External validation of the acute physiology and chronic health evaluation II in Korean intensive care units

Purpose: This study was designed to validate the usefulness of the Acute Physiology and Chronic Health Evaluation (APACHE) II for predicting hospital mortality of critically ill Korean patients. Materials and Methods: We analyzed data on 826 patients who had been admitted to nine intensive care units and were included in the Fever and Antipyretics in Critical Illness Evaluation study cohort. Results: Among the patients enrolled, 62% (512/826) were medical and 38% (314/826) were surgical patients. The median APACHE II score was 17 (11 to 23 interquartile range), and the hospital mortality rate was 19.5%. Age, underlying diseases, medical patients, mechanical ventilation, and renal replacement therapy were independently associated with hospital mortality. The calibration of APACHE II was poor (H=57.54, p<0.0001; C=55.99, p<0.0001), and the discrimination was modest [area under the receiver operating characteristic (aROC)=0.729]. Calibration was poor for both medical and surgical patients (H=63.56, p<0.0001; C=73.83, p<0.0001, and H=33.92, p<0.0001; C=33.34, p=0.0001, respectively), while discrimination was poor for medical patients (aROC=0.651) and modest for surgical patients (aROC=0.704). At the predicted risk of 50%, APACHE II had a sensitivity of 36.6% and a specificity of 87.4% for hospital mortality. Conclusion: For Koreans, the APACHE II exhibits poor calibration and modest discrimination for hospital mortality. Therefore, a new model is needed to accurately predict mortality in critically ill Korean patients.