| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|
|
|
|||||||
|
|||||||||
Ann Thorac Surg 2001;71:512-520
© 2001 The Society of Thoracic Surgeons
a Department of Surgery, Tulane University Medical School, New Orleans, Louisiana, USA
b Department of Cardiac Research, Denver Veterans Affairs Medical Center, Denver, Colorado, USA
c University of Colorado, Health Sciences Center, Denver, Colorado, USA
Accepted for publication April 14, 2000.
Address reprint requests to Dr Hartz, Department of Surgery, SL22, Tulane University, 1430 Tulane Ave, New Orleans, LA 70112
e-mail: rshartzmd{at}aol.com
 |
Abstract |
|---|
 Top Abstract IntroductionPatients and methodsResultsCommentReferences |
|---|
Methods. For 1994 to 1996, The Society of Thoracic Surgeons database records for 441,542 coronary artery bypass grafting-only procedures were analyzed. Baseline annual multivariate models were built. Gender and race were added to each model. Risk-adjusted OM rates were then calculated for race, gender, and their combination. Patients were also stratified into groups of comparable predicted OM to allow for a direct comparison of risk-matched Caucasians and non-Caucasians.
Results. Of the procedures, 28.2% were on women and 8.5% on non-Caucasians. Overall, OM was 3.29%. Multivariate risk-adjusted OM varied by gender and race (p < 0.10). Risk-adjusted OM rates (with 95% confidence intervals) were 4.0% (3.9% to 4.1%) for females and 3.2% (3.2% to 3.3%) for males. Risk-adjusted OM rates were 3.9% (3.7% to 4.1%) for non-Caucasians and 3.3% (3.2% to 3.3%) for Caucasians. Among equally risk-matched Caucasians and non-Caucasians, non-Caucasians had significantly higher (p < 0.005) mortality among the lower risk subgroups (up to 10% predicted OM) but not among the higher risk subgroups.
Conclusions. Race and gender are independent predictors of adverse outcome following coronary artery bypass grafting, holding all other risk factors constant.
|
Introduction |
|---|
|
TopAbstractIntroductionPatients and methodsResultsCommentReferences |
|---|
Numerous studies have documented racial variations in the application of diagnostic studies (cardiac catheterization) and revascularization procedures (angioplasty and CABG) for coronary artery disease. Less certain is the impact of race on outcomes after these interventions. In the 1995 CABG-only STS risk model, the Native American race was the only race-related variable that remained in the final predictive model [4].
For the first time, the 1996 STS CABG-only multivariate model indicated that race (defined as Caucasian and non-Caucasian) was an independent predictor of operative mortality (OM) after holding all other risk factors constant [5]. Although there may be important differences in access to care based on racial subgroupings, the existing literature reveals that few studies have addressed the issue of procedural outcomes as related to race. Hence, it was our collective perception that the potential influence of race on procedural outcomes related to CABG surgery required further investigation.
The purpose of this analysis, therefore, was to determine whether race alone or in combination with gender affected the risk of 30-day operative mortality after CABG-only surgery. A concern was raised that the influence of race on OM might be explained either fully or in part by a collection of other risk factors such as severity of coronary disease or presence of comorbid disease states. For example, other studies have indicated that factors related to body size and coronary diameter have been linked to female gender as an independent risk factor for adverse events postcardiac surgery [6, 7]. Thus, the relationships of race with many other patient risk characteristics and process of care measures (such as body size and use of internal mammary artery grafts) were also explored.
|
Patients and methods |
|---|
|
TopAbstractIntroductionPatients and methodsResultsCommentReferences |
|---|
Although the anthropologic correctness of the "Caucasian" versus "non-Caucasian" categorization may be questioned, we had available to us only the racial designation provided on the STS Database forms. A more precise anthropologic categorization may be "North Western European" versus "non-North Western Europe." Given the small number of records within any race subclassification, the 1997 Expert Advisory Panel recommended that the variable for race be reclassified into two categories (Caucasian and non-Caucasian) for the 1996 STS CABG-only risk modeling process. Thus, this recommendation was used as the basis of this special analysis also.
Potential univariate interrelationships between study variables were analyzed using a Chi-square test or a Wilcoxon rank sum test. For the Chi-square tests, an adjustment (reduction of Chi-square by factor of 10) was made to account for the large size of STS population studied and a Cramer’s V was also reviewed as a measure of association to compare nominal variables (specifically, race). For reference, a Cramer’s V value less than 0.10 generally indicates a weak relationship is present (although the relationship may be statistically significant) between the variables analyzed. For model development and testing purposes, one half of all records were randomly assigned to learning and test data sets [8]. The OM rates in the learning and test data sets were reviewed to determine that no inherent bias in these six data sets existed using a Chi-square analysis. Statistical test performance metrics were evaluated appropriately [9].
Three baseline annual multivariate logistic models (for 1994, 1995, and 1996) were built according to The Society of Thoracic Surgeons standard specifications excluding the race and gender variables for comparative purposes. Univariate screening was performed to determine variables eligible for model entry.
For all records, the standard STS approved clinical data substitutions were used [5]. For race, this substitution required the records with missing race (14,854 or 3.36%) to be reclassified to Caucasian, as the clinical default category. To test this key assumption, the study’s primary multivariate analyses were rerun after dropping all records with missing race data.
As a potential measure of body morphology, both body mass index (BMI) and body surface area (BSA) variables were explored. For purposes of this study, BMI was defined according to the American Society of Clinical Nutrition’s four categories: 20 to 25 kg/m2 = normal; 26 to 30 kg/m2 = grade I obesity; 31 to 40 kg/m2 = grade II obesity; and greater than 40 kg/m2 = severe obesity [10].
The two key variables of study (gender and race) were reinserted into each baseline model as independent main effect variables. Subsequently, the interaction effect (gender-race) was assessed. The statistical significance of each of these two key variables and the interaction variable on the relative improvement in risk model performance were evaluated. The purpose of this comparative analysis was to determine in a multivariate context whether these key variables or their combination were statistically significant predictors of OM [11].
The c-index, which is a measure of each models’ predictive power, was calculated. The c-index represents the area under the receiver operating characteristic curve (which reflects the relative sensitivity and specificity of the model in predicting OM). In theory, the c-index may range from 0.5 to 1.0, where a value of 0.5 is useless for prediction purposes and a value of 1.0 represents perfect prediction. Using the method identified by Hanley and McNeil, a nested c-index comparison was performed to determine difference between the annual models’ performance, which were derived from the same cases [12].
The calibration was assessed using a Hosmer-Lemeshow (H-L) test, which determines the association between the observed OM and the estimated risks for OM from the model across deciles of patient risk [13]. In general, a model with an H-L test p value less than 0.05 indicates a lack of fit, which is poor calibration across the spectrum of patient risk. Due to inherently large sample sizes, this test statistic has been adjusted to compare with other reported logistic regression models in the literature appropriately [14]. To eliminate the possibility of a multiple comparisons problem, the c-index and H-L test results for all three annual models were calculated in aggregate on the test data set.
Risk adjusted operative mortality rates were calculated for race and gender, as well as the four combined gender/race categories. Confidence intervals were calculated for the risk-adjusted mortality rates provided for both the gender and race variables, as well as the four combined categories. In addition, the final model was used to place all patients into 7 clinically relevant subgroups of similar predicted risk (0% to 2.5%, 2.5% to 5.0%, 5.0% to 10%, 10% to 20%, 20% to 30%, 30% to 50%, and 50% to 100%) to allow direct comparisons of observed/expected operative mortality.
|
Results |
|---|
|
TopAbstractIntroductionPatients and methodsResultsCommentReferences |
|---|
Univariate analysis
For the study population, 8.5% of CABG procedures performed were on non-Caucasians and 28.2% were on women. Univariate analysis revealed that a statistically significant relationship existed between gender and race, confirming the perception that women undergoing CABG are more likely to be non-Caucasian than are men undergoing the procedure. For Caucasians, 27.6% of the CABG procedures were performed on women whereas for non-Caucasians, 35.3% were performed on women (p < 0.001). Table 1 illustrates the distribution of the records and deaths by both race and gender categories.
|
|
Multivariate analysis results
The three annual baseline models were used as the comparison for other models developed for each key variable of interest (Table 3). As BMI category did not enter any of the baseline multivariate statistical models generated (p > 0.16 for all 3-year periods), body surface area (the traditional STS measure of body morphology) was used in the baseline multivariate modeling process. The baseline model (built without race or gender) had a c-index of 0.778 with an adjusted H-L test statistic of 0.85.
|
|
The potential for an additional influence of an enhanced interaction effect between race and gender was explored. While keeping both gender and race variables eligible for model entry, a new race-gender interaction term was added to the logistic regression analysis. Holding all other risk factors constant, this race-gender interaction term was not statistically significant predictor of OM. That is, there was no improvement in model performance when adding the designations male Caucasian, male non-Caucasian, female Caucasian, and female non-Caucasian to any of the three annual final study models (p > 0.50). Therefore, the impact of being female and being non-Caucasian upon risk of adverse outcome is related to these main race and gender effects.
Beyond the impact associated with being either female or non-Caucasian alone, there is no additional influence on risk-adjusted OM associated with being both female and non-Caucasian (based on using an interaction term in the analysis). Although the risk-adjusted OM rates were greatest for non-Caucasian females and least for Caucasian males, these categories were used for clinically relevant comparative purposes only.
Multivariate risk-adjusted OM varied be gender and race (p < 0.05). Risk-adjusted OM rates (with 95% confidence intervals) were 3.996% (3.896% to 4.099%) for females, and 3.236% (3.171% to 3.302%) for males. Risk-adjusted OM rates were 3.901% (3.704% to 4.105%) for non-Caucasians, and 3.286% (3.232% to 3.340%) for Caucasians. (See Fig 1.) Risk-adjusted OM rates varied by the four patient subgroups as follows: male Caucasians 3.236% (3.167% to 3.304%), male non-Caucasians 3.822% (3.561% to 4.098%), female Caucasians 3.994% (3.888% to 4.103%), and female non-Caucasians 4.741% (4.392% to 5.109%). (See Fig 2.)
|
|
|
|
|
Comment |
|---|
|
TopAbstractIntroductionPatients and methodsResultsCommentReferences |
|---|
Many studies have reported that radical variations exist in the application of diagnostic and therapeutic procedures for coronary artery disease including cardiac catheterization acute infarct intervention with lytic therapy, angioplasty, and bypass surgery [16–18]. The literature correlating race with outcomes after such interventions is less clear. Oberman and Cutter demonstrated that whites are more likely to have CABG than blacks, but that the two racial groups had similar survival rates after bypass surgery [19]. Maynard and colleagues showed that the black race is correlated with poor survival with medical therapy of coronary artery disease, but not with surgical therapy [20]. Data from the National Hospital Discharge Summary shows that black men with acute myocardial infarction have angioplasty and CABG less often than white men with myocardial infarction. Furthermore, even though this diagnosis was more common in black than in white women, black women had significantly lower rates of cardiac catheterization and CABG [18]. For over 10,000 patients reported in the National Hospital Discharge Summary in 1997, CABG and coronary angioplasty were performed most often in white men, next in black men, then in white women, and least often in black women even when controlling for other risk factors. Similarly, in a study published at Duke University in 1997, blacks with coronary artery disease were significantly less likely than whites to undergo coronary revascularization, especially bypass surgery. The differences in treatment were most pronounced among those patients predicted to benefit most from revascularization, suggesting that coronary revascularization was underused in blacks [21].
Medicare data also reveal striking differences in the application of CABG procedures, the overall national rate being 25.6 per 100,000 population, but 27.1 for whites and 7.6 for blacks. Interestingly, gender differences in the application of CABG appear to be less striking. In whites, the rate is 40.4 for men, 16.2 for women. In blacks, the rate is 9.3 for men and 6.4 for women. The Medicare study also concluded that there are geographic differences in the rates of CABG applications [22]. Information obtained from the Veterans Administration demonstrates that regionalization of services (cardiac catheterization, angioplasty, and CABG) results in an increased likelihood of having a procedure performed, especially in elderly and African American patients [23]. Although there is less information related to other racial subgroups, non-Caucasian women in California (Asian and Latino) received CABG less often than men, and minority patients less often than whites [24]. The reason for these racial differences are unclear but may be associated with a complex combination of related to environment, biology, and genetic predisposition.
We therefore sought to determine whether non-Caucasian patients undergoing CABG-only (based on the racial characteristic variable for 8.5% of procedures entered in the STS Database) have different outcomes than do Caucasian patients. In the process of our analysis, we discovered a statistically significant, although weak association between race and gender (women having CABG are more likely to be non-Caucasian than are men). This finding prompted us to further analyze the patients based on race-gender subgroups.
A preliminary univariate analysis of this large population of patients demonstrated striking differences in the risk profiles of the four separate subgroups (male Caucasian, male non-Caucasian, female Caucasian, and female non-Caucasian). The major study risk factors (race and gender) were analyzed separately and in combination, holding all other patient risk factors constant. Advanced age, severity of coronary artery disease (left main and triple vessel), nonelective operation, reoperation, renal failure, and chronic obstructive lung disease are each known to independently affect OM and CABG in an adverse fashion. All of these attributes were controlled in our multivariate analysis of race and gender.
Less certain are the effects of body size and of the use of the left internal mammary artery as a bypass graft. The tiered classification system established for BMI did not enter any of the baseline multivariate models. This BMI classification, although viewed as clinically relevant, did not appear to have an independent impact on risk of operative death holding all other factors constant. To account for body morphology, therefore, body surface area was used in the standardized STS risk modeling approach.
A problem with data completeness and quality was also encountered when we attempted to analyze the effects of left internal mammary artery grafting on outcome. In 24.3% of the CABG records, the IMA fields were missing, or completed with inconsistent data. For IMA use, we could not determine a satisfactory method of dealing with missing data. Due to an observed bias related to the member sites either completing or not completing these sections of the data form, the IMA variable was deleted from the remainder of the analysis.
We were able to conclusively demonstrate that non-Caucasian race is a multivariate predictor of operative mortality after CABG-only, holding all other patient risk factors constant. Female gender had a similar negative impact on operative outcome. Although there were significantly more non-Caucasian patients in the female group than in the male group (confirming an univariate relationship between race and gender), we could not demonstrate that the combination of race and gender interaction multivariately produced any additional impact beyond these two main attributes.
Across all 3 years, the addition of the race and gender variables to the baseline multivariate risk models did not result in uniform changes to the predictor variable test. Thus, the relationship between race and gender variables with other predictor variables appears to be complex. Further research may be needed to comprehensively identify the diverse set of factors influencing the differential 30-day operative death rates for non-Caucasian and female subgroups.
Risk stratification by commonly accepted clinical subgroups allowed for direct comparison by race and by gender. Both female gender and non-Caucasian race have significantly higher mortality rates among the lower- risk clinical subgroups. As predicted risk increases, the observed differences in 30-day operative mortality rates do not appear to be influenced by race or gender.
As with all other national databases, The Society of Thoracic Surgeons National Adult Cardiac Database faces inherent data completeness and quality challenges. The data collection form is complex and the data entry may be performed retrospectively. Second, the assigned risk estimate cannot easily be calculated remotely at the patient bedside to facilitate the clinical decision-making process. Finally, this database is composed largely of male Caucasian patients. Still, our analysis, which mandated exclusion of those records not containing a minimum set of hard data points (related to age, date of surgery, gender, and procedure type) yielded information concerning the effects of face on CABG outcome, and supplements the existing literature on gender.
Almost a decade ago, when the high operative mortality rates for women undergoing CABG first came under scrutiny, a review of the literature led Dr Janet Bickell to conclude that women were less likely than men to be referred for CABG if their risk was low, but equally likely to be referred if their risk for operative mortality was high. She postulated that the "referral bias" being observed in women was actually a reflection of more appropriate treatment for women than for men [25]. The present analysis indicates that the same principles may apply to non-Caucasian patients. However, our conclusions that non-Caucasian and female patients have worse risk-adjusted outcomes after CABG are not to be interpreted that these groups should have bypass surgery less often. Rather, they should challenge us to further investigate the basis of these discrepancies in the future, as well as to evaluate other outcomes of care (such as morbidity, length of stay, and cost) for these special "at risk" populations [26].
There are several limitations of this study. First, the non-Caucasian group was not further stratified. Since coronary artery disease is the leading cause of death in African-Americans, an analysis of this particular subgroup should be undertaken. Second, there was no designation for coding "diffuse coronary artery disease," a condition that may be more prevalent in one racial subgroup than in another. Even more importantly, we did not have access to those parameters that would have allowed an analysis of the socioeconomic status of each patient, a factor that may be a powerful determinant of outcome [27]. Additionally, the inability to evaluate the impact of use of the left internal mammary artery as a bypass graft conduit upon operative mortality is a major limitation of our analysis. Clearly, there may be complex interactions between other patient risk characteristics and processes of care with CABG outcomes that we were unable to evaluate using the current data available. The current efforts of the STS Database Committee (both to simplify the data collection form and prioritize the data capture efforts) should have a major impact on improving data quality and completeness. Thus, more complete and comprehensive data will be available to facilitate future studies in this area.
|
References |
|---|
|
TopAbstractIntroductionPatients and methodsResultsCommentReferences |
|---|
This article has been cited by other articles:
|
|
 |
|
  W. A. Cooper, V. H. Thourani, R. A. Guyton, P. Kilgo, O. M. Lattouf, E. P. Chen, C. D. Morris, J. D. Vega, T. A. Vassiliades Jr, and J. D. Puskas Racial Disparity Persists After On-Pump and Off-Pump Coronary Artery Bypass Grafting Circulation, September 15, 2009; 120(11_suppl_1): S59 - S64. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 R. Matyal Newly Appreciated Pathophysiology of Ischemic Heart Disease in Women Mandates Changes in Perioperative Management: A Core Review Anesth. Analg., July 1, 2008; 107(1): 37 - 50. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 S. H. Han, J. H. Bae, D. R. Holmes Jr, R. J. Lennon, E. Eeckhout, G. W. Barsness, C. S. Rihal, and A. Lerman Sex differences in atheroma burden and endothelial function in patients with early coronary atherosclerosis Eur. Heart J., June 1, 2008; 29(11): 1359 - 1369. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 C. N. Bairey Merz, L. J. Shaw, S. E. Reis, V. Bittner, S. F. Kelsey, M. Olson, B. D. Johnson, C. J. Pepine, S. Mankad, B. L. Sharaf, et al. Insights From the NHLBI-Sponsored Women's Ischemia Syndrome Evaluation (WISE) Study: Part II: Gender Differences in Presentation, Diagnosis, and Outcome With Regard to Gender-Based Pathophysiology of Atherosclerosis and Macrovascular and Microvascular Coronary Disease J. Am. Coll. Cardiol., February 7, 2006; 47(3_Suppl_S): S21 - S29. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
A. N. Trivedi, T. D. Sequist, and J. Z. Ayanian Impact of Hospital Volume on Racial Disparities in Cardiovascular Procedure Mortality J. Am. Coll. Cardiol., January 17, 2006; 47(2): 417 - 424. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
A. Zacharias, T. A. Schwann, C. J. Riordan, S. J. Durham, A. Shah, and R. H. Habib Operative and Late Coronary Artery Bypass Grafting Outcomes in Matched African-American Versus Caucasian Patients: Evidence of a Late Survival-Medicaid Association J. Am. Coll. Cardiol., October 18, 2005; 46(8): 1526 - 1535. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 D. M. Shahian, E. H. Blackstone, F. H. Edwards, F. L. Grover, G. L. Grunkemeier, D. C. Naftel, S. A.M. Nashef, W. C. Nugent, and E. D. Peterson Cardiac Surgery Risk Models: A Position Article Ann. Thorac. Surg., November 1, 2004; 78(5): 1868 - 1877. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
B. M. Rothenberg, T. Pearson, J. Zwanziger, and D. Mukamel Explaining disparities in access to high-quality cardiac surgeons Ann. Thorac. Surg., July 1, 2004; 78(1): 18 - 24. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 B. Perek, M. Jemielity, and W. Dyszkiewicz Why Are the Results of Coronary Artery Bypass Grafting in Women Worse? Asian Cardiovasc Thorac Ann, December 1, 2003; 11(4): 293 - 298. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
C. R. Bridges Cardiac surgery in African Americans Ann. Thorac. Surg., October 1, 2003; 76(4): S1356 - 1362. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
A. L. W. Shroyer, L. P. Coombs, E. D. Peterson, M. C. Eiken, E. R. DeLong, A. Chen, T. B. Ferguson Jr, F. L. Grover, and F. H. Edwards The society of thoracic surgeons: 30-day operative mortality and morbidity risk models Ann. Thorac. Surg., June 1, 2003; 75(6): 1856 - 1865. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 N. A. Nussmeier, M. R. Marino, and W. K. Vaughn Hormone replacement therapy is associated with improved survival in women undergoing coronary artery bypass grafting J. Thorac. Cardiovasc. Surg., December 1, 2002; 124(6): 1225 - 1229. [Abstract] [Full Text]
|
|
|
|
|
|
J. S. Rumsfeld, M. E. Plomondon, E. D. Peterson, M. G. Shlipak, C. Maynard, G. K. Grunwald, F. L. Grover, and A. L. W. Shroyer The impact of ethnicity on outcomes following coronary artery bypass graft surgery in the Veterans Health Administration J. Am. Coll. Cardiol., November 20, 2002; 40(10): 1786 - 1793. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
R.-K. R. Chang, A. Y. Chen, and T. S. Klitzner Female Sex as a Risk Factor for In-Hospital Mortality Among Children Undergoing Cardiac Surgery Circulation, September 17, 2002; 106(12): 1514 - 1522. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 C. Romualdi, D. Balding, I. S. Nasidze, G. Risch, M. Robichaux, S. T. Sherry, M. Stoneking, M. A. Batzer, and G. Barbujani Patterns of Human Diversity, within and among Continents, Inferred from Biallelic DNA Polymorphisms Genome Res., April 1, 2002; 12(4): 602 - 612. [Abstract] [Full Text] [PDF]
|
|
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
| 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 |
