Development and Validation of Elective and Nonelective Risk Prediction Models for In-Hospital Mortality in Proximal Aortic Surgery Using the National Institute for Cardiovascular Outcomes Research (NICOR) Database.

Background. To facilitate patient choice and the risk adjustment of consultant outcomes in aortic operations, reliable predictive tools are required. Our objective was to develop a risk prediction model for in-hospital mortality after operation on the proximal aorta. Methods. Data for 8641 consecutive UK patients undergoing proximal aortic operation from the National Institute for Cardiovascular Outcomes Research database from April 2007 to March 2013 were analyzed. Multivariable logistic regression was used to identify independent predictors of in-hospital mortality. Model calibration and discrimination were assessed. Results. In-hospital mortality was 4.6% in elective operations and 16.5% in nonelective operations. In the elective model, previous cardiac operation (adjusted odds ratio [OR] 4.1, 95% confidence interval [CI]: 3.0 to 4.7) and ejection fraction greater than 30% (adjusted OR 2.3, 95% CI: 1.7 to 3.1) were the strongest predictors of mortality (p < 0.001). The area under the receiver operating characteristic (AUROC) curve was 0.805 (95% CI: 0.802 to 0.807) with a bias-corrected value of 0.795. Model calibration was acceptable (p = 0.427) on the basis of the Hosmer-Lemeshow goodness-of-fit test. In the nonelective model, salvage operations (adjusted OR 9.9, 95% CI: 6.5 to 15.2) and previous cardiac operation (adjusted OF 3.9, 95% CI: 3.0 to 5.0) were the strongest predictors of mortality (p < 0.001). The AUROC curve was 0.761 (95% CI: 0.761 to 0.765) with a bias-corrected value of 0.756, and model calibration was also found to be acceptable (p = 0.616). Conclusions. We propose the use of these risk models to improve patient choice and to enhance patients' awareness of risks and risk-adjust aortic operation outcomes for case-mix. (C) 2016 by The Society of Thoracic Surgeons