The new EuroSCORE II does not improve prediction of mortality in high-risk patients undergoing cardiac surgery: a collaborative analysis of two European centres

Prediction of operative risk in adult patients undergoing cardiac surgery remains a challenge, particularly in high-risk patients. In Europe, the EuroSCORE is the most commonly used risk-prediction model, but is no longer accurately calibrated to be used in contemporary practice. The new EuroSCORE II was recently published in an attempt to improve risk prediction. We sought to assess the predictive value of EuroSCORE II compared with the original EuroSCOREs in high-risk patients. Patients who underwent surgery between 1 April 2006 and 31 March 2011 with a preoperative logistic EuroSCORE >= 10 were identified from prospective cardiac surgical databases at two European institutions. Additional variables included in EuroSCORE II, but not in the original EuroSCORE, were retrospectively collected through patient chart review. The C-statistic to predict in-hospital mortality was calculated for the additive EuroSCORE, logistic EuroSCORE and EuroSCORE II models. The Hosmer-Lemeshow test was used to assess model calibration by comparing observed and expected mortality in a number of risk strata. The fit of EuroSCORE II was compared with the original EuroSCOREs using Akaike's Information Criterion (AIC). A total of 933 patients were identified; the median additive EuroSCORE was 10 (interquartile range [IQR] 9-11), median logistic EuroSCORE 15.3 (IQR 12.0-24.1) and median EuroSCORE II 9.3 (5.8-15.6). There were 90 (9.7%) in-hospital deaths. None of the EuroSCORE models performed well with a C-statistic of 0.67 for the additive EuroSCORE and EuroSCORE II, and 0.66 for the logistic EuroSCORE. Model calibration was poor for the EuroSCORE II (chi-square 16.5; P = 0.035). Both the additive EuroSCORE and logistic EuroSCORE had a numerically better model fit, the additive EuroSCORE statistically significantly so (difference in AIC was -5.66; P = 0.017). The new EuroSCORE II does not improve risk prediction in high-risk patients undergoing adult cardiac surgery when compared with original additive and logistic EuroSCOREs. The key problem of risk stratification in high-risk patients has not been addressed by this new model. Future iterations of the score should explore more advanced statistical methods and focus on developing procedure-specific algorithms. Moreover, models that predict complications in addition to mortality may prove to be of increasing value.