Abstract
OBJECTIVE: The EuroSCORE I was one of the most frequently used pre-operative risk models in cardiac surgery. In 2011 it was replaced by its successor the EuroSCORE II. This study aims to validate the EuroSCORE II and to compare its performance with the EuroSCORE I in a Dutch hospital. METHODS: The EuroSCORE II was prospectively validated in 2,296 consecutive cardiac surgery patients between 1 April 2012 and 1 January 2014. Receiver operating characteristic curves on in-hospital mortality were plotted for EuroSCORE I and EuroSCORE II, 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. RESULTS: The observed mortality was 2.4% (55 patients). The discriminative power of the EuroSCORE II surpassed that of the EuroSCORE I (area under the curve EuroSCORE II 0.871, 95% confidence interval (CI) 0.832-0.911; area under the curve additive EuroSCORE I 0.840, CI 0.798-0.882; area under the curve logistic EuroSCORE I 0.761, CI 0.695-0.828). Both the additive and the logistic EuroSCORE I overestimated mortality (predictive mortality additive EuroSCORE I median 5.0%, inter-quartile range 3.0-8.0%; logistic EuroSCORE I 10.7%, inter-quartile range 5.8-13.9), while the EuroSCORE II underestimated mortality (median 1.6%, inter-quartile range 1.0-3.5). In most stratified analyses the EuroSCORE II performed better. CONCLUSION: Our results show that the EuroSCORE II produces a valid risk prediction and outperforms the EuroSCORE I in elective cardiac surgery patients.