Prospective validation of the EuroSCOREII risk model in asingle Dutch cardiac surgery centre

ObjectiveThe EuroSCOREI was one of the most frequently used pre-operative risk models in cardiac surgery. In 2011 it was replaced by its successor the EuroSCOREII. This study aims to validate the EuroSCOREII and to compare its performance with the EuroSCOREI in aDutch hospital.MethodsThe EuroSCOREII was prospectively validated in 2,296 consecutive cardiac surgery patients between 1April 2012 and 1January 2014. Receiver operating characteristic curves on in-hospital mortality were plotted for EuroSCOREI and EuroSCOREII, and the area under the curve was calculated to assess discriminative power. Calibration was assessed by comparing observed versus expected mortality. Additionally, analyses were performed in which we stratified for type of surgery and for elective versus emergency surgery.ResultsThe observed mortality was 2.4% (55patients). The discriminative power of the EuroSCOREII surpassed that of the EuroSCOREI (area under the curve EuroSCOREII 0.871, 95% confidence interval (CI) 0.832-0.911; area under the curve additive EuroSCOREI 0.840, CI 0.798-0.882; area under the curve logistic EuroSCOREI 0.761, CI 0.695-0.828). Both the additive and the logistic EuroSCOREI overestimated mortality (predictive mortality additive EuroSCOREI median 5.0%, inter-quartile range 3.0-8.0%; logistic EuroSCOREI 10.7%, inter-quartile range5.8-13.9), while the EuroSCOREII underestimated mortality (median1.6%, inter-quartile range1.0-3.5). In most stratified analyses the EuroSCOREII performed better.ConclusionOur results show that the EuroSCOREII produces avalid risk prediction and outperforms the EuroSCOREI in elective cardiac surgery patients.