Overriding of drug safety alerts in computerized physician order entry
Overriding of Drug Safety Alerts in Computerized Physician Order Entry was investigated in order to determine the reason physicians override safety alerts, the effect that overriding alerts has on patients and the article offers suggestions for improving safety alerts within the decision support system.
Using MEDLINE and EMBASE databases seventeen publications were found addressing the overriding of safety alerts in CPOE. It was found that 49% to 96% of all safety alerts are overridden.
Overriding alerts does not necessarily lead to medication errors. The authors point out that alerts are "justified" and "unjustified". Justified overriding of alerts are those overrides that occur when the physician uses his/her personal knowledge in caring for the patient. Overriding alerts can result in adequate therapy for a patient as well as medication errors. Unjustified overriding does not necessarily produce adverse events and justified overriding does not always prevent adverse events. Studies presented on the outcome of overriding alerts did not reveal adverse drug events in more than 97% of the cases, however, the overrides that do pose the greatest problem are unjustified alerts. These are the overrides that occur due to ignoring alerts, misinterpretation or wrong selection.
Three studies contributed the most important reason for overriding safety of alerts was due to “alert fatigue”. Alert fatigue was explained as being the mental state that is the result of too many alerts consuming time and mental energy. Alerts that were frequently displayed was considered not serious, irrelevant, or shown repeatedly.
The article explains other conditions that lead to a high override rate which are distrust in the system, lack of motivation, lack of time and the interruption of the workflow process. The lack of sufficient training for the physicians using the clinical decision support safety alerting system can contribute to some of these conditions. As a result of a high override rate, healthcare organizations may choose to disconnect a group of safety alerts which can result in relevant alerts no longer being active.
Areas that need to be corrected in order to help decrease the unjustified overrides within the alerting system are low sensitivity and specificity within the system and also provide clarity of the information within the alert. Two studies revealed that 36.5% and 39% of the alerts were inappropriate (false positive) in regards to the patient in question, thus decreasing specificity of the system. The first recommendation made by the authors was to increase specificity by reducing the amount of false positive alerts by excluding irrelevant and no urgent alerts. Such alerts would be interactions between systemic and topical drugs, alerts of drug allergy in case of medication intolerance, and alerts without urgent or possible action. Alerts that are considered low specificity alerts can be detoured to the pharmacist or nurse to be reviewed and discussed with the physician if needed.
Decreased sensitivity can be a result of alerts being turned off, an incomplete knowledge database and inaccurate patient information in the system such as body weight and drug allergies. The use of free text will also affect the sensitivity of the alerting system. Customizing the alerts to be patient specific according to age, body weight and gender can increase sensitivity, however; this process can be time consuming and have the potential to increase the number of inappropriate alerts thus decreasing specificity. To improve sensitivity the authors suggest using coded data to choose a drug name and regimen and keeping patient information current by updating the drug database with known allergies and drug interactions. As an alert is generated the patient's laboratory findings included with the alert can also help to increase the sensitivity of the alerting system.
Comments: The author suggests that physicians entering the reason for overriding an alert could help in adjusting the knowledge database thus improving the alerting system. It was also stated that oversight or forgetfulness are the primary reasons for the system to generate an alert and does not reflect ones knowledge on the subject. A well designed decision support can ultimately improve patient safety. Entering the reason for an override in coded data can prevent the system from showing a duplicate alert. SS
14:00, 21 October 2006 (CST)
An article published in the Journal of the American Medical Informatics Association by Van Der Sus, Heleen., et al. focused on reviewing literature on overriding drug safety alerts in computerized physician order entry (CPO). The authors utilized the Reason’s framework of accident causation to interpret their results. Seventeen papers met the author’s qualifications for inclusion in the review. This study sought to identify the issues facing clinical decision-making in correlation with computer aided alerting systems.
This publication was produced as a review of articles written on overriding clinical decision support system alerts and reminders. Many computerized physician order entry systems are designed to incorporate decision support systems that serve as a reminder to physicians entering patient orders of potential patient safety issues. Since clinical decision support has become and integral part of electronic ordering there has been studies done of its impact on patient care and clinical best practices. The authors sought to collectively review the articles published on decision support systems “to provide insight into physicians’ handling of safety alerts.” The reviewers asked a serious of questions one being: “How often and in what situations are safety alerts overridden?” The reviews concluded that except for serious alerts for overdose that alerts were overridden 49% to 96%. The article also mentioned that a study revealed that the override rate increased from 50% to 75% over a five-year period. The second question asked by the reviewers was “Why do physicians override them?” The authors concluded that more often than not physicians were suffering from “alert fatigue” which is what happens when they are exposed to a burdensome Decision Support System (DSS). They also stated that this fatigue causes physicians to ignore alerts that could be important to patient care. Yet another questions asked by the reviews was “What effects ensue?” The results were as follows: “Adverse events were observed in 2.3%, 2.5% and 6% of the overridden alerts, respectively, in studies with override rates of 57%, 90%, and 80%.” “Adverse events were preventable in 0.8% and none of the overrides, respectively.
The authors concluded by recommending that “entry of override decisions to gain a deeper insight into (justifying) override.” “Specificity or sensitivity should be increased as the result of consensus meetings between physicians and pharmacists.” Finally, “Decision support in CPOE can be a good tool to improve patient safety but can also hamper patient safety if it is poorly designed.”
Comments: Clinical decision support have been both praised and cursed by clinical users as helpful and hindering to patient care. We must continue to develop clinical decision support systems, however it is our duty as informatics professionals to design these systems in collaboration with users and reviewing lessoned learned and best practices from those who have gone before. I agree with the authors that DSS in CPOE can and is a good tool to improve patient safety, and yet we must also be aware that it can hinder patient care if we do not design it properly.
