Prognosis of acute renal failure requiring dialysis in an intensive care unit: a multivariate analysis

The persistently high mortality rate in acute renal failure (ARF) patients managed in intensive care units (ICUs) remains unexplained. The few available studies of prognostic factors have produced conflicting results. Objective: to identify prognostic factors in ICU patients with ARF requiring dialysis. Methods. patients with a negative history for chronic renal failure who developed ARF within 48 hours of an ICU admission between January 1, 1993, and June 30, 1995, were included retrospectively. The values of 27 variables (two demographic, 15 clinical, two therapeutic, and eight laboratory test variables) and the number of organ failures prior to the first dialysis session were subjected to univariate analysis then to multivariate multiple logistic regression. Performance of the multivariate model was evaluated in the same population using the Hosmer-Lemeshow goodness-of-fit test, and the discriminating value of the model was assessed based on the area under the Receiver Operating Characteristic (ROC) curve as described by Hanley, Results: 104 patients (82 male and 32 female) with a mean age of 63.6 +/- 14.5 years were included. The IGSI, IGSII, and APACHE II severity scores were 13.9 +/- 4.6, 48.2 +/- 14.3, and 21.6 +/- 5.9, respectively, Mean number of organ failures was 2.2 +/- 1.3. Artificial ventilation was used in 43.8% of patients, and 48.8% had sepsis, 22.8% shock, and 35% oliguria. In hospital mortality was 44.7% (95% CI, 4.01-49.4). Of the 18 variables significant in the univariate analysis, only five were significant in the multivariate analysis, namely a history of stage III-IV heart failure, a history of chronic obstructive lung disease, sepsis, oliguria (<0.5 L/d for 3 days), and the Glasgow score. The Hosmer-Lemeshow goodness-of-fit test showed excellent performance of the multivariate model (predicted mortality, 62.9% versus observed mortality, 63%; H score, 11.6; P>0.1). The area under the ROC curve was 0.93, indicating good discriminating power of the model. Conclusion: In this study, simple parameters available on admission provided valuable prognostic orientation in ARF patients requiring dialysis in an ICU. These results require confirmation by a prospective study in another population.