In this study, we compare the Trauma and Injury Severity Score (TRISS) and A Severity Characterization of Trauma (ASCOT) models by using NCKUH trauma registry to assess the performance of correct prediction in terms of sensitivity, specificity and misclassification rate. The database has accumulated to 5,672 cases, NCKUH 2,490; Chi-Mei 3,182 respectively. Blunt trauma mechanism was composed of 4, 892 (86.2%) while 552 (9.7%) were pertinent to penetrating. The male/female ratio is 2.4:1. Traffic accident is the major cause of injury (3, 472-(61.2%)), followed by work injury (723-(12.7%)); fall (702-(12.4%)) and burn injuries (160-(2.8%)). The category of traffic accident is comprised of motorcycle-related, (1,257-(69.14%)), followed by automobile-related was (301-(16.56%)) and bicycle injuries (123-(6.8%)). The category of working injury comprised by machine crushed cases (332-(45.92%)) followed by cutting (148-(20.47%)) and impacts (69-(9.5%)). The overall mortality rate in our registry was 8.3%. ASCOT and TRISS were compared using sensitivity, specificity and misclassification rates. Each method had disadvantages in predicting outcomes of particular subgroups of patients. ASCOT tends to underestimate the probability of survival among patients with head/spinal injuries; while TRISS had a similar effect on multiple trauma victims. In conclusion, ASCOT is superior to TRISS in correctly predicting severe head trauma cases. However, both methods have their limitations in terms of accurate prediction. It is our hope to develop a mixed, revised model to better predict patients survival probability. Therefore, it is feasible to adopt ASCOT methodology in prediction of trauma patients in Taiwan. Expanded database and better methodology need to be developed in further study.