ObjectiveTo compare prognostic scoring systems in a retrospective series of patients with severe acute pancreatitis admitted to a surgical intensive care unit (ICU).MethodBetween January 1992 and December 1996, the charts of all patients with a discharge code of acute pancreatitis were reviewed. There were 273 charts reviewed. Of these, 12 were admitted to the surgical ICU with a diagnosis of severe acute pancreatitis. A preliminary analysis of the data considers descriptive summary statistics, such as the mean and the range. The Spearman's rank-correlation test was computed to assess concordance between the following: a) length of stay and Ranson criteria; b) length of stay and Acute Physiology and Chronic Health Evaluation (APACHE) III score; and c) length of stay and modified Glasgow Coma score. Also, an unpaired t-test was used to obtain concordance between the following: a) death and Ranson; b) death and APACHE III; and c) death and modified Glasgow Coma score.Results30 at 96 hrs (mean 71 +/- 16 [SD]; p < .0) subsequently died. These two patients were excluded from the Spearman's rank-correlation tests. The mean length of stay in our sample was 61.8 (range, 7-201) days. The mean Ransom criteria was 4.3 (range, 1-9). The mean 96-hr APACHE III score was 33.3 (range, 0-83). The Spearman's rank-correlation between length of stay and Ranson criteria was 0.68, with a corresponding p value of .03. Similar results were observed for the length of stay and APACHE III at 96 hrs (correlation, 0.77; p = .0098) and the length of stay and the modified (correlation, 0.78; p = .007). These data reveal that the magnitude of correlation between the length of stay and the 96-hr APACHE III and modified Imrie is larger than that between length of stay and Ranson criteria.Conclusions60 days. These patients had major pancreatic complications that included pancreatic necrosis, pancreatic abscess, pseudocyst, hemorrhagic pancreatitis, and pancreatic ascites. (Crit Care Med 1999; 27:901-907)