Background: Several proposed definitions for acute renal failure (ARF) exist, but little is known of their significance in clinical practice. We evaluated the ability to predict hospital mortality in 2 ARF-specific severity-of-illness scoring methods, the Risk, Injury, Failure, Loss, End-Stage Renal Disease (RIFLE) score and the score presented by Bellomo et al in 2001.
Methods: The study included 668 consecutive patients with 694 treatment episodes treated in 2 intensive care units (ICUs) in a university hospital within 11 months. ARF prevalence was classified according to the RIFLE and Bellomo scores. As references, we evaluated 2 general severity-of-illness scoring systems, the admission Acute Physiology and Chronic Health Evaluation II (APACHE II) and Sequential Organ Failure Assessment (SOFA) scores.
Results: Admission SOFA scores and maximum RIFLE scores for the first 3 days in the ICU were independent predictors of hospital mortality by means of forward conditional logistic regression. In receiver operating characteristic analysis, SOFA and APACHE II scores performed better than ARF-specific scores, and discriminative powers for hospital mortality were only moderate for the RIFLE and Bellomo scores: areas under the curve were 0.653 (95% confidence interval, 0.588 to 0.719) and 0.587 (95% confidence interval, 0.514 to 0.660), respectively.
Conclusion: Neither of the ARF-specific scoring methods presented good discriminative power regarding hospital mortality. However, maximum RIFLE score for the first 3 days in the ICU was found to be an independent predictor of hospital mortality, along with admission SOFA score.