Clinical background
Before cardiac surgery, clinical staff need to be able to classify the patient's risk of mortality and morbidity quickly, both to inform the patient and to plan resources. The most widely used instruments, such as EuroSCORE and Parsonnet, rest on summing a large number of variables and can be laborious to calculate at the bedside. The Cardiac Anesthesia Risk Evaluation (CARE) score was developed as an alternative: a classification into five clinical classes resting on the anaesthetist's or surgeon's overall judgement rather than on a fixed checklist. The aim was to offer an instrument that can be used within minutes and yet performs on a par with the multifactorial indices.
Determining the Cardiac Anesthesia Risk Evaluation score
CARE consists of two components: a clinical class (1 to 5) and a modifier for emergency surgery.
The clinical class is assigned on three dimensions: the degree of control of medical comorbidities, surgical complexity and general condition:
| Clinical class | Criteria |
|---|---|
| 1 | Stable heart disease, no other comorbidity, non-complex surgery |
| 2 | Stable heart disease with one or more medically controlled comorbidities, non-complex surgery |
| 3 | Any uncontrolled medical condition, or complex surgery (e.g. reoperation, combined procedures) |
| 4 | Any uncontrolled medical condition together with complex surgery |
| 5 | Poor left ventricular function, dialysis-dependent renal failure, reoperation or surgery for active endocarditis |
To this is added a modifier: if the procedure is an emergency, that is, cannot be postponed for full investigation or optimisation, an E is added to the class (for example 3E). This marks the fact that the patient has not undergone complete preoperative optimisation.
The instrument was derived in a prospective study at the University of Ottawa Heart Institute, in which 3,548 patients undergoing cardiac surgery were classified with CARE and with three established multifactorial indices (Parsonnet, Tuman and Tu) [1]. The first 2,000 patients formed the reference group for discrimination analysis and the following 1,548 for calibration. Two independent assessors assigned CARE scores, and eight different anaesthetists scored their own patients to test the usability of the instrument in routine practice.
Interpretation in practice
CARE is not a continuous scoring scale but an ordinal classification. Each class corresponds to a stepwise higher risk, and the E modifier signals that risk may be underestimated because optimisation could not be carried out.
| Class | Risk level | Clinical action |
|---|---|---|
| 1–2 | Low | Standard perioperative management. Patient information can be restrained in its statement of risk. |
| 3–4 | Intermediate | Increased vigilance. Consider intensified postoperative monitoring. Identify modifiable factors that can be optimised if time allows. |
| 5 | High | A high-risk patient. Prepare for extended postoperative care, including a longer stay in intensive care. Discuss the risks explicitly with the patient and the team. |
| E modifier (any class) | Risk may be underestimated | The same measures as for the base class, but with awareness that incomplete optimisation adds uncertainty. Consider raising the level of preparedness by one step. |
In a modern cohort from the same institution (6,627 patients, 2009 to 2015), in-hospital mortality was clearly graded by CARE class: 0 per cent for class 1, 0.3 per cent for class 2, 1.3 per cent for class 3, 4.6 per cent for class 4 and 13.6 per cent for class 5 [2]. This confirms that the classification retains its ability to separate risk levels even in a contemporary population with lower overall mortality.
Validation and performance
In the derivation study, CARE performed on a par with the three multifactorial indices for both mortality and morbidity, with good calibration for all the models except Parsonnet, which failed calibration for morbidity [1]. The c-statistic for CARE was in the same range as for the more complex indices.
An external validation was carried out at the Centre Hospitalier Universitaire Pitié-Salpêtrière in Paris with 556 consecutive patients [3]. CARE was compared with EuroSCORE and the Tu index. All three models showed acceptable calibration, and discrimination for mortality was comparable between CARE and EuroSCORE. The study also examined inter-observer agreement between specialties: agreement was 90 per cent between two anaesthetists, 83 per cent between anaesthetists and surgeons and 77 per cent between anaesthetists and cardiologists [3]. This is remarkably high agreement for an instrument resting on clinical judgement rather than a checklist.
In a comparative study at the Ottawa Heart Institute from 2006 to 2009, with 3,818 patients, CARE was compared with both the additive and the logistic EuroSCORE [4]. CARE had lower discrimination than the logistic EuroSCORE but retained good calibration, while both EuroSCORE variants substantially overestimated mortality. Notably, CARE remained well calibrated a decade after it was developed, without any need for recalibration, which is a rare property among risk indices.
In the modern cohort from 2020, a c-statistic of 0.80 was reported for CARE, confirming good discrimination in a population with contemporary surgical technique and perioperative care [2].
Limitations
CARE rests on the assessor's clinical judgement and has no fixed scoring checklist. This is the instrument's strength at the bedside but also its weakness: the quality of the classification depends entirely on the assessor's experience. Inter-observer agreement is high among anaesthetists but falls when other specialties are involved, which means that CARE should be assigned by the person with the most experience of cardiac anaesthesia [3].
The instrument was developed and validated in patients undergoing cardiac surgery on cardiopulmonary bypass. Patients undergoing off-pump surgery, aortic surgery, heart transplantation or implantation of a mechanical assist device were excluded from the validation studies [2], and CARE has not been validated for these groups.
The E modifier is a binary marker and does not capture the degree of urgency. A patient who must be operated on within the hour and one who can wait 24 hours receive the same modifier, even though the risk profiles may differ substantially.
CARE gives a classification but no estimated probability of mortality as a percentage. For patient information and for comparisons between institutions, an instrument that gives a quantitative risk percentage, such as EuroSCORE II, may be preferable. On the other hand, EuroSCORE has shown a tendency to overestimate in contemporary populations, while CARE retains its calibration [4].
References
- Dupuis JY et al. The cardiac anesthesia risk evaluation score: a clinically useful predictor of mortality and morbidity after cardiac surgery. Anesthesiology 2001. PMID: 11176081
- Ristovic V et al. The Impact of Preoperative Risk on the Association between Hypotension and Mortality after Cardiac Surgery: An Observational Study. Journal of Clinical Medicine 2020. PMID: 32629948
- Ouattara A et al. Predictive performance and variability of the cardiac anesthesia risk evaluation score. Anesthesiology 2004. PMID: 15166559
- Tran DT et al. Comparison of the EuroSCORE and Cardiac Anesthesia Risk Evaluation (CARE) score for risk-adjusted mortality analysis in cardiac surgery. European Journal of Cardio-Thoracic Surgery 2012. PMID: 21803595
