BACKGROUND The aim of this study was to test whether early and intensive use of continuous venovenous hemofiltration (CVVH) achieved a better than predicted outcome in patients with severe acute renal failure undergoing cardiac operations, and whether a simple and yet accurate model could be developed to predict their outcome before starting CVVH. METHODS Medical record analysis with collection of demographic, clinical, and outcome information was used. RESULTS Sixty-five consecutive patients were treated with early and intensive CVVH (mean operation to CVVH time, 2.38 days; pump-controlled ultrafiltration rate, 2 L/h) after coronary artery bypass grafting (56.9%), single valve procedure (16.9%), or combined operations (26.2%). In 32.3% of patients, intraaortic balloon counterpulsation was required and 20% of patients were emergencies. Sustained hypotension despite inotropic and vasopressor support occurred in 40% of patients and prolonged mechanical ventilation in 58.5%. Using an outcome prediction score specific for acute renal failure, the predicted risk of death was 66%. Actual mortality was 40% (p = 0.003). Using multivariate logistic regression analysis and neural network analysis, patient outcome could be predicted with good levels of accuracy (receiver operating characteristic 0.89 and 0.9, respectively). CONCLUSIONS Early and aggressive CVVH is associated with better than predicted survival in severe acute renal failure after cardiac operations. Using readily available clinical data, the outcome of such patients can be predicted before the implementation of CVVH.