Objective: The present study aimed to examine the predictive performance of the logistic European System for Cardiac Operative Risk Evaluation (EuroSCORE) in a large cohort of patients undergoing cardiac surgery from 1999 through 2010 because methodologic shortcomings have hampered many previous studies questioning its predictive performance.
Design: Population-based prospectively registered data.
Setting: The Western Denmark Heart Registry, a multi-institutional registry.
Participants: Twenty-one thousand six hundred sixty-four patients.
Interventions: On-pump cardiac surgery.
Measurements and main results: The predictive ability of the logistic EuroSCORE was assessed using the area under the curve (AUC) for the discrimination test, the Hosmer-Lemeshow (HL) calibration test, and the mean estimated-to-observed mortality ratio (E/O). The overall AUC was 0.79 (95% confidence interval [CI] 0.77-0.81; HL test, p < 0.01; E/O 1.9). For coronary artery bypass grafting, the AUC was 0.78 (95% CI 0.75-0.81; HL test, p < 0.01; E/O 2.3). For coronary artery bypass grafting plus valve replacement, the AUC was 0.69 (95% CI 0.65-0.73; HL test, p = 0.02; E/O 1.5). For aortic valve replacement, the AUC was 0.76 (95% CI 0.72-0.80; HL test, p < 0.01; E/O 2.5). The overall and procedural specific E/O ratios tended to increase from 1999 to 2010. Mortality was overestimated across all levels of estimated risk, and in low-to-medium-risk patients, this overestimation increased most notably with time.
Conclusions: The EuroSCORE provides moderate-to-good discrimination and poor calibration. Despite substantial changes in risk factors during the study period, the EuroSCORE consistently overestimated 30-day mortality independent of the preoperative risk level and surgical procedure performed, indicating improved quality of surgery and patient care.
Copyright © 2012 Elsevier Inc. All rights reserved.