EuroSCORE and the patients undergoing coronary bypass surgery

Objective: The aim of this study was to assess the performance of the European Risk System in Cardiac Operations (EuroSCORE) model to predict mortality in patients undergoing myocardial revascularization at the Division of Cardiovascular Surgery of Santa Casa de São Paulo Medical School. Methods: From May 2005 to November 2006, 100 consecutive patients undergoing coronary artery bypass surgery were retrospectively analyzed. The records of these patients were reviewed in order to retrieve those variables included in the EuroSCORE risk scoring method. The correlation of predicted and observed mortality was compared. Statistical analysis was performed using chisquare test for univariate analysis and Hosmer-Lemeshow Test for logistic regression model. Results: Hospital mortality was 5%. For EuroSCORE univariate analysis, findings were as follows: score 0-2 predicted mortality 0.40%, observed 0.00%; score 3-5 predicted mortality 1.45%, observed 0.00%; score greater than 6 predicted mortality 3.15%, observed 7.94%. Although these differences, p-value was 0.213 with no statistical significance. The p-value for the Hosmer-Lemeshow Test was < 0.001 indicating poor calibration of the model for this sample. Conclusion: The EuroSCORE model is a simple, objective system to estimate hospital mortality. However, to validate the logistic regression analysis, it is necessary hundreds of patients, which limit its widespread application. Descriptors: Myocardial revascularization/mortality. Risk assessment. Survival analysis. Severity of illness index.

objective system to estimate hospital mortality. However, to validate the logistic regression analysis, hundreds of patients are necessary, which limits its widespread application.
Descriptors: Coronary artery bypass surgery/mortality. Risk assessment. Survival analysis. Severity of illness index.

INTRODUCTION
The diversification in the possibilities of treatment of cardiac ischemic disease, the perfecting of operative strategies and the frequent advances in the areas of technology applied to medicine have caused a change in the profile of patients submitted to coronary artery bypass grafting (CABG). Additionally, with the aging of the population, a considerable number of elderly patients are referred for surgical treatment [1][2][3]. Thus, this therapeutic option has been offered to an increasing number of the most varied patients.
Mortality rate is routinely applied as an indicator of the quality of medical services [4]. Frequently, this indicator is utilized in a general form without knowing the exact profile of the studied population. In an effort to better stratify the risk for death in heart surgery, analysis models were developed with an aim to better characterize factors that influence the results. The European System of Cardiac Operation Risk Evaluation (EuroSCORE) is one of them [4][5][6]. This model, with univariate statistical analysis and logistic regression, demonstrated accuracy even when applied to non-European populations [7][8][9]. Based on data collected from 128 centers of eight European countries, this system evaluated 68 pre-operative and 29 operative risk factors that may influence hospital mortality. A total of 17 real risk factors were identified and for each one a score was attributed in a univariate analysis, thereby classifying the patients in three groups according to the obtained risk (low, medium or high). In the analysis by logistic regression a weight of the same factors is attributed. This is an easyto-use model accessible via the web which is popularizing its use.
The aim of this work was to evaluate the current clinical profile of patients submitted to CABG in the Cardiovascular Surgery Division of Santa Casa de São Paulo Medical School, to check the expected risk of death in this group of patients by applying the EuroSCORE and compare this with the observed deaths.

METHODS
The basis to carry out this research was a retrospective analysis of data. Over a period of 18 months between May 2005 and November 2006, 553 patients were submitted to heart surgery in the Central Hospital of the Santa Casa de São Paulo Medical School. Among these patients, 103 were submitted to CABG. A total of 100 patients, whose hospital records contained all the information necessary for an analysis of the criteria of operative mortality risk according to the 17 items included in the EuroSCORE (Table 1), were selected. Patients with incomplete data were excluded. In Table 2 the characteristics of our patients are presented classified in their respective groups of risk. The expected mortality rates, by applying the EuroSCORE, were compared with the observed deaths in the sample. Statistical analysis of the sample was achieved using the chi-squared test for univariate analysis and the Hosmer-Lemeshow test to calibrate and adjust the logistic regression of the EuroSCORE to the sample. The confidentiality of the data obtained from the patients' records were guaranteed and utilized exclusively for the current study which was approved by the Research Ethics Committee of the institution (protocol number 011/08).

RESULTS
Of the 100 patients, 60 were male. The patients' ages varied from 26 to 89 years old (mean = 61.3 years and SD = 11.7 years) with 45% of patients having ages equal to or greater than 65 years old. Seventy-six patients suffered from systemic arterial hypertension, 47 from dyslipidemia, 40 were smokers, 34 diabetics and 14 were obese (Body Mass Index > 35). Only 4 reported coronary artery disease in the family. Thirty patients underwent elective surgeries and 70, due to instable angina, post-infarction angina or critical coronary artery lesions, were submitted to urgent surgery. Eighteen individuals had lesions of the left main coronary artery. The prevalences of risk factors of the study sample and of the EuroSCORE study are illustrated in Table  2. Eight patients were considered low risk (Score 0-2), 29 medium risk (Score 3-5) and 63 high risk (Score > 6). Fifty- For 5 years or fraction > 60 years Female Prolonged use of de bronchodilators or steroids Claudication or obstruction > 50% of carotid arteries or previous or planned intervention of the abdominal aorta, carotid arteries or peripheral arteries Disease affecting walking or daily activities Requiring opening of the pericardium >2.3 mg in the pre-operative period Using antibiotics at time of surgery Any of the following: ventricular tachycardia or fibrillation or aborted sudden death, pre-operative heart massage, pre-operative ventilation before arriving in the surgery room, pre-operative inotropic support, intra-aortic balloon or pre-operative acute renal failure (<10 mL/h) Resting angina requiring the use of nitrates before arrival in the anesthesia room EF 30 -50% EF < 30% < 90 days Systolic pressure of the pulmonary artery > 60mmHg Performed before the following day's shift Another heart procedure at the same time as CABG Ascending aorta, aortic arch or descending aorta four patients were operated on employing cardiopulmonary bypasses (CPB). Of these, 25 were classified as high-risk patients. Hospital mortality, observed only in high-risk patients submitted to CPB, was 5%. Table 3 presents differences in the predicted and observed deaths according to the groups, as defined by the Hosmer-Lemeshow test. The p-value for this test was < 0.001 indicating bad adjustment or bad calibration of the model for the current sample size, despite of the accuracy, as estimated by c-statistics of the logistic method, being very high (94.7%). Table 4 shows that the divergences in the percentages of predicted and observed deaths were not statistically different according to the chi-squared test and univariate analysis (p-value = 0.213).

DISCUSSION
The current clinical profile of patients submitted to heart surgery puts into question   Without adjustments, taking into consideration associated risk factors, the conclusions might be incorrect, mainly when applied to a specialty in which, it is well known, that the most critical patients are those that benefit most from surgical as opposed to conservative clinical treatment [5,10,11]. Different systems of risk stratification have been utilized over the two last decades with the objective of: predicting mortality, evaluating therapeutic trends and analyzing cost-benefits [5,12]. Additionally, the application of a scoring system provides information to patients and their families in respect to the risk that individual patients will be submitted. This resource may be a facilitating tool in the communication and in the comprehension of risks.
Among the different studies to stratify operative mortality risk, the EuroSCORE has proved to be accurate, even when applied to non-European populations [7][8][9]. It is easy to use at the bedside and allows an analysis very close to the results, particularly in surgeries of the coronary arteries.
We retrospectively applied the EuroSCORE to analyze operative death in 100 consecutive patients operated in Santa Casa de Sao Paulo, a tertiary hospital that attends a large number of urgent cases and emergencies in its central emergency room. The service attends approximately 12,000 patients monthly. Consequently, 70% of the patients in our sample are referred by the emergency unit and present with instable or post-infarction angina or with severe coronary artery lesions and are submitted to urgent surgeries, a great difference to the patients enrolled in the EuroSCORE trial.
Many risk factors were observed in our sample including chronic pulmonary disease, neurological dysfunction, renal failure and pre-operative critical states as well as factors related to the heart such as instable angina, ventricular dysfunction with ejection fraction < 30% and an incidence of 48% of recent myocardial infarction.
The overall operative mortality was 5% which only involved high-risk patients. This corresponds to a mortality rate of 7.94% for this group, a higher percentage than that predicted by the EuroSCORE. In the low-and medium-risk groups there were no deaths, which is also different from the rates expected by applying the EuroSCORE. Other authors have reported similar results, where the death of low-risk patients was overestimated and of high-risk patients underestimated [13,14].
In respect to the underestimated mortality rate, the difference in this study, may be attributed to the characteristics of our service, notably that it attends many emergencies. This suggests a bias in the application of this risk score in institutions with this characteristic. There are no published reports applying the EuroSCORE to populations with a predominance of emergencies with which the results of this study can be compared. Even so, the discrepancies between the observed and predicted death rates employing univariate analysis were not statistically significant (Table 4).
Another positive characteristic of our service is our experience of CABG without using CPB. There were no deaths among patients who were not submitted to CPB. As has already been reported by Calafiore et al. [15], this surgical strategy, particularly in high-risk patients, contributes to a reduction in the operative mortality.
We applied the Hosmer-Lemeshow test and verified that, in spite of the high accuracy (94.7%), a p-value < 0.001 was attained indicating bad calibration of the sample for the logistic regression analysis. A larger number of patients (hundreds) are necessary to better adjust the applicability of this test. There is, therefore, a limitation in the application of this score in its logistic regression analysis as many services do not treat this number of patients. In the original work [5], for example, Germany contributed 4779 patients from 23 centers (average of 207 patients per center) and Spain 2444 from 25 centers (average of 97 patients per center), while the sample from Switzerland was 111 patients from a single center, thus similar to the number in our study. The mean number of patients enrolled in the original study was 148 patients per center, which is less than the necessary to validate the logistic regression analysis.

CONCLUSION
The EuroSCORE is a simple and objective model to predict operative death. However, to validate the logistic regression analysis, hundreds of patients are necessary, thus limiting its application when this number is not available.