All Content

Alert fatigue contributes to task burden and can threaten patient safety. In this study, researchers at one academic medical center found that machine learning techniques could enable intelligent filtering of medication alerts and reduce alert volume by 54%.

Clinical decision support systems (CDSS) can help identify potential drug-drug interactions (DDI), but they can lead to alert fatigue and threaten patient safety. Based on an analysis of DDI alerts and survey data regarding physician experience using the DDI module in the CDSS, researchers identified barriers (i.e., lack of patient-specific characteristics and DDI-specific screening intervals) that contribute to false-positive alerts and alert fatigue.

Computerized decision support alerts for drug-drug interactions are commonly overridden by clinicians. This study examined fifteen well-known drug-drug interactions and identified risk factors that could reduce risk in the majority of interactions (e.g., medication order timing, medication dose, and patient factors).

Computerized prescriber order entry (CPOE) systems embedded in electronic health systems alert clinicians to potential safety concerns such as drug-drug interactions or medication dosage errors. Results of this review indicate that alerts influenced prescriber behavior in most of the included studies. However, it is unclear whether these behavioral changes improve patient safety outcomes. Recommendations for future research include randomized controlled trials to determine which alerts maximize patient safety, while minimizing prescribers’ alert fatigue.

Medication errors occur frequently in ambulatory care settings. This article describes the development and testing of an ambulatory medication safety evaluation tool, which is based on an inpatient version administered by The Leapfrog Group. Pilot testing at seven clinics around the US indicates that clinics struggled in areas of advanced decision support such as drug age and drug monitoring, and that most clinics lacked EHR-based medication reconciliation functions.

Clinical decision support (CDS) alerts can improve patient safety, and prior research suggests that monitoring alert overrides can identify errors. Over a one-year period, this study found that medication-related CDS alerts associated with renal insufficiency were nearly always deemed inappropriate and were all overridden. These findings highlight the need for improvements in alert design, implementation, and monitoring of alert performance to ensure alerts are patient-specific and clinically appropriate.  

Computerized provider order entry (CPOE) can prevent prescribing errors, but patient safety threats persist. Based on qualitative interviews with multidisciplinary prescribers, the authors identified several issues related to CPOE interacting within a complex prescribing environment, including alert fatigue, remote prescribing, and default auto-population of dosages.

Machine learning can improve the accuracy of clinical decision support (CDS) tools. This single-site study used data from the electronic health record (EHR) and clinical pharmacist review to test the accuracy of a hybrid CDS system to identify prescriptions with high risk of medication error. The machine-learning based approach was more accurate than existing techniques such as the traditional CDS system and can improve the reliability of prescription checks in an inpatient setting.  

Using data from the Computerized Physician Order Entry (CPOE) Evaluation Tool, this study compared hospital performance against fatal orders and nuisance orders. From 2017 to 2018, overall performance increased and fatal order performance improved slightly; there was no significant change in nuisance order performance; however, these results indicate that fatal alerts are not being prioritized and that over-alerting in some cases may be contributing to alert fatigue.

Smart pumps have been adopted as one approach to preventing medication errors, but less is known about their use in pediatric populations and contribution to NICU alert fatigue. This study examined NICU smart pump records from 2014 to 2016 and found that pump alerts do not contribute significantly to overall alert burden in the NICU, and alerts tended to cluster around specific patients and medications (such as fentanyl, insulin and vasopressin). The study also identified 160 attempts to exceed the programmed dosing limit; while these represented a small number of violations over the entirety of the study period, the attempts involved high-risk medications (including fentanyl, insulin, and morphine) and doses programmed at 5- to 24-times the maximum dose which could result in significant adverse patient outcomes.

Alerts designed to prevent inappropriate prescribing of medications are frequently overridden and contribute to alert fatigue. This study describes the use of machine learning to improve the clinical relevance of medication error alerts in the inpatient setting.

This systematic review examined the override rates of several different clinical decision support approaches. Researchers conclude that role tailoring—the provision of different alerts to prescribers versus pharmacists—was the most successful method to reduce alert fatigue. They recommend redesigning decision support to reduce alert fatigue.

Although clinical decision support is intended to improve safety, decision support alerts often result in alert fatigue and overrides. This prospective observational study examined overrides for exceeding the maximum dose of a medication in the intensive care unit. Researchers determined that insulin was the most frequent medication for which a maximum dosage alert was overridden. In almost 90% of cases, the overrides were deemed clinically appropriate. The authors conclude that more intelligent clinical decision support for medication dosing is needed to balance safety with alert fatigue in the intensive care unit. A past PSNet perspective discussed the challenges of implementing effective medication decision support systems.