Study: Early Nephrology Consultation Triggered by AI Fails to Prevent Acute Kidney Injury

A structured early nephrology consultation (ENC) triggered by a real-time machine-learning acute kidney injury (AKI) risk score did not improve outcomes for hospitalized patients at high risk for stage 2 AKI, according to results of a randomized clinical trial published in JAMA Network Open. The single-center study, conducted at the University of Chicago, found no significant difference in the peak change in serum creatinine (SCr)—the trial's primary outcome—between patients who received the intervention and those who received usual care.¹

“In the setting of early AKI or high AKI risk, drug and nephrotoxin stewardship has been shown to prevent the development of AKI and/or prevent AKI progression,” Jay L. Koyner, MD, professor of medicine in the Section of Nephrology at the University of Chicago and study author, told Pharmacy Times in an interview. “However, we can see from our study that there are lots of missed opportunities to intervene and provide good AKI care.”

Trial Design and Patient Population

Investigators used a gradient-boosted machine learning tool called the electronic signal to prevent AKI (ESTOP-AKI) to identify hospitalized adults without SCr-based AKI who were at elevated risk of developing the condition. Of 3929 patients screened, 180 were randomized 1:1 to either ENC, which included an in-person nephrologist assessment addressing volume status, kidney perfusion, medication dosing, electrolytes, nutrition, and further testing, or usual care, in which a nephrology consult occurred only if requested by the primary team. Enrollment spanned March 2019 through August 2024, with a pause during the COVID-19 pandemic.¹

No Difference in Kidney Outcomes

There was no significant adjusted difference in 7-day ΔSCr between the ENC and usual care arms (0.04 mg/dL vs −0.03 mg/dL; P = .30). Rates of any AKI development (42% vs 36%; P = .47), stage 2 or higher AKI (19% vs 13%; P = .28), need for kidney replacement therapy, and inpatient or 90-day mortality also did not differ significantly between groups. Investigators noted that patients on the hospital ward randomized to usual care were more likely to be readmitted within 90 days than those in the ENC arm, but overall readmission rates were not significantly different.¹

Recommendation Adherence Was Low

Among 306 total consultation recommendations across both arms, only 48% were completely followed. Notably, recommendations involving medication dosing, discontinuation, diuretics or fluids, and vasopressors were followed completely more often in the usual care arm (68%) than in the ENC arm (41%). Study authors, led by Koyner, suggested that this low adherence may have contributed to the lack of measurable benefit, noting that recommendations made before overt SCr-based AKI develops may be viewed by care teams as less urgent than those made after a diagnosis is established.¹

Context Within the Broader AKI Literature

The authors pointed to similarities with other trials testing structured interventions for AKI prevention or management. A 2024 trial of a kidney action team found only 33.8% of recommendations were followed, with no reduction in AKI progression. Two United Kingdom-based trials evaluating care-bundle approaches for established AKI likewise reported adherence rates below 50% alongside null or modest findings on AKI progression. By contrast, the international BigpAK-2 trial, which used a biomarker-guided Kidney Disease | Improving Global Outcomes care bundle in postoperative patients, reported a reduction in stage 2 or 3 AKI incidence alongside 47% bundle adherence in the intervention arm, suggesting that implementation fidelity may be central to whether these interventions succeed.^2-5

Implications for Practice

The authors concluded that the negative results should not be interpreted as a failure of machine-learning risk-prediction tools themselves, but rather as evidence that such tools need to be paired with interventions reliably capable of changing bedside management.¹

“Pharmacists play a crucial role in providing care to patients with AKI, as well as those at risk for AKI,” Koyner explained. “They can perform medication reconciliation, medication regimen reviews, and nephrotoxin stewardship for hospitalized patients.”

Koyner noted that, based on the study results, these interventions “do not always happen in patients with established AKI, let alone those at high risk for AKI.” He highlighted the importance of limiting the incidence and severity of AKI through these interventions, in addition to reducing the possibility of secondary episodes by avoiding exposure to new insults.

Koyner called for future research into which specific recommendations, such as nephrotoxin avoidance or hemodynamic monitoring, are most critical to prioritize and how adherence to early, pre-AKI recommendations might be improved across care teams, including pharmacists involved in medication dosing and stopping decisions.¹

“Future studies are needed to determine which medication changes/adjustments/avoidance lead to improved outcomes in patients at risk for AKI without causing other harms,” Koyner concluded.

REFERENCES

1. Churpek MM, Fatima A, Anjorin O, et al. Early nephrology consultation and acute kidney injury in hospitalized patients: a randomized clinical trial. JAMA Netw Open. 2026;9(7):e2622554. doi:10.1001/jamanetworkopen.2026.22554

2. Aklilu AM, Menez S, Baker ML, et al. Early, individualized recommendations for hospitalized patients with acute kidney injury: a randomized clinical trial. JAMA. 2024;332(24):2081-2090. doi:10.1001/jama.2024.22718

3. Selby NM, Casula A, Lamming L, et al. An organizational-level program of intervention for AKI: a pragmatic stepped wedge cluster randomized trial. J Am Soc Nephrol. 2019;30(3):505-515. doi:10.1681/ASN.2018090886

4. Hodgson LE, Roderick PJ, Venn RM, Yao GL, Dimitrov BD, Forni LG. The ICE-AKI study: impact analysis of a clinical prediction rule and electronic AKI alert in general medical patients. PLoS One. 2018;13(8):e0200584. doi:10.1371/journal.pone.0200584

5. Zarbock A, Ostermann M, Forni L, et al. A preventive care strategy to reduce moderate or severe acute kidney injury after major surgery (BigpAK-2): a multinational, randomised clinical trial. Lancet. 2025;406(10521):2782-2791. doi:10.1016/S0140-6736(25)01717-9