Does the euroSCORE equally well predict perioperative cardiac surgical risk for men and women?

Abstract

The euroSCORE is a widely applied scoring system to estimate perioperative risk in patients undergoing cardiac surgery. It acknowledges an increased risk in women but does not apply a gender-specific weighing of individual risk indicators. We separately investigated the performance of the euroSCORE for men and women. Cardiac surgical patients operated in 2002 and 2003 (n = 1930) served as training group. Patients operated in 2004 served as validation group (n = 870). Using binary regression analysis, we developed a gender-specific model on the basis of the original euroSCORE risk indicators. The individual risk indicators were assigned gender-specific correlation coefficients. Applying ROC analysis to the validation group, the original euroSCORE performed well, with an area under curve (AUC) of 0.758, for men but poor, with an AUC 0.676, for women. After gender-specific re-weighing, ROC analysis revealed considerably improved predictive values of the modified euroSCORE. AUC for women increased to 0.755, AUC for men further increased to 0.772. The original euroSCORE failed to accurately predict perioperative mortality in women. A binary regression model, assigning a gender-specific weight to each of the risk indicators of the original euroSCORE, considerably improved the predictive power of the modified euroSCORE.

Key words::

• Gender
• statistics
• risk analysis/modeling

Declaration of interest: The authors report no conflicts of interest. The authors alone are responsible for the content and writing of the paper.