Coronary artery disease is a leading cause of morbidity and mortality in octogenarian patients [1, 2]. Octogenarians represent the fastest growing segment of our population; over 40% of the very elderly manifest cardiovascular disease [3]. The indications for coronary artery bypass grafting (CABG) surgery are well defined and the number of octogenarians being referred for surgical coronary revascularization is increasing [4]. Accurate risk stratification and the prediction of operative mortality is essential as this aids the clinical decision-making process, helps estimate the need for resources, facilitates proper patient counseling, informed consent, and allows for monitoring of surgeon and institution performance through risk-adjusted outcomes.

Risk-scoring algorithms based on both patient history and functional status have been established. The most widely used are the Society of Thoracic Surgeons Predicted Risk of Mortality (STS) score and the European System for Cardiac Operative Risk Evaluation (EuroSCORE) [5, 6]. Despite the acceptance of these risk scores in the context of CABG, there is little data on the performance of the revised EuroSCORE II in comparison to the established STS, additive EuroSCORE and the logistic Euroscore. Furthermore, these scores have not been validated to specifically predict operative mortality in octogenarians.

The aim of the current study was to evaluate and compare the performance of the recently introduced EuroSCORE II with its previous version EuroSCORE (additive and logistic) and the STS score in predicting perioperative mortality in patients age 80 and older undergoing isolated CABG at our institution.

Relative to younger patients, those aged 80 and older had a significantly higher mean predicted mortality by the STS Score (3 ± 2% vs. 1 ± 1%; p < 0.001), additive EuroSCORE (8 ± 3% vs. 4 ± 3%; p < 0.001), logistic EuroSCORE (15 ± 14% vs. 5 ± 6%; p < 0.001), and EuroSCORE II (4 ± 3% vs. 2 ± 2%; p < 0.001) (Fig. 1). The total observed perioperative mortality of 2% for patients aged 80 years and older was not significantly different from the observed perioperative mortality of 1% for patients under age 80 (p = 0.323).

The receiver operator characteristic (ROC) for both patient groups are shown in Fig. 2a and b respectively. Patient’s age 80 and older had an AUC for the STS score of (0.672; p = 0.151, 95% confidence interval [CI] 0.459 to 0.883), additive EuroSCORE (0.709; p = 0.079, 95% CI 0.523 to 0.896), logistic EuroSCORE (0.694; p = 0.104, 95% CI 0.499 to 0.889) and EuroSCORE II (0.794; p = 0.014, 95% CI 0.696 to 0.893). Patients under age 80 had an AUC for the STS score of (0.829; p = 0.003, 95% CI 0.710 to 0.947), additive EuroSCORE (0.750; p = 0.023, 95% CI 0.544 to 0.955), logistic EuroSCORE (0.785; p = 0.010, 95% CI 0.585 to 0.985) and EuroSCORE II (0.845; p = 0.002, 95% CI 0.709 to 0.980). The p-value under the AUC represents the ability and accuracy of the score in predicting perioperative mortality in patients who undergo CABG surgery. Overall, all scores assessed underperformed patients age 80 and older compared, with EuroSCORE II demonstrating the best predictive ability in this group.

The net reclassification improvement index (NRI) of STS compared to EuroSCORE II was calculated. Area under the curve analysis demonstrates that the areas of the three curves are similar for patients under age 80, but different for patients age 80 and over where EuroSCORE II had the largest area. The Delong test indicates that the difference of the AUC between the three curves is not significant for both groups. However, NRI analysis reveals that EuroSCORE adds 16% reclassification improvement compared to STS for age under 80, and 67% improvement for in those age 80 and older. The 67% improvement results from ability of EuroSCORE II to discriminate all 6 deaths, while STS is able to discriminate only half (3 out of 6) of deaths.

We demonstrate in a large series of consecutive octogenarians undergoing isolated CABG that EuroSCORE II had the best performance for risk assessment among the tools available. However, the expected mortality calculated by all four risk scores overestimated perioperative risk for patients of all ages, most particularly in octogenarians.

Our study is unique in that it includes all four risk scores with a focused evaluation of their performance in isolated CABG surgery. This is particularly relevant given that octogenarians represent a growing proportion of patients being referred for CABG surgery [10]. Early studies report cardiac surgery operative mortality rates of up to 11% [11, 12]. Refinements in surgical technique and perioperative management has resulted in significantly improved outcomes [13], as well, differences in observed operative mortality in other reports may stem from inherent country or population specific outcomes [14, 15]. We report a low observed perioperative mortality of 2% in patients age 80 and older which is not different from existing contemporary reports [4, 16, 17]. Although elderly patients have worse early outcomes compared to younger patients, long-term outcomes for octogenarians at our institution after CABG are similar to, if not better than, the age-adjusted Canadian population. These are demonstrated by previously published reports on our cohort examining intermediate and long-term survival post-operatively for octogenarians compared to younger subjects [4, 18, 19].

Accurate risk stratification facilitates improved patient selection, prevents complications [20], improves patient quality of life and can justify the cost effectiveness of the procedure [21]. However, risk models, although helpful and objective for operative risk assessment and clinical decision-making, are not without their limitations. The original EuroSCORE and EuroSCORE II were developed with a patient population mean age of 62.5 years and 64.9 years, respectively [13, 22]. Our results show that all scores assessed significantly overestimated operative risk in octogenarians, and were well calibrated in patients under age 80. This identified imprecision in operative risk prediction has important implications for clinical decision making in elderly patients.

Frailty has been shown to be an independent predictor for both postoperative complications and in-hospital mortality after adjusting for age [23]. It is estimated that 20% of octogenarians are frail [24]. Although considered difficult to quantify as compared to the comorbidity-based risk assessment models currently in use, several scores have emerged as an attempt to assess frailty including the Edmonton Frailty Scale, Comprehensive Assessment of Frailty test, Fried Frailty Scale and others [24–26] These frailty scores are a multidimensional assessment of health and functional status incorporating sociodemographic, biomedical, cognitive capacity, independence with daily activities, social support and mood [27, 28].

Incorporating a quantitative measure of the degree of frailty can introduce objectivity to the assessment of biological status to complement conventional comorbidity risk assessment and improve the accuracy of operative risk assessment in the elderly. This would allow more accurate stratification of risk than current models and aid in selecting the most appropriate modality of intervention to best optimize outcomes for patients. In an age 70 and older population, Prudon et al. [29] demonstrated that the addition of gait speed to logistic EuroSCORE improved the accuracy of the model. Sundermann et al. [24] used a frailty score in addition to conventional risk scores and found frailty to only moderately correlate with STS and EuroSCORE risk models. Other investigators have demonstrated the utility of a frailty score in the preoperative assessment of their patients [24, 30]. Afilalo et al. demonstrate that both frailty and disability (as defined by a 5-m gait speed ≥ 6 seconds and the presence of 3 or more impairments on the Nagi scale) are complementary and additive to existing risk scores where inclusion of them improved the discrimination of the STS score [30]. Thus, incorporation of various risk factors specific to the elderly population as measures of physiological vulnerability may improve the effectiveness of cardiac surgical risk models [31].

Frailty is an emerging concept; collaborative efforts towards defining the optimal method towards measuring frailty and disability deserve further exploration. Multidisciplinary teams involving cardiac surgery, cardiology, geriatric medicine, physiotherapy and occupational therapy may help address the diverse elements specific to the elderly that contribute to increased risk [32]. A 64-slice-coronary computed tomography shows promise as a non-invasive alternative to the standard coronary angiography to detect CABG graft stenosis [33]. However, in a frail octogenarian population, symptoms and quality of life should be the driver for any further invasive or non-invasive investigations.

Currently, all risk prediction models overestimate surgical risk, particularly in octogenarians. The EuroSCORE II demonstrated better discriminatory accuracy for predicting operative mortality than STS, additive and logistic EuroSCORE in this population. Inclusion of new variables into these risk models, such as frailty, that are of particular interest in the elderly, may allow for more accurate prediction of true operative mortality and thus lead to improved decision making in this important group of patients.