The Society of Thoracic Surgeons risk model for operative mortality after multiple valve surgery.

BACKGROUND: This study was undertaken to develop The Society of Thoracic Surgeons (STS) mortality risk models for multiple valve procedures, which comprise 12% of total valve operations. METHODS: Data were obtained from the STS Adult Cardiac Surgery Database for 50,231 patients undergoing combinations of aortic (A), mitral (M), and tricuspid (T) surgery between January 1, 2004, and December 31, 2010, divided into developmental (2004 to 2009) and validation (2010) samples. Pulmonary valve operations, aortic root replacements, and dissection procedures were excluded, and insufficient AT procedures were available to model. Using stepwise logistic regression, the risk of operative mortality was estimated for each valve surgery type: AM, n=27,035; MT, n=18,686; and AMT, n=4,510. Two separate models were estimated, one that included only patient characteristics and status at presentation, and thereby would be suitable for performance profiling; and another that added discretionary operative variables such as arrhythmia ablation or valve repair. RESULTS: Unadjusted operative mortality was 7.6% for MT, 9.4% for AM, and 13.1% for AMT procedures. Significant risk factors for mortality included emergency presentation, advanced age, renal failure, reoperation, endocarditis, diabetes mellitus, severe chronic lung disease, peripheral vascular disease, coronary artery disease, and female sex. In models containing intraoperative variables, performance of arrhythmia ablation and atrioventricular valve repair were protective for mortality. In the validation sample, the model exhibited acceptable discrimination in each of the three procedural subgroups (C=0.711 to 0.727). CONCLUSIONS: Risk models were developed to predict operative mortality for patients having multiple valve procedures. These models may be useful for outcome assessment, quality improvement, patient counseling, shared decision making, and research.

Duke Scholars

Author Sean Michael O'Brien Biostatistics & Bioinformatics, Division of Biostatistics