The aim of this study was to validate and compare the acute physiology and chronic health evaluation (APACHE) II, simplified acute physiology score (SAPS) II, and SAPS 3 in a Korean intensive care unit (ICU). A retrospective analysis of the prospective ICU registry was conducted in the medical and surgical ICUs of Samsung Medical Center. Calibration and discrimination were assessed by the Hosmer-Lemeshow goodness of fit C and H test and area under the receiver operating characteristic (aROC) curve from 1,163 patients. The ICU mortality was 9% (105/1,163) and the hospital mortality was 14.9% (173/1,163). The mortalities (%) predicted by APACHE II, adjusted APACHE II (Adj-APACHE II), SAPS 3, Australagia SAPS 3 (AUS-SAPS3), and SAPS II were mean 29, 26.2, 27.3, 25.2 and 21, respectively. The calibration of all severity scores were poor (APACHE II: H=222.82, p＜0.0001, C=212.23, p＜0.0001, Adj-APACHE II: H=191.75, p＜0.0001, C=165.36, p＜0.0001, SAPS 3: H=132.84, p＜0.0001, C=131.80, p＜0.0001, AUS-SAPS 3: H=106.36, p＜0.0001, C=97.32, p＜0.0001, SAPS II: H=89.01, p＜0.0001, C=64.68, p＜0.0001). The discrimination of APACHE II was modest (aROC=0.779), and SAPS 3 and SAPS II showed good discrimination (aROC=0.876 and 0.906, respectively). In Korea, hospital mortality rates predicted using APACHE II, SAPS 3, and SAPS II exhibited poor calibration and modest to good discrimination. To better predict mortality, a new model or customization specifically for Korea is needed.