Validation of a modified EuroSCORE risk stratification model for cardiac surgery: the Swedish experience.

Objective: The European System for Cardiac Operative Risk Evaluation (EuroSCORE) is used to identify patients at high risk for aortic valve replacement (AVR) in whom alternative procedures, such as trans-catheter aortic valve implantation (TAVI), may be appropriate. The aim of the present study was to calibrate and validate the EuroSCORE for different cardiac surgery procedures to improve patient selection for valve surgery. Methods: The study included 46516 patients undergoing open cardiac surgery during 2001-2007. A fivefold cross-validation technique was used to calibrate four different models. Model discrimination was determined by the area under the receiver operating characteristic (ROC) curve and model calibration by the Hosmer-Lemeshow (H-L) test. Results: The actual and predicted 30-day mortality was 3.2%. The discrimination (ROC area) of the calibrated 30-day mortality prediction models was 0.79 for coronary bypass surgery, 0.77 for mitral valve surgery (MVS), and 0.75 for miscellaneous procedures, compared with 0.78 (p=0.199), 0.74 (p=0.077), and 0.72 (p=0.001), respectively, for the original EuroSCORE. The discrimination for AVR was the same for the calibrated and the original EuroSCORE model (0.70). The H-L test gave a p-value of 0.104 for the calibrated and <0.001 for the original EuroSCORE model. Conclusions: A calibration of EuroSCORE resulted in an acceptable predictive capacity for 30-day mortality, and improved discrimination and calibration for MVS and miscellaneous procedures. However, the poor discriminatory for the AVR procedure suggests that the EuroSCORE may not be satisfying for assessing risk prior to TAVI and that more optimized risk stratification models may be needed.