Validation of 5 prediction models for acute kidney injury and its outcome after cardiac surgery procedures in Chinese patients

ObjectiveTo assess the clinical usefulness and value of the 5 models for the prediction of acute kidney injury (AKI), severe AKI which renal replacement treatment was needed (RRT-AKI) and death after cardiac surgery procedures in Chinese patients. Methods One thousand and sixty - seven patients who underwent cardiac surgery procedures in the department of cardiac surgery in the Zhongshan Hospital, Fudan University between May 2010 and January 2011 were involved in this research. The predicting value for AKI (AKICS), RRT-AKI (Cleveland, SRI and Mehta score) and death (EURO score) after cardiac surgery procedures was evaluated by Hosmer-Lemeshow goodness-of-fit test for the calibration and area under receiver operation characteristic curve (AUROC) for the discrimination. ResultsThe incidence of AKI was 20.34%(217/1067), and 63.13% of their renal function recovered completely. The incidence of RRT-AKI was 3.56%(38/1067) and the mortality of AKI and RRT - AKI was 9.68%(21/217) and 44.73%(17/38) respectively. The total mortality was 3.28%（35/1067）. The discrimination and calibration for the prediction of AKI of AKICS were low. For the prediction of RRT-AKI, the discrimination and calibration of Cleveland score were high enough, but the predicated value was lower than the real value (1.70% vs 3.86%). The discrimination of Mehta score and the calibration of SRI were low. The discrimination and calibration for the prediction of death of EURO score was low. ConclusionAccording to the 2012 KDIGO AKI definition, none of the 5 models above is good at predicting AKI after cardiac surgery procedures. Cleveland score has been validated to have a proper impact on predicting RRT-AKI after cardiac surgery procedures, but the predicting value is still in doubt. EURO score has been validated to have an inaccurate predicting value for death after cardiac surgery procedures.