The Society of Thoracic Surgeons 2018 Adult Cardiac Surgery Risk Models: Part 1-Background, Design Considerations, and Model Development.

BACKGROUND: The last published version of The Society of Thoracic Surgeons (STS) Adult Cardiac Surgery Database (ACSD) risk models were developed in 2008 based on patient data from 2002 to 2006 and have been periodically recalibrated. In response to evolving changes in patient characteristics, risk profiles, surgical practice, and outcomes, the STS has now developed a set of entirely new risk models for adult cardiac surgery. METHODS: New models were estimated for isolated coronary artery bypass grafting surgery (CABG [n = 439,092]), isolated aortic or mitral valve surgery (n = 150,150), and combined valve plus CABG procedures (n = 81,588). The development set was based on July 2011 to June 2014 STS ACSD data; validation was performed using July 2014 to December 2016 data. Separate models were developed for operative mortality, stroke, renal failure, prolonged ventilation, reoperation, composite major morbidity or mortality, and prolonged or short postoperative length of stay. Because of its low occurrence rate, a combined model incorporating all operative types was developed for deep sternal wound infection/mediastinitis. RESULTS: Calibration was excellent except for the deep sternal wound infection/mediastinitis model, which slightly underestimated risk because of higher rates of this endpoint in the more recent validation data; this will be recalibrated in each feedback report. Discrimination (c-index) of all models was superior to that of 2008 models except for the stroke model for valve patients. CONCLUSIONS: Completely new STS ACSD risk models have been developed based on contemporary patient data; their performance is superior to that of previous STS ACSD models.

Duke Scholars

Author Pamela Susan Douglas Medicine, Cardiology

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