OBJECTIVE To assess the accuracy of the APACHE II score, the Mannheim Peritonitis Index (MPI), and the Peritonitis Index Altona (PIA) II in the prediction of outcome of patients with peritonitis. METHODS Prospective, multicentre study. METHODS 12 Departments of Surgery in Europe. METHODS 271 Patients with peritonitis confirmed at laparatomy. METHODS Computation of the three scores on one set of data for each patient. METHODS The ability to predict death or survival within 30 days of operation with each of the three scores. The prediction were evaluated according to the following criteria: discriminatory ability (areas under the receiver-operator characteristic (ROC) curves relating sensitivity to specificity); sharpness (level of confidence that was associated with a prediction); and reliability (agreement between predicted and observed mortality within equidistant intervals on the scale). RESULTS APACHE II was superior to both the MPI and PIA II in its discriminatory ability and reliability, but the MPI and PIA II made more "sharp" predictions. CONCLUSIONS None of the three scores is of any use for predicting the outcome for individual patients. APACHE II is the current standard for assessing the severity of peritonitis.