Prediction of outcome in acute renal failure by discriminant analysis of clinical variables

Abstract

To define factors of prognostic importance in patients with acute renal failure, we studied 23 clinical variables (including preexisting conditions, the acute clinical setting, and complications) by univariate and multivariate analysis in 148 patients. Ten variables contributed to discrimination in the overall analysis. Using a function derived by stepwise discriminant analysis, death could be predicted with a positive predictive value of 100% and a sensitivity of 58%. The performance of this discriminant score was validated in a subsequent group of 113 patients with a positive predictive value of 100% and a sensitivity of 26%. Thus, a fatal outcome of acute renal failure can be accurately predicted in a considerable proportion of patients by a weighted evaluation of clinical variables. The discriminant score may be useful in comparing the severity of illness among series of patients with acute renal failure.