Objective: To compare the value of ATN-ISI (acute tubular necrosis-individual severity index) and APACHE II (acute physiology and chronic health evaluation) score models in predicting mortality and renal outcome in patients with acute renal failure (ARF).
Methods: 422 in-hospital patients with ARF were retrospectively analysed, the value in predicting prognosis and renal outcome of these two score models was compared. Renal outcome was observed three times, i.e. >or= 30 days, >or= 45 days and >or= 60 days after the occurrence of ARF and discriminant analysis was used in the study.
Results: The scores of the two models both correlated significantly with the mortality of patients with ARF. When ATN-ISI scores were >or= 0.85, the mortality was 100% and it was the same with APACHE II when the scores were >or= 35. The areas under ROC curve of APACHE II and ATN-ISI were 0.817 +/- 0.021 and 0.880 +/- 0.018 respectively, suggesting good model discriminantion. For the correctly classificated rate of renal outcome, ATN-ISI was better than APACHE II at any time. When ATN-ISI scores were >or= 0.75, patients were dialysis dependent. When the scores were >or= 0.58 but less than 0.75, the patient's renal function did not recover but was not dialysis dependent; when the scores were < 0.58, the patients's renal function would recover. When APACHE II scores were >or= 26, patients were dialysis dependent. There was no cut-off score point to separate renal function recovery from renal dysfunction. However, when APACHE II scores were <or= 22, 19.6% of the patients had renal dysfunction and 80.4% of them had renal function recovered.
Conclusion: Both the score models are of value in predicting mortality and renal outcome in patients with ARF. ATN-ISI score model was better than APACHE II.