Creating Meaningful Decision Support to Reduce Drug-Drug Interactions

An overabundance of alerts can lead to alert fatigue and compromise patient safety

Exposure to life threatening drug–drug interactions (DDIs) occurs frequently, despite widespread use of clinical decision support (CDS) systems to warn prescribers and pharmacists of potentially harmful medication combinations. Most DDI tools are off-the-shelf systems that use a simple approach of triggering alerts based on the presence of potentially interacting drug pairs in a patient’s medication regimen. Unfortunately, these types of tools have overly sensitive systems that generate inappropriate or nonmeaningful alerts that become excessive. These excessive messages lead to alert fatigue by clinicians who either override or ignore them, which can cause critical alerts to be ignored or missed and result in a major patient safety issue.

Creating more meaningful alerts may reduce drug–drug interactions

Dr. Daniel Malone, an expert in medication safety at the University of Utah, wants to improve these DDI systems by making the alerts more meaningful. He hypothesized that contextualizing DDI alerts to specific patient circumstances would reduce overall alert burden and elevate those alerts that would result in patient harm.

He and a team of researchers identified eight high-priority DDIs that are frequently overridden by prescribers and that can cause significant harm to patients. They developed and validated alert algorithms for these DDIs using EHR data, so that the DDI alert algorithms provide clinically meaningful information to healthcare providers when a patient is at risk but does not interrupt the workflow when the medications are not likely to cause harm.

“The construction of meaningful DDI algorithms will permit healthcare providers, organizations, and systems to provide useful decision support to reduce patient harm due to these drug–drug interactions.” – Dr. Daniel Malone

Integration into CDS systems can support more widespread use

The research team tested the algorithms using real patient data in a simulation study and found that using the eight DDI algorithms would result in a 52 percent reduction in alerts. “This is great for clinicians,” coinvestigator Dr. Richard Boyce notes, “because there will be less alert fatigue, and the alerts are more precise, are more meaningful. You get a reduction in alerts, but the alerts that did go off were more patient-specific.”

One of the major outputs of the project was the creation of the website DDI-CDS.org, which provides documentation for the various algorithms and the apps that were developed, and additional materials to support knowledge sharing with other healthcare organizations and vendors. For example, the team developed and tested three apps for DDI-CDS (https://ddi-cds.org/apps/): one for over-the-counter pain medications in combination with blood thinners (NSAIDS with warfarin), and two apps that examined the impact on the concentration of a target medication in an individual when used in combination with a specific group of different medications (tizanidine and cytochrome P450 1A2 inhibitors, and colchicine and cytochrome P450 3A4/p-glycoprotein inhibitors). The dissemination and integration of these DDI algorithms into CDS systems will support clinicians in decision making and reduce patient harm due to these DDIs.