Prediction of Perioperative Mortality in Cardiac Surgery – Risk Scoring Systems: In Quest of the Holy Grail

Risk models are widely used to predict patient outcomes after cardiac surgery. Majority of models incorporate operative mortality which is considered to be a measure of the quality of care in cardiac surgery. These models allow accurate prediction of mortality, thereby judiciously offering a surgical strategy for high-risk patients and to achieve optimal outcomes. Over the last quarter of a century, more than 20 of the risk stratifications models have been used in cardiac surgery and the majority of them (over 20) are for adult cardiac surgery and a few for congenital heart defect surgeries. Most of them have been proposed from North America or Europe. Preoperative risk assessments have important implications and may facilitate informed consent, influence consideration for the timing, and choice of surgical interventions. The risk prediction scores have an impact not only on the patient's well-being but also to benchmark institutions, surgical procedures, and surgeons. Scoring systems were developed to predict mortality rates after high-risk cardiac surgery. There was a paradigm shift in the manner coronary artery disease which was managed in the 1970s when coronary artery bypass grafting (CABG) was the gold standard. With the introduction of percutaneous coronary intervention to tackle coronary artery disease, the spectrum of patients undergoing CABG was rapidly changing, and currently, more and more high-risk patients are undergoing CABG surgery. In the late 1980s Parsonnet score, the first-ever risk stratification score was applied by several surgeons and more variables were used to derive either simplified or complex scoring systems.[1,2] Later on, the Euro SCORE I,[3] Euro SCORE II,[4] Society of Thoracic Surgeons (STS) risk score,[5,6,7] age-creatinine-ejection fraction (ACEF) scores,[8] ACEF II score,[9] and Northern New England score[10] were developed. All these scores use multiple combinations of patient factors such as age, comorbid conditions, cardiovascular risk factors, and extracardiac risk factors. Euro SCORE II and STS scores are the most frequently used risk scores in the United States and Europe. Euro SCORE was developed and validated on data collection from 128 European surgical centers in 1995 and the results were published in 1999. European System for Cardiac Operative Risk Evaluation The European System for Cardiac Operative Risk Evaluation (Euro SCORE) was initially published in 1999 and was the most rigorously evaluated risk stratification system in cardiac surgery.[3,4] Combining 17 risk factors (patient, cardiac, and operation), the Euro SCORE method is a valuable measure for the prediction of immediate death after adult cardiac surgery. The logistic Euro SCORE was aimed at developing a scoring system predicting early mortality in cardiac surgical patients in Europe based on operative risk factors. It was developed from a large European database and included 13,302 patients. The Euro SCORE provides two methods for calculating predicted outcomes: the original additive model and the more recent logistic model.[11,12] Validation of the Euro SCORE took place worldwide in different population groups. Although well-established and validated in several patient population groups, additive Euro SCORE model overestimates mortality in low-risk patients and underestimates mortality in high-risk patients.[13,14] On the other hand, the logistic Euro SCORE model has been reported to over predict risk despite worsening risk profiles of patients and improvements in cardiac surgical outcomes. This scoring model is more accurate in predicting mortality in combined CABG and valve surgery.[15] The Euro SCORE can be calculated online. Society of Thoracic Surgeons Score The STS risk models for various cardiac surgical procedures have been developed since 1999[5,6,7] and there is more extensive data entry required to calculate the score and is also available online. A wide variety of endpoints (mortality and multiple morbidities) are included in some of the models calculating risk for isolated CABG, valve surgery, and combined CABG and valve surgeries. The risk models have been modified periodically and 27 new adult cardiac surgery models have been developed and validated.[16] The predictive performance of the STS risk score in general compared to other risk models and remains the most widely used model in the United States. The models include mortality as well as multiple morbidity endpoints including cerebrovascular events, renal failure, prolonged ventilation, and deep sternal wound infection. The STS score was not included in the most comprehensive analysis comparing 19 different scoring systems, while other studies demonstrated that the STS score model showed similar discriminative capability and predictive performance with Euro SCORE and other risk scores.[17] Parsonnet Score The Parsonnet score was described by Victor Parsonnet in 1989 in the United States and it was a uniform reporting system for operative mortality risk in all cardiac surgical procedures. It was developed from data on 3500 patients collected between 1982 and 1987.[1] Parsonnet score received widespread acceptance but the predictive accuracy has been reduced as a result of advances in treatment. The advantage of the Parsonnet score is the variables representing risk factors which are objectively measurable and easily recordable and well validated by several clinicians and variables that are completely or partly based on a subjective assessment such as the class of angina and diffuseness of coronary artery disease, operative priority is not included in the risk scoring model.[2] Age-Creatinine-Ejection Fraction Score The ACEF score were introduced in 2009 and was derived from a cohort of the single institution.[8] It is presently included in the guidelines for myocardial revascularization of the European Society of Cardiology and Association for Cardiothoracic Surgery as a risk stratification model for surgical and percutaneous myocardial revascularization. An updated version of the ACEF (ACEF II) score inclusive of emergency surgery and preoperative anemia was well calibrated. Discrimination of the ACEF II was significantly better than ACEF and with good calibrated properties.[9] Northern New England Score The Northern New England scoring system was initially published in 1992 with the data collected from the United States.[10] The initial model only predicted CABG mortality; later, the new Northern England Cardiovascular Disease study group developed another scoring system based on the data between 1996 and 1998 that included neurological events and mediastinitis. This scoring model was adopted by the American College of Cardiology/American Heart Association guideline committee on CABG.[17] The evidence on the relative utility of multivariable models used to predict mortality in cardiac surgery is scarcely reported in the literature. A few evaluated the evidence on comparisons that have been made among the Euro SCORE II, STS cardiac surgery risk models, the ACEF score, and Northern New England score. These are the established multivariable core models used currently to predict perioperative mortality aiming cardiac surgery that has been adopted by several recent guidelines. In a meta-analysis, 22 articles were included that compared established risk models: Euro SCORE II, STS score, and ACEF score for perioperative mortality during cardiac surgery. In this meta-analysis, 33 pairwise comparisons among 3 score risk models that have been recommended for use in risk stratification before cardiac surgery.[18] Euro SCORE II and STS score performed similarly and tended to outperform the ACEF score in terms of discrimination measured by area under the curve (AUC). The three core score models did not show any consistent differences in calibration performance with the exception that calibration of the STS score appeared superior to the ACEF score. The superior performance of the Euro SCORE II and STS score likely reflects that the ACEF score was developed specifically for patients who underwent elective cardiac surgical procedures. In numerous studies, conclusions often drawn indicate that Euro SCORE II and STS scores are widely used, even though these two scores give different predictions in different groups. Several models in current usage have predicted a rising probability of operative mortality while the observed mortality has decreased. This is due to an increasing prevalence of high-risk patients and significant advances made in the diagnosis and interventional or surgical procedures in cardiology and cardiac surgery. Currently, the risk models are developed from the prospectively maintained specialty-specific databases that contain crucial variables that have been demonstrated to be associated with outcome. Some models have been criticized for not being able to predict individual risk. The prediction of morbidity and mortality varies among several risk prediction models based on the variables used to gauge accurate predicting ability: the area under the receiver operating characteristic curve and C-index are variable for some patient populations utilizing different scoring systems. Good predictive ability is expected when the C-index is above 0.7 and strong predictive ability of the C-index is above 0.8.[19] Another meta-analysis[20] examined the performance of the Euro SCORE II and STS score specifically in-patients who underwent aortic valve replacement (AVR). The meta-analysis found that Euro SCORE II slightly outperforms STS score concerning calibration (Euro SCORE II O: E ratio 0.94 vs. STS score O: E ratio 0.84), whereas STS score slightly outperformed the Euro SCORE II concerning discrimination (STS score AUC 0.75 vs. Euro SCORE II AUC 0.73). The adoption of a single system is preferred to assess the risk of mortality for patients undergoing cardiac surgery. The Euro SCORE II and the STS scoring systems yield mortality rates that are both close to the actual rates observed at several institutions. The discriminatory ability of the Euro SCORE II and the STS risk score was superior to that of Euro SCORE I.[21] All these risk stratification models share important limitations. Preoperative risk factors included in a model can change significantly overtime leading to substantial underestimation or overestimation of postoperative risk. Caution should be exercised when interpreting the results of these models for individual patient's risk as these are derived from large databases. The STS risk score and the Euro SCORE II are the most widely used risk score systems all over the world although the STS risk score is widely used in the United States. The Euro SCORE II is almost as effective as the STS risk score in predicting operative mortality. The Euro SCORE has the added benefit of having greater flexibility in terms of its applicability to variety of cardiac surgical procedures. In terms of specific surgical groups, Euro SCORE II is most useful for establishing the risk in patients undergoing either CABG only or mitral valve surgery only, whereas the STS score is most useful for patients who undergo either AVR or CABG with concomitant valve surgery. All the risk score systems have benefits and drawbacks and are meant to add the clinical decision-making.[21] There is no risk stratification model currently available to apply for Indian patients undergoing cardiac surgery. We may have to include risk variables peculiar to the Indian population in developing the risk scores.