BACKGROUNDSeveral risk stratification models have been developed for cardiac surgery. This study aimed to evaluate the accuracy of four existing risk stratification models, the Fuwai System for Cardiac Operative Risk Evaluation (FuwaiSCORE), the Society of Thoracic Surgeons 2008 cardiac surgery risk model for isolated valve surgery (the STS model), the European System for Cardiac Operative Risk Evaluation (EuroSCORE) and the initial Parsonnet's score (the Parsonnet model) in predicting prolonged intensive care unit (ICU) stay in Chinese patients undergoing heart valve surgery.

METHODSData were collected retrospectively from records of 1333 consecutive patients who received heart valve surgery in a single center between November 2006 and December 2007. Prolonged ICU stay was defined as not less than 124 hours. Calibration was assessed using the Hosmer-Lemeshow (H-L) goodness of fit test. Discrimination was assessed using the receiver-operating-characteristic (ROC) curve area.

RESULTSThe FuwaiSCORE showed good calibration and discrimination compared with other risk models. According to the H-L statistics, the value of the FuwaiSCORE was 12.82, P > 0.1. The area under ROC curve of the FuwaiSCORE was 0.81 (95%CI 0.78 - 0.84).

CONCLUSIONSOur study suggests that the FuwaiSCORE is superior to the other three risk models in predicting prolonged length of ICU stay in Chinese patients with heart valve surgery. Having fewer variables, the system is much easier for bedside use than other systems.