| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|
|
|
|||||||
|
|||||||||
|
Interact CardioVasc Thorac Surg 2007;6:192-195. doi:10.1510/icvts.2006.138313 © 2007 European Association of Cardio-Thoracic Surgery
The performance of the EuroSCORE and the Society of Thoracic Surgeons mortality risk score: the gender factorCardiac Surgery Research, Inova Heart and Vascular Institute, Falls Church, VA 22042, USA Received 26 June 2006; received in revised form 19 December 2006; accepted 20 December 2006
*Corresponding author. Tel.: +1-703-776-8308; fax: +1-703-776-8303.
The purpose of this study was to explore potential differences in the performance of the EuroSCORE and the STS mortality risk score in the prediction of operative mortality following cardiac surgery with special focus on the impact of gender. We retrospectively reviewed 3125 consecutive cases of coronary artery bypass surgery performed at our institution between 2001 and 2004. STS and EuroSCORE (logistic [E-log] and additive [E-add]) operative mortality risk scores were calculated for each patient and stratified by gender (female: n=692; male: n=2433). Mortality risk scores were compared between the STS and EuroSCORE using C-statistics and likelihood ratios (LR). Stratified by gender, the E-log and E-add correlated well with the STS (female: r=0.77, 0.78, P<0.001; male: r=0.78, 0.79, P<0.001). Using C-statistics generated from logistic regression, both EuroSCORE models correctly modeled operative mortality compared to the STS. Among male patients, the EuroSCORE C-statistic (E-log, 0.808; E-add, 0.809; STS, 0.796) performed more comparable to the STS than female patients (E-log, 0.853; E-add, 0.855; STS, 0.827). Our results suggest that both the STS risk scores and the EuroSCORE are good predictors for operative mortality with slight advantage for the STS risk score. Combined with the ease of use, we conclude that the EuroSCORE is another viable, bedside instrument for surgeons looking for a preoperative assessment of mortality risk, particularly in female patients undergoing cardiac surgery.
Key Words: Coronary artery bypass grafting EuroSCORE; Society of Thoracic Surgeons; Risk score; Gender
The Society of Thoracic Surgeons mortality risk score (STS) and the European System for Cardiac Operative Risk Evaluation (EuroSCORE) scoring system are the two most frequently used risk profile systems used within the United States and Europe. The STS score is comprised of over 40 clinical parameters [1], whereas the EuroSCORE involves 18 clinical characteristics among three categories, each weighted accordingly [2]. Previous work suggests that despite substantial demographic differences between European and North American patient populations, the EuroSCORE was found to be very reliable in assessing operative risk in North American populations [3]. Results suggesting a disparity in mortality rates following a cardiac surgical procedure based solely on gender are mixed [4–6]. Further complicating the issue of the gender and mortality are two recent reports by Blankstein et al. [7], and Toumpoulis et al. [8]. Blankstein et al. present the results of an analysis of 15,440 patients following CABG and conclude that despite adjustment for known risk factors, gender alone is an independent risk factor for operative mortality following CABG [7]. However, Toumpoulis et al. concluded after reviewing 3760 patients following isolated CABG that despite similar early outcomes and long-term survival, there are far fewer predictors of long-term mortality among female patients compared to male patients [8]. The authors suggest gender profiles may be at the root of the matter and suggest that a gender independent assessment of mortality may be the preferred method. Female CABG patients have been associated with increased rates of postoperative depression, diminished functional gains, and increased readmission rates despite adjustment for presurgical comorbidities, age and body surface area [9, 10]. Other investigators claim that after adjustment for presurgical comorbidities and physiologic factors, rates of postoperative complications and mortality are similar [11]. Both the STS and EuroSCORE risk algorithm account for gender by the inclusion of gender term in each respective model; thus, with all other preoperative risk variables being equal, female patients will always demonstrate increased risk compared to males. Given that the sole purpose of risk scoring systems such as the STS and EuroSCORE is to adjust for presurgical clinical characteristics, including gender, this study was designed to explore potential differences in the performance of the EuroSCORE and the STS mortality risk scoring systems in the prediction of operative mortality following cardiac surgery with particular attention to the impact of gender.
We retrospectively reviewed 3125 consecutive cases of isolated coronary artery bypass patients (CABG) performed in our institution between January 2001 and August 2004. Data were abstracted from a computerized medical record system developed in accordance with the Society of Thoracic Surgeons National Database (STS). Preoperative risk factors were defined by the STS Adult Cardiac Surgery Database (version 2.41). Mortality was defined as all deaths occurring during hospitalization or those deaths occurring after discharge but within 30 days following surgery.
STS and EuroSCORE (logistic [E-log] and additive [E-add]) mortality risk scores were calculated for each patient. In addition, EuroSCORE calculations were validated against a downloaded EuroSCORE calculator (www.EuroSCORE.org) for 100 randomly selected patients. Spearman rank-order correlation coefficients were used to compare E-add and E-log scores to those of the STS. Mortality risk scores were compared between the STS and EuroSCORE using C-statistics, and likelihood ratios (LR+). C-statistics were calculated by logistic regression models to assess the predictive ability of either the STS or EuroSCORE system to correctly model an actual death after adjustment for potential confounders. The C-statistic determines the discriminatory power of each logistic regression equation. LR+ was calculated as the ratio of the true positive rate to the false positive rate. All statistical analyses were conducted using SAS software (Ver. 8.2, Cary, NC).
Demographic characteristics are presented in Table 1. Subjects averaged 63 years in age (females: 65.8 years, males: 62.6; P<0.001) and were primarily male. Female patients presented with statistically significant increased rates of the following comorbidities: cerebrovascular disease, diabetes, renal failure, dialysis, hypertension, morbid obesity, and peripheral vascular disease. Male patients presented with a statistically significant increased rate of smoking.
The overall mortality rate was 1.8%; 2.9% for female patients and 1.5% for males. The predicted overall STS mortality was 2.6%; 4.1% and 2.1% for female and male patients, respectively. The predicted overall EuroSCORE mortality using the E-log was 5.2%; 7.9% and 4.5% for female and male patients, respectively. The predicted overall EuroSCORE mortality using the E-add was 4.3±3.2; 5.8±3.2 and 3.9±3.1 for female and male patients, respectively. For both female and male patients, the E-log (r=0.77, 0.78) and E-add (r=0.78, 0.79) correlated well with the STS score. Among both female and male patients, C-statistics for the correct identification of observed mortality were highest for the E-log and E-add compared to the STS score (Fig. 1). No statistical significance was observed between either the E-log or E-add and the STS C-statistics estimates.
Likelihood ratio positives for selected sensitivities are presented in Table 2. At predetermined sensitivity levels, STS LR+ estimates are very similar between female (LR+: range: 2.00–4.53) and male (LR+: 1.50–4.45) patients. However, at predetermined sensitivity levels, female E-log and E-add LR+ estimates (LR+: E-log range: 2.40–6.83; E-add range: 1.49–8.23) are demonstrably increased compared to their male counterparts (LR+: E-log range: 1.94–3.48; E-add range: 2.06–3.22).
In this study we demonstrated comparable mortality estimates between the two scoring systems in the prediction of mortality following a CABG. This was even more significant in the female patient population. The EuroSCORE allows the clinicians to use an easy calculator to predict mortality. The findings in this study demonstrate that the EuroSCORE is valid especially when risk modeling female patients. Comparing the results of various scoring systems side by side is difficult because the instruments are scaled differently. For example, an STS risk score of 4.0 is not the same as a 4.0 for either iteration of the EuroSCORE. We have attempted to standardize these scales by the use of sensitivity and specificity. These values have no clinical application towards the patient and can only be used for comparison purposes. The issue of comparing scoring systems presents a dilemma for any surgeon when deciding which instrument to choose. When comparing risk models, two issues should be considered: 1) the predicted individual mortality estimate and 2) the subsequent aggregate predictor of mortality. The STS risk score is a complex tool to implement at the bedside but provides mortality risk estimate much closer to actual observed rates. The EuroSCORE, however, is very easy to use at the bedside but provides individual mortality estimates exceeding the STS estimates and actual observed estimates (Fig. 2).
The utilization of preoperative mortality risk algorithms to determine the best treatment modality for a given patient with cardiovascular disease is a common practice among cardiac surgeons and cardiologists. The EuroSCORE has gained increasing popularity due to its format, and is now used quite extensively. However, potential criticisms for its use are over simplification, inaccuracy and a lack of reliability. Thus, given similar predictive abilities between risk scores, would the model choice be facilitated by ease of use? In practice at the bedside, the use of risk scoring algorithms serves two purposes beyond simply giving both the surgeon and patient a predicted numerical outcome profile of the upcoming surgery: 1) the patient is more fully informed of the risks inherent in their upcoming risky procedure and 2) provide the surgeon with an objective, measurable risk profile for comparison purposes should the surgery be delayed, thereby allowing time for a diminished risk profile. Research comparing the EuroSCORE to other risk systems has resulted in mixed results [13–17]. Kawachi et al. compared the EuroSCORE additive model to the classic Parsonnet model with excellent result [12]. Among over 800 cardiac surgery patients with an overall mortality rate of 4.5%, EuroSCORE predicted mortality of 1.4% compared to 2.4% for the Parsonnet model. Nilsson et al. found the EuroSCORE to be a much better predictor of 30-day mortality among CABG patients compared to the STS model [13]. Among over 4400 CABG patients with an actual mortality rate of 1.89%, the EuroSCORE AUC was 84% vs. 71% for the STS. Furthermore, in a sample of over one-half million coronary bypass or valve surgery patients, Nashef et al. recommended the use of the EuroSCORE over the STS [3] after demonstrating that the EuroSCORE predicted mortality was remarkably similar to the observed mortality (3.994% vs. 3.992%). The case for which scoring system to use may rest on ease of use. The STS estimates are closer to observed rates compared to the EuroSCORE but that is to be expected; the STS prediction model is extremely comprehensive and includes over 50 clinical variables of preoperative clinical variables versus 17 for the EuroSCORE. However, given the conservative nature of both models, model ease of use may dictate the use of the EuroSCORE. It would be of note to understand more fully how the surgeon uses any tool in the preoperative period. Are risk estimates generated at the bedside or only after surgery during the data collection, cleaning, and review period some months after surgery? The STS clearly states that their estimates are designed for performance review purposes but in practice, is this the case? Currently, our surgical group uses the STS estimates months following any given individual case as a part of a quarterly review process. With the EuroSCORE, however, we have implemented a process by which we stratify the additive model into either low (0–2), medium (3–5) and high risk (6+) categories and present the patient with past mortality estimates within these categories. This process achieves the benefits of presenting the patient with preoperative risk estimates, along with a stated reason for their design, and allows for a more complete statistical analysis upon later review after subsequent postoperative data collection and review. In this report, we found that the EuroSCORE more accurately predicted mortality among female patients compared to the STS risk score. This may be due to one of several reasons. Firstly, the results are simply an artifact unique to our sample alone and not applicable to other sample populations. Secondly, while women compared to men, have been shown to have smaller coronary arteries after controlling for body size [14], present with increased angina severity [15], and congestive heart failure [15] and present with increased severity of disease at time of referral [14, 15], the STS risk model with its large number of clinical variables included may actually over capture and suppress results. Traditionally, statistical modeling procedures must balance a full and accurate model with parsimony, e.g. the number of terms used in the model. The inclusion of over 50 clinical parameters, as well as potential interaction terms would lead to a risk algorithm involving hundreds of terms. This in turn would require a sample size of many thousands of patients to derive sufficient statistical power for meaningful results – a sample size few institutions could meet. The EuroSCORE with its relatively few terms may strike a better balance between sample size and parsimony. The gender difference we report may be useful to the clinician during the preoperative assessment period. Extra attention to those clinical variables not included within the risk algorithms may warrant further attention, especially among male patients. Such knowledge may help lessen their traumatic experience. Our own experiences suggest that female patients presenting for open-heart surgery tend to be slightly more depressed, thereby requiring increased preoperative and postoperative counseling. Female patients, however, might be served better by the increased knowledge that research has demonstrated that the presented risk scores may be more accurate among female patients. We conclude that among CABG only patients, the EuroSCORE and STS mortality risk scores are very similar in predictive performance. Furthermore, when compared to the EuroSCORE, while the STS predicted mortality estimates presented at the bedside are closer to future actual observed estimates, the EuroSCORE upon further analysis is actually a slightly better predictor of mortality. Combined with the ease of use, we conclude that the EuroSCORE is a viable, easy to use alternative for surgeons looking for a preoperative assessment of mortality, particularly in female patients undergoing cardiac surgery.
|
||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
| ANN THORAC SURG | ASIAN CARDIOVASC THORAC ANN | EUR J CARDIOTHORAC SURG |
| J THORAC CARDIOVASC SURG | ICVTS | ALL CTSNet JOURNALS |
Top
Abstract
1. Introduction
