• Joachim Kötting, Wolfgang Schiller, Andreas Beckmann, Elke Schäfer, Klaus Döbler, Christian Hamm, Christof Veit, and Armin Welz.
• Department of Biostatistics, BQS-Institute for Quality and Patient Safety, Düsseldorf, Germany.
• Eur J Cardiothorac Surg. 2013 May 1; 43 (5): 971-7.

ObjectivesThe aim of the study was to establish a scoring system to predict mortality in aortic valve procedures in adults [German Aortic Valve Score (German AV Score)] based upon the comprehensive data pool mandatory by law in Germany.MethodsIn 2008, 11 794 cases were documented who had either open aortic valve surgery or transcatheter aortic valve implantation (TAVI). In-hospital mortality was chosen as a binary outcome measure. Potential risk factors were identified on the basis of published scoring systems and clinical knowledge. First, each of these risk factors was tested in an univariate manner by Fisher's exact test for significant influence on mortality. Then, a multiple logistic regression model with backward and forward selection was used. Calibration was ascertained by the Hosmer-Lemeshow method. In order to define the quality of discrimination, the area under the receiver operating characteristic (ROC) curve was calculated.ResultsIn 11 147 of 11 794 cases (94.5%), a complete data set was available. In-hospital mortality was 3.7% for all patients, 3.4% in the surgical group (95% confidence interval 3.0-3.7%, n = 10 574) and 10.6% in the TAVI group (95% confidence interval 8.2-13.5%, n = 573). Based on multiple logistic regression, 15 risk factors with an influence on mortality were identified. Among them, age, body mass index and left ventricular function were categorized in three (body mass index, left ventricular dysfunction) or 6 subgroups (age). The Hosmer-Lemeshow method corroborated a valid concordance of predicted and observed mortality in 10 different risk groups. The area under the ROC curve with a value of 0.808 affirmed the quality of discrimination of the established scoring model.ConclusionsIt is well known that a predictive model works best in the setting where it was developed; therefore, the German AV Score fits well to the patient population in Germany. It was designed for fair and reliable outcome evaluation. It allows comparison of predicted and observed mortality for conventional aortic valve surgery and transcatheter aortic valve implantation in low-, moderate- and high-risk groups. Thus, it enables primarily a risk-adjusted benchmark of outcome and fosters the efforts for continuous improvement of quality in aortic valve procedures.

20 keywords...

hide…