We performed a prospective study in the 28 multidisciplinary intensive care units (ICUs) in the Rhône-Alpes area in France to investigate the role of initial versus delayed occurrence of acute renal failure (ARF) in patient outcome. ARF was defined as a serum creatinine concentration > 300 micromol/L, urine output < 500 ml/24 h (or < 180 ml/8 h), or hemodialysis requirement. Over the 1-yr study period, 1,086 patients presented with ARF on ICU admission or during the first 2 d of ICU stay (Group A; 736 patients), from Day 3 to Day 6 (Group B; 202 patients), or from Day 7 (Group C; 148 patients). The overall hospital mortality rate was 66% (61% in Group A, 71% in Group B, and 81% in Group C; p < 0.0001). Logistic regression analysis of a random sample of 510 patients showed that SAPS II score on ICU admission, number of ARF episodes, previous health status, absence of oliguria, absence of hemodialysis, and absence of ischemic acute tubular necrosis were predictive of patient survival. This model was tested and validated on the basis of the remaining patients. Thus, in this population, late ARF was not a predictive factor for patient outcome.