Clinical background
EuroSCORE II exists to quantify operative mortality risk before cardiac surgery, a decision in which both patient and surgeon need as objective a basis as possible for consent and for planning the operation. Its predecessor, the original EuroSCORE (both the additive and the logistic version), systematically overestimated risk when applied to modern patient cohorts. In the derivation study, actual mortality was 3.9%, while the additive model predicted 5.8% and the logistic model 7.57% [1]. EuroSCORE II was developed to correct this failure of calibration and to give a more realistic picture of risk.
The score is not a decision tool in itself, but support for clinical judgement within a multidisciplinary heart team. It should help to identify patients in whom the operative risk is so high that alternative strategies should be considered, and to give patients with a low predicted risk a well-informed basis for consent.
Calculating EuroSCORE II
EuroSCORE II rests on a logistic regression model with 18 variables. Predicted mortality is calculated as:
where
The age term is coded as 1 up to and including 60 years and increases by 1 for each additional year. The other variables and their coefficients are shown in the calculator's fields. The variables comprise patient-related factors (age, sex, renal function, insulin-treated diabetes, chronic lung disease, extracardiac arteriopathy, poor mobility, previous cardiac surgery, active endocarditis, a critical preoperative state, NYHA class, CCS class 4 angina, left ventricular function, recent myocardial infarction, pulmonary hypertension) and operation-related factors (urgency, the extent of the procedure, surgery on the thoracic aorta).
The derivation cohort consisted of 22,381 consecutive patients undergoing major cardiac surgery at 154 hospitals in 43 countries over a 12-week period from May to July 2010 [1]. The primary outcome was mortality at the operating hospital. Secondary outcomes were mortality at 30 and 90 days. The dataset was split into a development part for model building and a validation part of 5,553 patients. In the validation part, actual mortality was 4.18% and predicted mortality 3.95%, with an area under the ROC curve of 0.8095 [1].
Interpretation in practice
EuroSCORE II gives a predicted probability of operative mortality as a percentage. Unlike score-based instruments with fixed risk bands, the model gives a continuous figure, and there are no generally accepted cut-offs defining "low", "intermediate" or "high" risk. Interpretation must therefore be made in relation to the nature of the procedure, the patient's preferences and the alternatives available.
| Predicted mortality | Clinical interpretation |
|---|---|
| Under 2% | Low operative risk. The decision to operate should generally be governed by the cardiological indication, not by the operative risk. |
| 2 to approximately 10% | Moderately increased risk. An individual judgement is required, particularly where an alternative treatment exists (for example TAVI in aortic stenosis). |
| Over 10% | Increased operative risk. Multidisciplinary discussion is indicated. The model overestimates risk in this group, so the actual probability may be lower than stated. |
An important clinical point is that EuroSCORE II is best calibrated in the low and intermediate risk range. At a predicted mortality above approximately 10 to 30%, the value should be interpreted with caution, since several validation studies show that the model tends to overestimate risk in this group [2, 3].
Validation and performance
Several external validation studies have been published since EuroSCORE II was introduced.
Multicentre validation (Italy). Barili et al. validated EuroSCORE II in 12,325 consecutive patients from three institutions over a six-year period [2]. Discrimination was high, with an AUC of 0.82 (95% CI 0.80 to 0.85), on a par with the derivation study. Calibration was good up to approximately 30% predicted mortality, but the model overestimated risk in patients with higher predicted values. The authors noted that removing statistically non-significant variables did not affect performance, suggesting that the model could be simplified without loss of predictive power [2].
Age-specific validation (France). Provenchère et al. analysed 7,161 consecutive patients at a single centre over seven years, of whom 832 (12%) were 80 years or older [3]. For patients under 80, the AUC was 0.81 (95% CI 0.79 to 0.84) with good calibration. For patients aged 80 or over, discrimination fell to an AUC of 0.67 (95% CI 0.60 to 0.73), which is unsatisfactory. Calibration was acceptable up to 10% predicted mortality in the older group but overestimated risk above that. This corresponds to approximately 20% of the older population; for the remaining 80% with a predicted mortality below 10%, the estimate was reliable [3].
Endocarditis population (the Netherlands). Heinen et al. validated EuroSCORE II in 2,569 patients operated on for infective endocarditis between 2013 and 2021, with 30-day mortality as the outcome [4]. The AUC was 0.72, lower than in elective cardiac surgery. The model overestimated mortality, particularly at predicted values above 20%, and discrimination was poor in emergency surgery. The authors recommended that EuroSCORE II can be used in endocarditis up to a predicted probability of approximately 20%, but that above that the value should be halved to approach the actual risk [4].
Isolated CABG (the Netherlands and Sweden). Roefs et al. validated EuroSCORE II for 120-day mortality after elective or urgent isolated CABG using data from national quality registries in both the Netherlands (45,016 procedures, 2013 to 2019) and Sweden via SWEDEHEART (16,374 procedures) [5]. In the Dutch cohort the c-statistic was 0.79 (95% CI 0.77 to 0.81) for the original model. In the Swedish validation cohort, a refitted model achieved a c-statistic of 0.80 (95% CI 0.78 to 0.83) with improved calibration compared with the original EuroSCORE II. The original model tended to underestimate risk in the small group of high-risk patients undergoing isolated CABG, which differs from the pattern in mixed cardiac surgical populations, where overestimation predominates [5].
Comparison with other models. A meta-analysis by Sullivan et al. compared EuroSCORE II, the STS Score and the ACEF Score across 22 studies with 33 comparisons [6]. EuroSCORE II and the STS Score performed equivalently (pooled difference in AUC = 0.00), while both outperformed the ACEF Score (difference in AUC 0.10 and 0.08 respectively, p < 0.05). Calibration and reclassification measures were, however, reported unevenly across the studies [6].
Limitations
EuroSCORE II has several important limitations that clinicians must be aware of:
Advanced age. In patients aged 80 years or older, discrimination falls markedly and the model overestimates risk at a predicted mortality above 10% [3]. Frailty, which the model does not capture, is an important risk factor in older patients and may explain part of the prediction error.
High-risk patients in general. At a predicted mortality above approximately 20 to 30%, calibration is poor whatever the age. The model overestimates risk in this group in mixed procedures [2, 3], but may underestimate it in isolated CABG [5]. High-risk patients are under-represented in the derivation cohort, which contributes to the uncertainty.
Infective endocarditis and emergency surgery. In endocarditis, discrimination is lower (AUC 0.72) and the model overestimates risk, particularly in emergency procedures [4]. EuroSCORE II must not be used to refuse indicated surgery in endocarditis solely on the basis of a high score.
Frailty and functional status. EuroSCORE II includes poor mobility of neurological or musculoskeletal origin, but does not capture the frailty phenotype in full. Cognitive function, nutritional status and sarcopenia are absent, even though these factors affect operative mortality in older patients.
Outcome definition. The primary outcome in the derivation study was mortality at the operating hospital, not 30-day mortality [1]. This can give a somewhat different picture of risk compared with studies using 30 or 120 days as the endpoint, particularly with a long hospital stay.
A static instrument in a dynamic environment. The derivation data were collected over 12 weeks in 2010. Since then, surgical technique, anaesthesia and perioperative care have developed, which may affect how current the model is. Roefs et al. showed that a model refitted with data from 2013 to 2019 improved calibration [5], suggesting that the coefficients of the original model gradually become outdated.
External validation in the SWEDEHEART registry
SWEDEHEART, the Swedish national quality registry for cardiac care, collects data on all CABG procedures in Sweden and has been used for external validation of EuroSCORE II [5]. In the Swedish cohort of 16,374 isolated CABG procedures (2013 to 2019), 120-day mortality was 1.5%, the median EuroSCORE II was 1.4, and the original model performed well but with scope for improvement in calibration in high-risk patients. A refitted model based on Dutch data was validated in the Swedish cohort with a c-statistic of 0.80 and good calibration [5]. This indicates that EuroSCORE II works well in routine Swedish practice for isolated CABG, but that high-risk patients should be assessed with additional clinical judgement.
References
- Nashef SA, Roques F, Sharples LD et al. EuroSCORE II. Eur J Cardiothorac Surg 2012;41(4):734-44. PMID: 22378855
- Barili F, Pacini D, Capo A et al. Does EuroSCORE II perform better than its original versions? A multicentre validation study. Eur Heart J 2013;34(1):22-9. PMID: 23028171
- Provenchère S, Chevalier A, Ghodbane W et al. Is the EuroSCORE II reliable to estimate operative mortality among octogenarians? PLoS One 2017;12(11):e0187056. PMID: 29145434
- Heinen FJ, Peijster AJL, Fu EL et al. External validation of EuroSCORE I and II in patients with infective endocarditis: results from a nationwide prospective registry. Eur J Cardiothorac Surg 2024;66(6):ezae418. PMID: 39579090
- Roefs MM, Sartipy U, Friberg Ö et al. Refitting EuroSCORE II for 120-Day Mortality After Coronary Artery Bypass Grafting Using Nationwide Registry Data. Interdiscip Cardiovasc Thorac Surg 2026;41(4):ivag076. PMID: 41803632
- Sullivan PG, Wallach JD, Ioannidis JP. Meta-Analysis Comparing Established Risk Prediction Models (EuroSCORE II, STS Score, and ACEF Score) for Perioperative Mortality During Cardiac Surgery. Am J Cardiol 2016;118(10):1574-82. PMID: 27687052
