Mixed Results With Alerts for AKI-Related Medications in Hospital

While medication-targeted acute kidney injury (AKI) alerts led to certain kidney-affecting medications being stopped, the overall clinical impact was limited, a researcher reported.

The open-label, randomized trial included over 5,000 patients hospitalized for AKI with an active order for one of the medication of interest (MOIs) classes that were selected as being possibly nephrotoxic -- non-steroidal anti-inflammatory drugs (NSAID), renin-angiotensin-aldosterone system inhibitors (RAASi), or proton pump inhibitors (PPI) -- explained F. Perry Wilson, MD, MSCE, of Yale School of Medicine in New Haven, Connecticut.

Although NSAIDs are frequently discontinued in AKI, RAAS inhibitors are only sometimes stopped, while PPIs are rarely discontinued, Wilson explained during a presentation at the American Society of Nephrology Kidney Week.

Half of these patients had automated alerts pop-up with a medication order entry, flagging these medications for potential cessation. These alerts were effective, increasing the relative risk of MOI discontinuation by 9% within 24 hours of randomization versus patients who didn't have these automatic alerts pop up (RR 1.09, 95% CI 1.04-1.14). MOIs were discontinued in 61.1% of the alert group versus 55.9% of the usual care group.

When broken down across MOI class, all three saw a significantly greater discontinuation associated with the alerts, with the biggest impact seen with PPIs:

• NSAID: RR 1.09 (95% CI 1.04-1.14)
• RAASi: RR 1.14 (95% CI 1.08-1.21)
• PPI: RR 1.26 (95% CI 1.09-1.45)

While the alerts were effective at bolstering medication discontinuation, they didn't really seem to have all that large of a clinical impact. Missing the trial's primary outcome, these medication alerts didn't reduce the risk for a composite outcome of progression of AKI, dialysis, or death within 14 days or hospital discharge (overall RR 0.92, 95% CI 0.83-1.01). This occurred in 23.1% of the alert group versus 25.3% of the usual care group.

The composite outcome wasn't reduced by these alerts when isolated to those just on NSAIDs (RR 0.92, 95% CI 0.76-1.12) or RAASi (RR 0.99, 95% CI 0.85-1.16). There was a significant benefit seen for patients on PPIs (RR 0.88 ,95% CI 0.79-0.98).

This finding led the researchers to perform a post-hoc analysis on what made PPI users different from the rest of the cohort, Wilson explained. "They tend to be sicker, they were much more likely to be in the ICU -- 28% versus 14% -- and had a higher modified SOFA [sequential organ failure assessment] score at the time of randomization."

Wilson suggested that alerts could most benefit patients on PPI as the latter may be an underrecognized contributor to AKI in hospitalized patients due to interstitial inflammation.

"Importantly, this also represents a unique phenotype," he said. "PPI might be a marker of severity of illness."

Kidney Week session co-chair Karen Griffin, MD, of Loyola University Medical Center in Chicago, said the study demonstrated how clinicians can "leverage the power" of electronic health records system. But alert fatigue is something to be concerned about, she noted.

"We're well aware of alert fatigue," Wilson said. "The reason we're doing these trials is to reduce the number of alerts. Most clinicians that we speak to say if we can prove that it improves patient outcomes, that they would then actually embrace these alerts. What their concerned about is alerts that they don't think make a difference."

"I will say also at Yale, the committee that oversees alerts has a policy of discontinuing two alerts for every one new alert that they implement, which I think is another good idea for health systems to consider to avoid alert fatigue," he added.

Wilson and colleagues reported that, compared with usual care, hospitalized patients with these medication alerts didn't see a significant association with any of the key secondary endpoints of inpatient kidney consult, progression of AKI, progression to stage 2 or 3 AKI, need for dialysis, 30-day readmission, or death. The two groups also had a similar duration of AKI and median length of stay.

Safety wasn't different for any of the three MOI classes in regards to alert versus usual care patients.

There also weren't any significant differences seen in the primary outcome in regards to most prespecified subgroup analyses looking at certain age groups, sex, race, medical versus surgical admission, and ICU versus non-ICU patients. The only subgroup that saw a significant benefit associated with these alerts were those with a baseline creatinine below 0.5 mg/dL, which Wilson pointed out has also been seen in two previously alert studies.

"I think it may be a real finding," he said. "It has some biologic plausibility in that the people with a low baseline creatinine, when they develop AKI, may have a creatinine that rises to a level that is still within the normal limit and can then therefore fly under the radar of clinicians."

The four-center trial enrolled 5,060 hospitalized patients with KDIGO stage 1 AKI. About half were male, 19% were Black, and the median age was 70. End-stage kidney disease, an initial hospital creatinine of 4+ mg/dL, or dialysis within a year prior were some of the exclusion criteria.

A total of 31% of the cohort had an active order at baseline for an NSAID, 53% for an RAASi, and 65% for a PPI.

Comment