Objective: To examine the epidemiology of acute renal failure (ARF) and to identify predictors of mortality in patients treated by continuous venovenous haemodiafiltration (CVVHDF).
Design: Uncontrolled observational study.
Setting: One intensive care unit (ICU) at a surgical and trauma centre.
Patients: A consecutive sample of 3591 ICU treatments.
Measurements and results: Demographic data, indications for ICU admission, severity scores and organ system failure at the beginning of CVVHDF were set against the occurrence of ARF and ICU mortality. 154 (4.3% of ICU patients and 0.6% of the hospital population) developed ARF and were treated with CVVHDF. Higher American Society of Anaesthesiologists (ASA) status and higher Apache II score were associated with ICU incidence of ARF. However, these criteria were not able to predict outcome in ARF. A simplified predictive model was derived using multivariate logistic regression modelling. The mortality rates were 12% with one failing organ system (OSF), 38% with two OSF, 72% with three OSF, 90% with four OSF and 100% with five OSF. The adjusted odds ratio (OR) of death was 7.7 for cardiovascular failure, 6.3 for hepatic failure, 3.6 for respiratory failure, 3.0 for neurologic failure, 5.3 for massive transfusion and 3.7 for age of 60 years or more.
Conclusion: General measures of severity are not useful in predicting the outcome of ARF. Only the nature and number of dysfunctioning organ systems and massive transfusion at the beginning of CVVHDF and the age of the patients gave a reliable prognosis in this group of patients.