Previously published investigations have suggested that the Injury Severity Score (ISS) can be replaced by the New Injury Severity Score (NISS) which takes into account the three most severe injuries regardless of body region. This study was conducted to evaluate whether NISS can also give a better mortality prediction than ISS in a different setting. The objective of this study was to compare the accuracy between ISS and NISS in predicting mortality of trauma patients. The study population consisted of trauma patients admitted to a 650-bed university hospital in Thailand from June 1996 to May 1999. Data of patients admitted to the hospital were prospectively collected to identify the injuries and outcomes of treatment. Each injury was reviewed and coded according to the Abbreviated Injury Scale 1990 revision (AIS-90) and computed for ISS and NISS. Areas under receiver operating characteristic (ROC) curves were employed to compare the abilities of the ISS and the NISS in predicting patients' mortality. The results of the study showed that 2,044 trauma patients were admitted to the hospital during the three-year study period and 114 patients died in the hospital. The median scores of the ISS and the NISS of the survivors were 4 and 8 respectively. The median scores of the ISS and the NISS of the non-survivors were 25 and 38 respectively. The area under ROC curve derived from the ISS (0.966; 95%C.I = 0.965 - 0.967) was significantly (p < 0.05) less than the NISS (0.974; 95%C.I = 0.973 - 0.975). We concluded, under our setting, that NISS also performed the mortality prediction in trauma patients better than ISS and confirmed the results of the previous studies.