Support Alerts for Contraindicated Medications Using Computerized Physician Order Entry." J Am Med Inform Assoc. 2005; 12:269-274.
  7. Anderson HJ Avoiding 'Alert Fatigue'; A children's hospital uses business intelligence software to fine-tune how CPOE functions." Health Data Management 1 Oct. 2009: 42. Academic OneFile. Web. 22 Oct. 2015.
  8. Cash JJ Alert Fatigue Informatics Exchange; Am J Health Syst Pharm Vol 66 Dec 1 2009
  9. Kasselheim AS Health Affairs, 30, no.12 (2011):2310-2317 Still Minimizing The Risk Of LitigationClinical Decision Support Systems Could Be Modified To Reduce 'Alert Fatigue' While
  10. Phansalkar S, vander Sijs H, Tucker AD, et al.J Am Med Inform Assoc 2013;20:489–493 Drug–drug interactions that should be non interruptive in order to reduce alert fatigue in electronic health records