Objective To compare the predictive effectiveness of five severity scoring systems on the risk assessment of death in patients readmitted to ICU after cardiac surgery. Methods A total of 184 cardiac surgery patients discharged from cardiac surgery ICU and then readmitted to ICU were selected, and the Acute Physiology and Chronic Health Evaluation (APACHE Ⅱ), European System for Cardiac Operative Risk Evaluation (EuroSCORE), Early Warning Score (EWS), Modified Early Warning Score (MEWS) and the Early Warning Score-Lactate (EWS-L) were used to assess their risk of death.Hosmer-Lemeshow goodness of fit test was used to determine the calibration ability of each tool, and the area under ROC curve was utilized to determine their ability distinguishing death or not. Results Totally 30 patients died during hospitalization. EuroSCORE，APACHEⅡ，EWS and EWS-L had good calibration in predicting the number of death because the actual number of deaths was included in their 95% CI of the predicted deaths. According to the result of Hosmer-Lemeshow test, there were no statistical differences between the predicted and actual number of deaths (P＞0.05 for all), and the discrimination ability of EWS-L was the best (AUC=0.794), when the value of EWS-L was 6.45, the Yuden index was 0.534, which achieved the best predictive performance. Conclusion It is recommended to use EWS-L assessing the risk of death for patients readmitted to ICU after cardiac surgery. And the predictive performance of existing assessment tools can be further improved through the studies of risk factor analysis for death.