Performance of Three Prognostic Models in Patients with Cancer in Need of Intensive Care in a Medical Center in China

OBJECTIVE: The aim of this study was to evaluate the performance of Acute Physiology and Chronic Health Evaluation II (APACHE II), Simplified Acute Physiology Score 3 (SAPS 3), and Acute Physiology and Chronic Health Evaluation IV (APACHE IV) in patients with cancer admitted to intensive care unit (ICU) in a single medical center in China. MATERIALS AND METHODS: This is a retrospective observational cohort study including nine hundred and eighty one consecutive patients over a 2-year period. RESULTS: The hospital mortality rate was 4.5%. When all 981 patients were evaluated, the area under the receiver operating characteristic curve (AUROC, 95% Confidential Intervals) of the three models in predicting hospital mortality were 0.948 (0.914–0.982), 0.863 (0.804–0.923), and 0.873 (0.813–0.934) for SAPS 3, APACHE II and APACHE IV respectively. The p values of Hosmer-Lemeshow statistics for the models were 0.759, 0.900 and 0.878 for SAPS 3, APACHE II and APACHE IV respectively. However, SAPS 3 and APACHE IV underestimated the in-hospital mortality with standardized mortality ratio (SMR) of 1.5 and 1.17 respectively, while APACHE II overestimated the in-hospital mortality with SMR of 0.72. Further analysis showed that discrimination power was better with SAPS 3 than with APACHE II and APACHE IV whether for emergency surgical and medical patients (AUROC of 0.912 vs 0.866 and 0.857) or for scheduled surgical patients (AUROC of 0.945 vs 0.834 and 0.851). Calibration was good for all models (all p > 0.05) whether for scheduled surgical patients or emergency surgical and medical patients. However, in terms of SMR, SAPS 3 was both accurate in predicting the in-hospital mortality for emergency surgical and medical patients and for scheduled surgical patients, while APACHE IV and APACHE II were not. CONCLUSION: In this cohort, we found that APACHE II, APACHE IV and SAPS 3 models had good discrimination and calibration ability in predicting in-hospital mortality of critically ill patients with cancer in need of intensive care. Of these three severity scores, SAPS 3 was superior to APACHE II and APACHE IV, whether in terms of discrimination and calibration power, or standardized mortality ratios.