The definition, classification, and choice of management of acute renal failure (ARF) in the setting of the intensive care unit (ICU) remain subjects of debate. To improve our approach to ARF in the ICU setting, we retrospectively applied the new classification of ARF put forward by the Acute Dialysis Quality Initiative group, RIFLE (acronym indicating Risk of renal failure, Injury to the kidney, Failure of kidney function, Loss of kidney function, and End-stage renal failure), to evaluate its sensitivity and specificity to predict renal and patient outcomes. RIFLE classification was applied to 183 patients with ARF admitted to the ICU (2002 to 2003) at the Northern General Hospital, Sheffield, UK. Patients were divided into 4 groups according to percentage of decrease in glomerular filtration rate from baseline. The risk group included 60 patients; injury group, 56 patients; failure group, 43 patients; and control group, 24 patients. Demographic, biochemical, hematologic, clinical, and long-term health status were studied and compared in the 4 groups. An attempt was made to evaluate, by means of logistic regression analysis and receiver operator characteristic curve analysis, the predictive value of RIFLE classification for mortality in the ICU. The failure group showed the worst parameters with regard to Acute Physiology and Chronic Health Evaluation (APACHE) II score, pH, lowest and highest mean arterial pressures, and Glasgow Coma Scale (P < 0.001). Mortality rate in the ICU (1 month) was significantly greater in the failure group compared with all groups (32 of 43 patients [74.4%]; P < 0.001) and, again, 6-month mortality rate (37 of 43 patients [86%]; P < 0.001). Receiver operator characteristic curve analysis showed that Simplified Acute Physiology Score (SAPS) II was more sensitive than APACHE II score for prediction of patient death in the risk and injury groups compared with the failure and control groups (risk group: SAPS II, 0.8 +/- 0.06; P < 0.001; APACHE II, 0.63 +/- 0.07; P = 0.14; injury group: SAPS II, 0.76 +/- 0.08; P < 0.001; APACHE II, 0.72 +/- 0.07; P = 0.006). RIFLE classification can improve the ability of such older and established ICU scoring systems as APACHE II and SAPS II in predicting outcome of ICU patients with ARF.