A comparison of RIFLE, AKIN, KDIGO, and Cys-C criteria for the definition of acute kidney injury in critically ill patients.

Abstract

AKI is a major clinical problem and predictor of prognosis in critically ill patients. The aim of our study was to determine whether the new Cys-C criteria for identification and prognosis of AKI were superior to the RIFLE, AKIN, and KDIGO criteria. In the retrospective and multicenter study, the incidence of AKI was identified by the four criteria. Receiver operating characteristic (ROC) curve was applied to compare the predictive ability for 28-day mortality, and logistic regression analysis was used for the calculation of odds ratios and 95 % confidence intervals. In the 1036 patients enrolled, the incidences of AKI were 26.4, 34.1, 37.8, and 36.1 %, respectively, under the four criteria. Patients with AKI had higher mortality and longer length of stay than those without in all definitions. Concordance in AKI diagnosis between Cys-C and KDIGO criteria was 95.9 %, higher than AKIN and RIFLE criteria (p < 0.0001). The area under ROC curves was 0.7023 for Cys-C criteria, which was a significantly greater discrimination (p < 0.05). KDIGO criteria identified significantly more AKI and AKI patients had significantly higher 28-day mortality than patients without AKI. The Cys-C criteria were more predictive for short-term outcomes than other three criteria among critically ill patients.