UCSD-Mayo AKI Score Leads Emergency ICU Validation

Key Takeaways

• Among patients in one Chinese emergency ICU, in-hospital AKI occurred in 23.6%.
• UCSD-Mayo had the highest observed discrimination among four scores for each of the three evaluated AKI outcomes.
• UCSD-Mayo did not differ significantly from AKI-RiSc in the comparison of seven-day AKI discrimination.

Acute kidney injury (AKI) risk scores have uncertain comparative performance among critically ill patients in an emergency intensive care unit (ICU). An ICU-derived score may distinguish AKI risk differently from scores developed in broader acute-care populations across the outcomes investigators evaluated.

Investigators conducted a single-center retrospective validation study of four acute kidney injury risk scores in the same 1,723 patients admitted from 2014–2021 to one emergency ICU in China. They assessed UCSD-Mayo, AKI-RiSc, Bell, and APS as prediction scores, not treatment assignments. The evaluated outcomes were in-hospital AKI, AKI within seven days, and stage 2–3 AKI. Discrimination was assessed using the area under the receiver operating characteristic curve (AUROC) and paired DeLong tests; calibration was examined with plots and cross-validated local recalibration. Integrated discrimination improvement (IDI), categorical net reclassification improvement (NRI), decision-curve analyses, and prespecified sensitivity analyses addressed other aspects of score performance.

UCSD-Mayo had the highest observed AUROC among the evaluated scores for each outcome: 0.713 for in-hospital AKI, 0.690 for AKI within seven days, and 0.706 for stage 2–3 AKI. The seven-day and stage 2–3 outcomes occurred less often than in-hospital AKI. The ranking describes observed discrimination in this cohort; it does not establish statistically significant superiority over every comparator.

The paired comparison for AKI within seven days gave a UCSD-Mayo minus AKI-RiSc AUROC difference of 0.029 (95% confidence interval [CI], −0.006 to 0.064; P=0.103), which was not statistically significant. Reclassification and decision-curve net benefit varied with the outcome and risk threshold, so the comparison did not favor one score uniformly across these measures. Prespecified sensitivity analyses did not materially change the overall pattern reported by the investigators.

Investigators characterized the performance gains as modest and not uniform across outcomes, measures, or decision thresholds. The retrospective comparison at one emergency ICU in China describes how existing scores performed in that cohort, rather than testing whether care guided by a score changes patient outcomes.

According to the authors, the risk scores may support AKI risk awareness and monitoring, but they do not serve as stand-alone treatment rules.

Clinician Questions

What cutoff was used for net reclassification improvement in the Chinese emergency ICU AKI score comparison?

Investigators assessed categorical net reclassification improvement for AKI risk scores in the Chinese emergency ICU using the observed event rate as the cutoff.

Was a time window specified for stage 2–3 AKI in the Chinese emergency ICU risk-score validation?

Investigators evaluated stage 2–3 acute kidney injury in patients in a Chinese emergency intensive care unit. Stage 2–3 AKI was assessed during the hospital stay, using KDIGO creatinine criteria, as a separate outcome from AKI within 7 days.

Recommended Reading

• For more on inpatient AKI risk modeling: AKI Risk Model in Older Heart Failure Inpatients