Loading Curie Briefs...
Getting the latest healthcare news for you
Getting the latest healthcare news for you
A new randomized trial found that an AI-powered tool accurately flagged hospitalized patients at high risk for acute kidney injury (AKI), but triggering early nephrology consultations didn't actually improve outcomes. Nearly 39% of flagged patients developed AKI anyway, and clinicians followed early recommendations far less often than in usual care. The takeaway: prediction is not prevention.
A randomized clinical trial published in JAMA Network Open tested whether an AI-driven risk score (ESTOP-AKI) could help prevent acute kidney injury (AKI) in hospitalized patients by prompting early nephrology consultations — before any kidney damage was detectable. The result? The AI identified high-risk patients effectively, but the intervention didn't move the needle on outcomes.
The trial enrolled 180 high-risk patients at the University of Chicago, splitting them between early nephrology consultation and usual care. Despite generating far more recommendations (270 vs. 36), the early consultation group showed no significant improvement in peak serum creatinine change, AKI development, need for dialysis, or 90-day mortality compared to usual care.
By the Numbers:
Why it matters: The study highlights a critical gap in AI-assisted clinical care — predicting risk is only the first step. When clinicians don't perceive urgency (because creatinine hasn't visibly spiked yet), recommendations go unheeded. Future research needs to pair smarter prediction with targeted, actionable interventions and better implementation strategies to close the loop between AI alerts and real patient benefit.
