Analysis of risk factors for valve replacements in 5,128 cases from a single heart center in China

Numerous studies have developed a "severity score" or "risk index" for short-term mortality associated with coronary artery bypass grafting (CABG). Due to the different distribution of disease types, the number of valve surgeries in the US and Europe is relatively small. Thus, a risk-scoring system for valve surgeries was developed later and used less than that for the CABG surgery. We retrospectively reviewed 5128 cases of heart valve replacement, to quantitatively assess the risk factors for hospital mortality, and establish risk models for the hospital mortality of cardiac valve replacement patients.

A total of 1549 cases of aortic valve replacement, 2460 cases of mitral valve replacement, and 1119 cases of combined aortic valve and mitral valve replacement that were recorded from January 2005 to December 2009 in the cardiac surgery database at Beijing Anzhen Hospital were selected for this study. The cases were randomly assigned to a model group (n = 3657) and a validation group (n = 1471) with a ratio of 7:3. Thirty-two pre- and intra-operative clinical indicators were selected as possible influencing factors for hospital mortality. Single-factor analysis was performed to screen these factors, and then multi-factor analysis was used to determine the risk factors for hospital mortality in the three surgeries and to establish risk models.

In the multi-factor analysis, age, body surface area, etiology, cardiopulmonary bypass time, preoperative cardiothoracic ratio, cardiac functional classification, and preoperative creatinine were risk factors for aortic valve replacement. Etiology, preoperative history of heart failure, cardiopulmonary bypass time, preoperative cardiothoracic ratio, and preoperative left ventricular end systolic diameter were risk factors for mitral valve replacement. Age, body mass index, cardiopulmonary bypass time, and cardiac function classification were risk factors for combined aortic valve and mitral valve replacement. The risk models showed good predictive ability (Hosmer-Lemeshow test: P = 0.981 in the model for aortic valve replacement, P = 0.503 in the model for mitral valve replacement, and P = 0.154 in the model for combined aortic valve and mitral valve replacement). The area under the ROC curve of the validation group was 0.958 (95%CI: 0.936 - 0.975) for the aortic valve replacement model, 0.876 (95%CI: 0.805 - 0.948) for the mitral valve replacement model, and 0.845 (95%CI: 0.753 - 0.939) for the combined aortic valve and mitral valve replacement, indicating that the risk models were good in predicting hospital mortality for surgeries.

The three risk models can quantitatively assess the hospital mortality risk in the patients treated with cardiac valve replacement.