Pairwise AUC comparisons revealed that Ranson's, MMS, BISAP, and SOFA had higher accuracy than SIRS, Ranson's and MMS had higher accuracy than APACHE-II in predicting SAP; Ranson's had the same accuracy with BISAP, but higher than other four criteria in predicting PNec; APACHE-II had higher accuracy than SIRS in predicting IPN.