Race and the decision to refer for coronary revascularization

The effect of physician awareness of patient ethnicity

Study design and patients

The sample was drawn from 1,022 consecutive cardiac catheterizations from 1993 to 1995. Eighty-four patients undergoing repeat catheterization during the study period were excluded from consideration, because the procedure was done to evaluate valvular heart disease, myocardial or pericardial disease. Patients undergoing repeat catheterization (n = 56) in this interval were included if the catheterization resulted in a new treatment decision. A final cohort of 938 catheterizations in 882 patients forms the data base. Clinical data collected included that accessible from the Decentralized Hospital Computer Program data base, as well as data from detailed angiographic reports. Coronary anatomy (location, number and extent of coronary stenoses), LV function, as assessed by either angiography, gated radionuclide ventriculography or echocardiography, and comorbidity were noted for each patient. Data requiring a patient interview or chart review, such as severity of angina or symptoms of congestive heart failure, were not available for this retrospective study. Additional details of coronary anatomy requiring angiographic review, such as diffuseness of the disease and adequacy of target vessels, were also not collected.

Classification of CAD was based on visual interpretation using standard criteria (23,24). Grading of the severity of CAD was in accordance with the American Heart Association’s Ad Hoc Committee for Grading of Coronary Artery Disease (25). The degree of stenosis was quantitated from the moving cineangiogram by visual estimation of the percent reduction in diameter relative to the caliber of the apparently nondiseased adjacent segments, with the range of lumen diameter reductions from 0% (normal coronary segment) to 100% (completely occluded vessel). When left main coronary artery (LMCA) disease was noted, measurements were made with handheld calipers and a magnifying scale. A lumen diameter reduction of ≥50% at the LMCA (equal to a 75% reduction in cross-sectional area) was considered significant. In CAD other than that in the LMCA, stenoses >70% were significant and considered proximal if they were located in the right coronary artery before the acute marginal branch, in the left anterior descending coronary artery (LAD) before the first septal perforator branch or in the left circumflex artery before the obtuse marginal branch. Significant lesions in the major branches of the left coronary artery system (i.e., anterior descending, diagonal, circumflex and obtuse marginal arteries) were measured and recorded.

Left ventricular function was measured by LV angiography in the majority of cases (n = 682). Calculation of ejection fraction by LV angiography was done using single-plane cine left ventriculography in the right anterior oblique projection using the area-length method. Cine angiographic data were processed by Trinity Computing systems (Greenway Plaza, Houston, Texas) and LV ejection fraction ≥50% was interpreted as normal, 35% to 50% as mild LV systolic dysfunction, 20% to 35% as moderate LV systolic dysfunction and <20% as severe LV systolic dysfunction. In 123 patients in whom a ventriculogram was not obtained, functional evaluation of the LV was determined from two-dimensional echocardiography or radionuclide ventriculography. The same quantitative criteria were used to describe LV function on a gated blood-pool study. Echocardiography of LV function was assessed semiquantitatively by experienced echocardiographers and was described as normal or mild, moderate or severe LV dysfunction. Estimation of ejection fraction by this echocardiographic method has been shown to be accurate and reliable when assessed by experienced echocardiographers (26,27).

Comorbidity was defined in the following manner: 1) the presence of clinically significant renal disease was defined by serum creatinine ≥2.0 mg/dl; 2) significant pulmonary disease was identified by pulmonary function tests (when available) and/or by the presence of bronchodilators or methylxanthines on the patient’s computerized medication profile; and 3) diabetes was identified by the presence of either insulin or oral hypoglycemic agents on the patient’s medication profile. Although hypertension is clearly more prevalent in African American than white patients, this variable (in contrast to the other co-morbidities measured) was notspecifically noted, because its presence or absence neither contributes to a decision to revascularize nor influences the choice of the procedure.

All angiographic cases were discussed weekly in a cardiac catheterization conference by a committee consisting of an interventional cardiologist, a cardiothoracic surgeon and a clinical cardiologist. The committee reviewed all angiographic studies from the previous week, including those patients for whom urgent or emergent treatment had been required. Data presented for each patient were strictly limited to clinical factors that might influence the treatment choice. For this reason, the patients’ ethnicity was not presented. The recommendation of this committee was considered as the final treatment recommendation.

Decisions regarding the mode of therapy were based on clinical and angiographic variables. Angiographic variables included the number and location of diseased vessels, the presence or absence of LMCA disease, the extent of disease in the distal vessel, LV systolic function and regional contractile function of the territory supplied by the diseased vessel. The clinical variables considered included the presence or absence of unstable angina, the severity of angina in patients with stable angina treated with medication, objective evidence of ischemia on radionuclide stress perfusion imaging or stress echocardiography and the presence of diabetes mellitus, renal or pulmonary disease. Patients with significant atherosclerotic disease in the LMCA and in all three major coronary arteries were referred for CABG. Most patients with double-vessel disease involving the proximal LAD and proximal right coronary arteries were also referred for CABG. Patients with single-vessel disease technically amenable to a coronary intervention and with objective evidence of ischemia were generally referred for a catheter-based intervention. Although patients with two-vessel disease not involving the LAD were often referred for PTCA, many also were referred for either medical management or CABG depending on the particular anatomic and physiologic findings. These treatment recommendations are consistent with those described for the Coronary Artery Surgery Study (22,23).

Statistical analysis

Statistical analysis involved three steps. First, the demographic and clinical characteristics of white and African American patients undergoing left heart catheterization were compared by using the chi-square test or the Wilcoxon signed-rank test. Second, treatment recommendations in white and African American patients with significant stenosis in one or more coronary arteries were compared by using the chi-square test. Analyses compared recommendations for CABG, PTCA and revascularization by either CABG or PTCA. Third, to adjust for the potential confounding influence of demographic and clinical factors in recommendations, additional logistic regression analyses were performed. Separate analyses were performed for CABG and PTCA. Independent variables in these analyses included age, African American ethnicity, the presence of specific comorbid disease (diabetes, pulmonary or renal disease), LV dysfunction (classified as mild, moderate or severe), the number of coronary arteries with significant stenosis and involvement of specific arteries. The multivariate odds ratio (OR) associated with African American ethnicity was used to estimate the relative rates of CABG and PTCA in African American and white patients.

Ethnic patterns of coronary revascularization

Although the recommendation for treatment is different from utilization, it is a key step in treatment utilization. Only one previous report specifically addressed racial differences in physicians’ recommendation for CABG, and thus could be compared with our study (28). Maynard et al. reported that 46.5% of African American patients, compared with 59.4% of white patients, were recommended for CABG. This highly significant difference paralleled the observed difference in procedure utilization in their cohort. Importantly, the investigators specifically noted that the clinical and angiographic characteristics of the two groups were not sufficiently different to account for the difference in treatment recommendation (28), but no details of the recommendation process were provided.

A review of Medicare and Veterans Affairs data bases (5,9,17)has found significantly greater utilization of CABG and PTCA in white than in African American patients. When Medicare patients were stratified by race and income to adjust for differences in socioeconomic status (5), the racial difference in procedure utilization persisted across all income ranges. When cardiac catheterization and revascularization rates were compared in African American and white veterans after an acute myocardial infarction (10), African Americans were significantly less likely to undergo cardiac catheterization, PTCA or CABG. Although these registry studies provide important information on the rates of procedure utilization, factors guiding utilization were not examined.

Two studies in which patients were stratified by race and disease severity (6,18)reported that African Americans were less likely to undergo CABG at all levels of disease severity, including LMCA disease (6). Of note, the outcome of African American patients was significantly poorer (18). In these and other cohorts (7,15), investigators noted a lesser prevalence of LMCA and three-vessel coronary disease in African Americans.

The contrast between these previously reported patterns of revascularization referral and procedure utilization and the results of our study suggests that race may be an important independent factor for physicians making treatment decisions.

Factors contributing to procedure use

We recognize that the etiology of the racial difference in procedure utilization is both multifactorial and complex. Patient-based biologic factors may contribute to ethnic variation in utilization patterns. Data from several studies have suggested that differences exist in the anatomic distribution or severity of CAD between African Americans and whites (3,7,12,15,16). Although many studies have involved large numbers of patients (7,15,16), the percentage of African Americans included in these studies was quite small, leaving open the possibility that the African American patients in these cohorts consisted of selected subgroups not representative of the African American population as a whole. Importantly, not one study supported the concept that lesser procedure utilization in African Americans could be exclusively attributed to lesser disease burden.

Patient socioeconomic status and culturally based determinants of physician decision-making, such as the level of expertise of the physician caring for the patient (specialist vs. generalist), access of particular groups of patients to different levels of physician expertise, the physician’s perception of the patient, as well as cultural differences that interfere with the patient’s acceptance of physician recommendations (23,28), may also have an impact on treatment recommendations. The role played by each of these factors requires separate study.

Study limitations

Although the results of this retrospective review are highly provocative, there are some study limitations requiring further evaluation of this model. It must be emphasized that our study does not examine how patients are referred for cardiac catheterization—another crucial step in the decision process leading to revascularization. We were able to examine the revascularization decision process only after catheterization. However, other studies scrutinizing patient management after catheterization have found different patterns of procedure utilization (6,9). Other variables, such as the appearance of target vessels, collateralization, the presence of diffuse as opposed to focal lesions, the severity of angina and symptoms of congestive heart failure, which might influence treatment decisions, could not be readily examined in this review. Another study limitation was the inability to assess patients’ reception of the treatment recommendation. Although a similar proportion of African American and white patients (46 vs. 39%, p = NS) underwent the recommended procedure at this Veterans Affairs Medical Center, we have no information on the patients who did not. Finally, this study was not a prospective study with strict physician blinding to patient race; rather, it was a retrospective study taking advantage of a decision process in which patient ethnicity was generally unknown to those making the decision regarding procedure referral. Nonetheless, our findings that the overall revascularization rate did not differ between racial groups and that differences in treatment did not exist when patients with one-, two- or three-vessel disease were compared, in contrast to other reports in the published data, suggests a critical need for additional research on the clinical and psychosocial factors directing physician utilization of specific cardiovascular procedures.