RESUMEN
INTRODUCTION: Acute kidney injury (AKI) is an important lifethreatening complication in patients hospitalized in intensive care units (ICU). This study was conducted to determine the incidence of AKI in the medical intensive care unit of a tertiary university hospital and to compare the predictive performance of three different AKI criteria (RIFLE, AKIN, and KDIGO) for in-hospital mortality. METHODS: The data of all consecutive patients were evaluated from their hospitalization to ICU until discharge or death, retrospectively. Patients with end-stage renal disease, history of kidney transplantation, those who stayed in the ICU for less than 72 hours, who underwent dialysis before admission to the ICU, and those with incomplete medical records were excluded. AKI was defined using serum creatinine criteria of RIFLE, AKIN, and KDIGO. RESULTS: 303 patients were included in this study. According to RIFLE, AKIN, and KDIGO criteria the incidence of AKI were 47.9 %, 44.6%, and 50.2%; respectively. In-hospital mortality rates were higher in AKI patients (P < .05 according to all three criteria). Regression analysis revealed that AKI was a predictor of in-hospital mortality (P < .05, for all). The ROC analyses showed that each of these criteria had similar abilities to predict in-hospital mortality (area under (Au) ROC for RIFLE = 0.76, AuROC for AKIN = 0.72, and AuROC for KDIGO = 0.76). CONCLUSION: The incidence of AKI was higher with KDIGO criteria. In-hospital mortality rates were higher in patients with AKI. Each criteria had similar abilities to predict in-hospital mortality.