The effect of race on coronary bypass operative mortality
Author + information
- Received April 6, 2000
- Revision received June 9, 2000
- Accepted July 20, 2000
- Published online November 15, 2000.
Author Information
- Charles R Bridges, MD, ScD, FACC∗,* (cbridges{at}mail.med.upenn.edu),
- Fred H Edwards, MD, FACC†,
- Eric D Peterson, MD, MPH, FACC‡ and
- Laura P Coombs, PhD§
- ↵*Reprint requests and correspondence: Dr. Charles R. Bridges, Department of Surgery, Fourth Floor, Silverstein Building, Hospital of the University of Pennsylvania, 3400 Spruce Street, Philadelphia, Pennsylvania 19104
Abstract
OBJECTIVES
The study was done to determine whether race is an independent predictor of operative mortality after coronary artery bypass graft (CABG) surgery.
BACKGROUND
Blacks are less frequently referred for cardiac catheterization and CABG than are whites. Few reports have investigated the relative fate of patients who undergo CABG as a function of race.
METHODS
The Society of Thoracic Surgeons National Database was used to retrospectively review 25,850 black and 555,939 white patients who underwent CABG-alone from 1994 through 1997. A multivariate logistic regression model was developed to determine whether race affected risk-adjusted operative mortality.
RESULTS
Operative mortality was 3.83% for blacks versus 3.14% for whites (unadjusted black/white odds ratio [OR] 1.23 [1.15–1.31]). Blacks were younger, more likely female, hypertensive, diabetic and in heart failure. Nonetheless, the influence of these and other preoperative risk factors on procedural mortality was quite similar in black and white patients. After controlling for all risk factors, race remained a significant independent predictor of mortality in the multivariate logistic model (adjusted black/white OR 1.29 [1.21, 1.38]). Proportionately, these differences were greatest among lower-risk patients. The race-by-gender interaction was significant (p < 0.05). The unadjusted mortality for black men, 3.30% and white men, 2.64% differed significantly (p < 0.05), whereas for women there was no difference (black, 4.49%; white 4.41%).
CONCLUSIONS
Black race is an independent predictor of operative mortality after CABG except for very high-risk patients. The difference in mortality is greatest for male patients and, though statistically significant, is small in absolute terms. Therefore, patients should be referred for CABG based on clinical characteristics irrespective of race.
Various studies have shown that black patients are less frequently referred for cardiac catheterization (1–5), percutaneous transluminal coronary angioplasty (PTCA) (1,2,4–13) and coronary artery bypass graft surgery (CABG) (1,2,4–8,10–14) than white patients. In general, these differences cannot be accounted for on the basis of disease severity (3,4,7–9,13,14) or the availability of health insurance or socioeconomic status (2–4,6,8,13). Racial bias on the part of physicians has been implicated as one possible contributing factor (3). It is also worth noting that this racial disparity in the use of revascularization procedures has been associated with lower long-term survival among blacks (7). These observations galvanize the view that efforts need to be made to encourage wider use of revascularization procedures in blacks to address this apparent underutilization. As part of this effort, patients and their physicians should understand the unique risks and benefits of these procedures among black patients. However, to date, there has been relatively limited information on the influence of race on the acute outcomes of bypass surgery.
The Society of Thoracic Surgeons National Cardiac Surgery Database, commonly called the STS Database, provides a unique opportunity to address this important question with statistical rigor and power. This report will analyze the national CABG surgical results taken from patients entered into the STS Database from 1994 through 1997. We sought to answer the following questions: Do blacks have poorer acute outcomes following CABG compared with whites? Are the important preoperative predictors of outcomes similar in black and white patients? Do blacks present later or more emergently for bypass surgery than whites? Is race a significant independent predictor of CABG operative mortality after accounting for other known risk factors?
Materials and methods
The population is taken from patients who underwent CABG alone without concomitant procedures, registered in the STS Database from 1994 through 1997. The data represent 47 states (Idaho, Vermont and Maine are not represented), and 5 Canadian provinces (Alberta, Manitoba, Nova Scotia, New Brunswick and Quebec). To be included in the study, patients had to be at least 20 years old and recorded as either black or white on their data collection form. The final analysis included 581,789 patients, of whom 25,850 were black and 555,939 were white.
Patient characteristics were tabulated by race (Table 1), and the differences in proportions for categorical variables were tested using chi-square tests; mean differences for continuous variables were tested using t tests. Overall rates of operative mortality (defined as “mortality occurring at any time in the hospital during the same admission or within 30 days of surgery”), procedural complications and length of stay for blacks and whites were also examined (Table 2). Complications included the following: central neurologic deficit persisting more than 72 h; pulmonary insufficiency requiring ventilatory support; operative re-intervention for bleeding; acute postoperative renal insufficiency; and deep sternal infection involving muscle, bone, and/or mediastinum. Univariate operative mortality rates by risk factor for blacks and whites were also compared (Table 3).
Patient Characteristics by Race
CABG Mortality, Complications, and Length of Stay by Race
Comparison of Operative Mortality for Specific Subpopulations
Next, a logistic regression model of CABG operative mortality was developed using stepwise logistic regression and including all the risk factors as the set of possible predictors. Given the large sample size, independent predictors of mortality were retained in the model if their significance was <0.001. Race was then added to the model to determine whether it had any predictive ability after adjusting for the risk factors already in the model.
Interactions between race and other risk factors were examined in two ways. First, interactions between race and risk factors of interest (e.g., gender, diabetes, hypertension, cerebrovascular [CVA], age, reoperation, shock, obesity) were added to the logistic model to determine their significance. Alternatively, the risk factors from the model developed on all patients were used to model white patients and black patients separately. A comparison of the odds ratio (OR) for the two groups made any differences between blacks and whites more apparent (Table 4).
Multivariate Logistic Regression Model by Race
Finally, the relationship between preoperative risk and race was examined more closely. Specifically, black patients and white patients were classified into decile groupings based on their preoperative risk scores (obtained from the multivariable model with race excluded). Observed mortality rates for blacks and whites in each risk group were then calculated. Graphs of observed mortality by risk decile grouping are presented in Figure 1. Graphs of the black/white ORs using the same decile groupings are presented in Figure 2. To determine whether the difference in mortality rate between blacks and whites was constant across risk score, the race-by-risk interaction was tested for significance using logistic regression.
Race by risk interaction. Observed operative mortality is shown for black and white patients as a function of the preoperative risk group.
Comparison of the black:white operative mortality OR for different preoperative risk groups.
Results
General demographics
A comparison of clinical preoperative risk factors for blacks and whites (Table 1) shows significant differences between the two groups. Because of the large number of patients in this analysis, nearly all univariate risk factors differ significantly as a function of race (p < 0.01). Several differences are striking, however. In general, black CABG patients were more likely to be female than were white patients (44.4% vs. 27.9%, p < 0.001). Black patients were also younger, having an average age of 61.5 years versus 64.8 years for whites. Black patients were more likely to have hypertension (80.5% vs. 62.4%), diabetes (43.8% vs. 27.8%), morbid obesity (15.4% vs. 11.4%) and renal disease (6.4% vs. 2.3%) than whites. However, we did not find strong evidence that blacks were referred for bypass later or in more emergent states than whites. Specifically, blacks were less likely to have their cases done under urgent, emergent or salvage conditions and had less New York Heart Association (NYHA) class IV symptoms at the time of bypass. Additionally, black patients did not appear to have more severe coronary disease anatomy or markedly worse left ventricular function (Table 1).
Operative outcomes by race
Table 2 provides the acute outcomes of bypass surgery for black patients and white patients. The operative mortality was 3.83% for blacks and 3.14% for whites, p < 0.001, unadjusted black/white (B/W) OR 1.23 [1.15–1.310]. Similarly, blacks had higher rates of postoperative complications, including stroke (B/W OR 1.36 [1.25, 1.48]), reoperation for bleeding (B/W OR 1.10 [1.02–1.20], prolonged ventilation (B/W OR 1.34 [1.28, 1.40]), renal failure (B/W OR 1.42 [1.34, 1.51]) and deep sternal wound infection (B/W OR 1.22 [1.05, 1.40]). Postoperative length of stays were also slightly longer for black patients (Table 2).
Table 3 shows the operative mortality for blacks and whites among important clinical subsets. The operative mortality was higher for blacks in most risk categories. For one important risk factor, however, blacks had a significantly lower operative mortality. The operative mortality for blacks with renal failure (7.6%) was lower than the operative mortality for whites with renal failure (10.5%), p = 0.0003.
Risk model comparisons
Using stepwise logistic regression, we identified 27 preoperative characteristics associated with operative mortality. A summary of the results of the logistic regression model appears in Table 4. The risk factors in Table 4 have been arranged according to their contribution to the predictive power of the overall model (as assessed by the value of the Wald chi-square statistic). Thus, age, reoperation, the need for preoperative intraaortic balloon pump (IABP) and ejection fraction emerge as the strong multivariate predictors of operative mortality. The overall ability of this model to discriminate patients who would live from those who would die was excellent (area under the ROC curve or C-statistic = 0.78).
To test whether race was an important independent predictor of operative mortality, we added race to our overall multivariable model. After controlling for the 27 other known risk factors, blacks had 29% higher operative mortality risk compared with whites (adjusted B/W OR 1.29 [1.21, 1.38]). Thus, patient race is an independent risk factor for bypass mortality even after accounting for other differences in patient clinical profiles.
While race is an independent risk factor for procedural outcomes, other known preoperative mortality risk factors among blacks and whites were strikingly similar (Table 4). Specifically, when the multivariable model was applied in black patient and white patient subsets separately, the directionality and magnitude of the ORs for these risk factors were quite similar in blacks and whites (Table 4). We also more formally tested for race by risk factor interactions in the overall patient population. Of the interactions examined, only race by gender was significant (p < 0.05), indicating that the difference in mortality between blacks and whites was more pronounced for men than for women. In fact, there was no significant difference in unadjusted mortality between black and white women (4.49% vs. 4.41%; p = NS).
Finally, the relationship between race and overall mortality risk was examined. Specifically, black patients and white patients were grouped into equal-sized deciles by their calculated operative mortality risk (based on the multivariable model). As demonstrated in Figure 1, blacks had slightly higher observed mortality across all risk deciles. When examined as risk ratios (Fig. 2), however, it becomes clear that B/W mortality differences are greatest among those patients at lowest predicted operative risk. For example, in Decile 1 (lowest-risk patients), the B/W OR for observed mortality was 1.83, whereas among those in Decile 10 (highest risk) this observed B/W mortality risk was only 1.03. In fact, a formal test of race by risk interaction was significant (p < 0.05), indicating that the difference in mortality rates for blacks and whites is not the same across risk groups.
Discussion
We used the Society of Thoracic Surgery National Database to study the effect of race on CABG operative mortality for 555,939 white and 25,850 black patients who underwent CABG from 1994 through 1997. Our analysis has indicated that blacks have a higher operative mortality after CABG than white patients with similar risk profiles. This increase in the relative risk of CABG among blacks, however, is modest in absolute terms (3.83% vs. 3.14%) and is concentrated among males and those with low preoperative risk profiles. In fact, black women and white women have equal unadjusted mortality risk, as do patients with high operative risk.
Comparison with previous studies
Prior to the present study, there have been relatively few published investigations of the effect of race on CABG operative mortality or postoperative long-term survival (15–21). Gray et al. (21) reviewed the records of 3,113 white and 115 black patients who underwent CABG at Cedars-Sinai Medical Center from 1984 through mid-1994. The operative mortality was 5.5% for blacks and 4.6% for whites, p = 0.48. Although there was no significant difference in operative mortality in this study, a multivariate analysis was performed only for long-term survival (21). Higgins et al. (19) analyzed the results of 2,282 white and 494 black patients undergoing CABG from 1990 to 1996 at Henry Ford Hospital in Detroit. The overall operative mortality was 5.5% for black patients and 2.5% for white patients (p < 0.003). However, using multivariate logistic regression, race was not a significant predictor of mortality (19).
Maynard et al. (18) using the Coronary Artery Surgery Study (CASS) database found no statistically significant difference in the five-year survival of patients treated surgically as a function of race. Taylor et al. (20) also using the CASS database, examined the records of 22,714 white and 571 black patients who underwent angiography between 1974 and 1979. Long-term (>10-year follow-up) survival for surgically treated patients was significantly lower for black patients than for white patients, with black race emerging as a significant adverse predictor of long-term survival in a multivariate Cox proportional hazards model.
The search for alternative risk factors
The reasons for the higher operative mortality observed in black patients in this study are not clear. It is difficult to understand how race per se should have a significant impact on operative mortality once all known risk factors have been considered in a multivariate model. Our results suggest several possible hypotheses. First, we found that the average age of the black CABG patients in our population (61.5 years) is significantly lower than the average age of the white patients (64.8 years; p = 0.0001). Yet if a multivariate analysis is performed excluding race as a variable, the predicted mortality for black patients is 3.12%, whereas the predicted mortality for whites is 3.14%. Despite younger age, black patients have similar predicted mortality independent of race. Thus, black patients were generally younger and slightly sicker for their age than were white patients. Therefore, race in this study may be a marker for patients with more rapid rate of progression of coronary artery disease (CAD), in particular, and vascular disease, in general.
This finding of potentially more aggressive atherosclerosis in black patients is also consistent with the observed higher rates of cerebrovascular disease, hypertension, and renal disease in blacks relative to whites. The factors that correlate with vascular disease progression, such as elevations in C-reactive protein, are likely to be more prevalent in these patients (22). The associated inflammatory milieu is likely to exacerbate end organ function to a greater extent after cardiopulmonary bypass and to increase the risk of cardiac, neurological, pulmonary and renal complications after surgery, as we observed (23)(Table 2). These and other serum factors may also be important in determining CABG operative mortality, considering that the removal of inflammatory mediators using conventional and modified ultrafiltration has recently been shown to improve pulmonary, hemostatic and cardiac function after cardiac surgery in adults (24). Endothelial dysfunction has also been raised (25) as a risk factor for hypertension, and perhaps it could lead to complications in patients after CABG.
Blacks, in the absence of angiographic evidence of CAD, have a significant reduction in endothelium-independent coronary flow reserve compared with whites (25). Additionally, the depressed coronary vascular relaxation in response to both acetylcholine and adenosine in black patients has been associated with the higher prevalence of left ventricular hypertrophy (LVH) seen in blacks (26). For a given degree of hypertension, black patients are more likely to have significant LVH (27). In this analysis, although LVH was not specifically studied, we did, however, note that racial differences in operative mortality appeared to be limited to those patients with hypertension (Table 3). Specifically, although hypertensive blacks had significantly higher operative mortality than whites, race was not a significant predictor of surgical outcome among normotensive patients. The LVH may be more prevalent in the black patients studied with hypertension, perhaps increasing the incidence of postoperative arrhythmias or compromising myocardial protection during cardiopulmonary bypass.
Finally, as previous researchers have indicated, the study of racial differences in care and outcome is complex (28). Dichotomous variables based on skin tone do not capture the ethnic and biological diversity inherent among humans. Additionally, race is often a surrogate marker for underlying socio-economic, educational, behavioral and cultural differences. Unfortunately, these factors were not collected within our study, but the potential influence of these and other nonbiologic factors on both the short- and long-term outcomes of CABG should be the focus of future studies.
Conclusions
In summary, the finding that race is an independent predictor of operative mortality after CABG should prompt a more careful search for additional risk factors for CABG operative mortality. The unexpected finding of a significant race/gender interaction is important. In the general population, female gender is an independent predictor of operative mortality after CABG (29,30)(Table 4). Thus, further analysis of the differential effects of race and gender on outcome, as functions of gender and race, respectively, is required. Although the operative mortality after CABG is significantly higher for blacks than for whites, the differences are relatively small in absolute terms. Therefore, patients should continue to be referred for CABG based on established clinical criteria, independent of race.
- Abbreviations
- BSA
- body surface area
- B/W
- black/white
- CABG
- coronary artery bypass graft
- CAD
- coronary artery disease
- CVA
- cerebrovascular accident
- CVD
- cerebrovascular disease
- COPD
- chronic obstructive pulmonary disease
- IABP
- intraaortic balloon pump
- LVH
- left ventricular hypertrophy
- MI
- myocardial infarction
- NYHA
- New York Heart Association
- OR
- odds ratio
- PVD
- peripheral vascular disease
- Received April 6, 2000.
- Revision received June 9, 2000.
- Accepted July 20, 2000.
- American College of Cardiology
References
- ↵
- Gillum R.F.,
- Gillum B.S.,
- Francis C.K.
- ↵
- ↵
- ↵
- Blacks in the Coronary Artery Surgery Study,
- Maynard C.,
- Fisher L.D.,
- Passamani E.R.,
- Pullum T.
- Laouri M.,
- Kravitz R.L.,
- French W.J.,
- et al.
- ↵
- ↵
- ↵Higgins RS, Paone G, Borzak S, Jacobsen G, Peterson E, Silverman NA. Effect of payer status on outcomes of coronary artery bypass surgery in blacks. Circulation 1998;98:II46–9; Discussion II49–50.
- ↵
- Taylor H.A.,
- Mickel M.C.,
- Chaitman B.R.,
- Sopko G.,
- Cutter G.R.,
- Rogers W.J.
- ↵
- ↵
- Liao J.K.
- ↵
- ↵Kiziltepe U, Uysalel A, Corapcioglu T, Akan H, Akalin H. Effects of combined conventional and modified ultrafiltration in adult patients undergoing cardiac surgery. Ann Thorac Surg 2000 (in press).
- ↵
- Houghton J.L.,
- Prisant L.M.,
- Carr A.A.,
- Flowers N.C.,
- Frank M.J.
- ↵
- ↵
- Chapman J.N.,
- Mayet J.,
- Chang C.L.,
- Foale R.A.,
- Thom S.A.,
- Poulter N.R.
- ↵
- Laviest T.A.
- ↵