Abstract:Objective To validate the accuracy and feasibility of acute physiology and chronic health evaluation (APACHE) Ⅳ model in critically ill elderly patients. Methods A total of 421 elderly critically ill patients admitted in our hospital from July 2011 to July 2015 were prospectively recruited in this study. Their APACHE Ⅳ scores were calculated within 24 h in ICU. Discrimination and calibration were respectively assessed by using area under the receiver operating characteristic curve (ROC AUC) and Hosmer-Lemeshow (HL) test. Standardized mortality ratio (SMR) was reported to quantify the mortality variation between observation and prediction and the brier score (BS) to represent the overall performance. Results The mortality rate of the 421 subjects was 33.3%(140/421). The discrimination of the APACHE Ⅳ model was good to excellent (AUC=0.82,5%CI:0.78-0.86), but the calibration was inadequacy according to the results of HL test (Chi square=20.49, P=0.009). The BS showed good accuracy of APACHE Ⅳ model (BS=0.18), but it may underestimate the in-hospital mortality (SMR=1.50,5%CI:1.30-1.70) in elderly critically ill patients. ConclusionThe APACHE Ⅳ model shows good discrimination in critically ill elderly patients, but has poor calibration and underestimated in-hospital mortality. Therefore, before we use the risk prediction models for geriatric patients, we should validate it and employ customized models to fit the population to be predicted.