Risk stratification in cardiac surgery: Algorithms and applications
Abstract: The aims of this research was to compare different risk score algorithms with regard to their validity to predict 30-day and one-year mortality after open-heart surgery, to evaluate if the preoperative risk stratification model EuroSCORE predicts the different components of resource utilization in cardiac surgery, and to systematically evaluate the accuracy and performance of artificial neural networks (ANNs) to select and rank the most important risk factors for operative mortality in open-heart surgery.
Preoperative evaluation of the surgical risk is an important component in cardiac surgery. Risk stratification can give patients and their relatives insight into the existent risk of complications and mortality, and aid in the selection of cases for surgery versus alternative, non-surgical therapies. It may also predict the need for hospital care resources and improve the quality of care. A few comparative studies of different risk algorithms exist, but the relative performance of the risk scoring systems currently used is unclear, and it still remains difficult to risk-stratify individual patients.
The present work identified four cardiac surgical risk models with a superior performance, with the EuroSCORE algorithm performing best. Though the algorithms were originally designed to predict early mortality, the one-year mortality prediction was also reasonably accurate. The additive EuroSCORE algorithm was also shown to be useful to predict intensive care unit (ICU) cost and an ICU stay more than two days after open-heart surgery. In an attempt to improve the mortality prediction further, a machine-learning technique, ANNs, was used. This identified mortality risk factors in a ranked order and defined a minimal set of risk variables resulting in a superior mortality prediction, compared with previously developed algorithms.
CLICK HERE TO DOWNLOAD THE WHOLE DISSERTATION. (in PDF format)