The Effect of Ethnicity on Survival in Male Veterans Referred for Electrocardiography and Treadmill Testing

Background  Ethnic differences in the relationship between access to health care and survival are difficult to define because of many confounding factors, such as socioeconomic status and baseline differences in health. Because the Veterans Affairs health care system offers health care largely without financial considerations, it provides an ideal setting in which to identify and understand ethnic differences in health outcomes. Previous studies in this area have lacked clinical and cardiovascular data with which to adjust for baseline differences in patients' health.

Methods  Data were collected from consecutive men referred for resting electrocardiography (ECG) (n = 41 087) or exercise testing (n = 6213) during 12 years. We compared ethnic differences in survival between whites, blacks, and Hispanics after considering baseline differences in age and hospitalization status. We also adjusted for electrocardiogram abnormalities and cardiac risk factors, exercise test results, and cardiovascular comorbidities.

Results  White patients tended to be older and had more baseline comorbidities and cardiovascular interventions when they presented for testing. White patients had increased mortality rates compared with blacks and Hispanics. In the ECG population, after adjusting for demographics and baseline electrocardiogram abnormalities, Hispanics had improved survival compared with whites and blacks. In the exercise test population, after adjusting for the same factors, as well as adjusting for the presence of cardiovascular comorbidities, cardiac risk factors, and exercise test findings, Hispanics also exhibited improved survival compared with the other 2 ethnicities. There were no differences in mortality rates between whites and blacks.

Conclusion  Our findings demonstrate that the health care provided to veterans referred for routine ECG or exercise testing is not associated with poorer survival in ethnic minorities.

ACCESS TO high quality health care for all segments of society is an important issue in the United States. Several studies^1-3 suggest that blacks and other ethnic minorities fail to receive health care that is equivalent to that received by their white counterparts. For example, some studies^4-6 have reported ethnic differences in the use of cardiac procedures. However, "use rates" alone are not indicators of quality; discrepancies in use rates may reflect differences in need, access to care, overuse in one group, or underuse in another. The important question is not whether ethnic minorities receive fewer treatments, but whether they receive health care that unfavorably affects their survival.

The Department of Veterans Affairs (VA) offers a distinct advantage for the study of ethnic differences in the delivery and outcomes of health care. Treatment within the VA system does not depend on a patient's ability to pay, and there are no financial incentives for physicians to perform procedures. Therefore, whites, blacks, Hispanics, and other ethnicities should theoretically receive equal access to health care. Also, because of the financial restrictions for eligibility for VA care, most veterans are in a similar socioeconomic class, and few can afford other sources of health care. Jha and colleagues⁷ recently reported better survival in black patients compared with white patients admitted to VA hospitals. They analyzed a large administrative database and studied the outcomes of patients after admission. The authors explored several possible reasons for ethnic differences in mortality rates, including severity of illness and other comorbidities at admission, but these explanations could not be evaluated adequately because the database lacked detailed clinical information.⁸

We performed the present analysis to assess whether ethnicity affects survival in male veterans. Using our database of men referred for resting electrocardiography (ECG) or exercise testing during 12 years, we compared differences in population demographics, clinical history, cardiac risk factors, electrocardiogram abnormalities, exercise test responses, and history of cardiac procedures and their relation to long-term survival among whites, blacks, and Hispanics. Electrocardiography and exercise test results provide important findings of the extent of cardiovascular disease at baseline, and the exercise test is the best measure of baseline functional status. Given that death is most often due to cardiovascular disease among veterans within the age range studied (aged 22-93 years; mean, 55 ± 11 years),⁹ consideration of these measures as confounders is appropriate. In previous outcome studies assessing health care and ethnicity, these cardiac data have been unavailable to adjust for baseline differences in cardiovascular health.

From June 1, 1987, to July 25, 2000, 42 417 consecutive men were referred for resting ECG and another 6213 for exercise testing. On arrival for the procedure, patients were classified ethnically as white, black, Hispanic, or other (including Asian, Pacific Islander, and Native American). Classification was accomplished through patient self-categorization. Because of the ethnic heterogeneity within the "other" category, we excluded these patients from the analysis.

To evaluate our population samples for referral bias, we compared their demographics with national, state, and VA Palo Alto Health Care System ethnic data (Table 1).

All patients undergoing routine ECG at the VA Palo Alto Medical Center were included in the analysis. Computerized 12-lead resting ECG findings were digitally recorded on the Marquette MAC system (GE Marquette, Milwaukee, Wis). Each electrocardiogram was labeled with the patient's social security number, and the storage file included height, weight, date of birth, ethnicity, and location of testing. Only the first ECG results were included for patients with more than 1 study in the database. The 1330 patients with no ethnic data available were excluded from analysis. The electrocardiogram abnormalities were defined according to 12 single-lead criteria of the Marquette ECG analysis program read by a board-certified cardiologist (V.F.F.).

Consecutive patients referred to 2 clinical exercise laboratories (Long Beach VA Medical Center, 1987-1991; and Veterans Affairs Palo Alto Medical Center, 1992-2000) underwent a symptom-limited treadmill test.^10,11 No ethnic data were available for 47 patients, and they were excluded from analysis. The exercise laboratories were directed in a consistent fashion by 2 of us (V.F.F. and J.M.). Patients who were subjects in research protocols were not included in the analysis. Medications were not changed or stopped before testing. A thorough clinical history, list of medications, and cardiac risk factors were recorded prospectively at the time of exercise treadmill testing using computerized forms.^12,13 The forms included standard definitions of clinical conditions and exercise responses.

After careful skin preparation, a standard 12-lead electrocardiogram was obtained before and continuously during exercise and for at least 5 minutes during recovery.¹⁴ Blood pressure was taken manually, and metabolic equivalents (METs) were estimated from treadmill speed and grade. Patients did not perform a cooldown walk, but were placed supine as soon as possible after exercise.¹⁵ No test result was classified as indeterminate.¹⁶ The exercise tests were performed, analyzed, and documented per standard protocol using a computerized database.¹⁷ Normal standards for age-predicted exercise capacity were derived from regression equations published previously¹⁸: predicted peak METs = 18.0 – 0.15(age); and age-adjusted exercise capacity = (achieved METs/predicted METs) × 100. No imaging modality was performed in conjunction with the tests. Both laboratories had affiliations with universities and had academic medical staff with rotating house officers and fellows. All tests were supervised by these individuals or by nurse practitioners; all test results were reviewed by 2 of us (V.F.F. and J.M.).

The Social Security Death Index was used to match all of the patients using name and social security number. The Social Security Death Index was used to ascertain vital status as of December 1999 for the ECG database and July 2000 for the exercise test database. Institutional review board approval was obtained at both institutions, and consent forms were signed by the patients undergoing exercise testing at both institutions before data collection. Institutional review board approval was obtained after ECG data collection at Veterans Affairs Palo Alto Medical Center. After using social security numbers to obtain death data, these numbers and names were removed, and an auto-number was generated to maintain patient confidentiality.

Quality control was safeguarded by data entry using specialized software at the time of testing and then by cross-entry into the VA administrative and clinical databases. These parallel data sets resulted in no more than 5% missing data in any field entered into the Cox proportional hazards regression model. Number Crunching System Software (Kaysville, Utah) was used for all statistical analyses after transferring the data from an Access (Microsoft, Redmond, Wash) database. Outliers were scrutinized and corrected when appropriate at all points of data entry and transference. Demographics and prevalence of selected ECG criteria were compared between ethnicities in the ECG population. In the exercise test population, demographics, cardiovascular comorbidities, cardiac risk factors, history of cardiac interventions, and exercise test results were compared between ethnicities. χ² Analysis was used to compare proportions, and analysis of variance was used to compare means between the 3 ethnic groups. P<.05 was considered significant.

Total (all-cause) mortality was used as the end point for survival analysis. Because the data regarding subsequent interventions were not available for all patients, patients who underwent interventions were not removed from observation. Survival analysis was displayed using Kaplan-Meier curves plotted for the ECG and exercise test populations. Log-rank tests were performed to assess if statistically significant differences in mortality rates were present between ethnicities.

Age, inpatient status, the presence of electrocardiogram abnormalities, and black and Hispanic ethnicities were included in a proportional hazards analysis to determine which variables were independent predictors of mortality. Electrocardiogram abnormalities included in the model were right ventricular hypertrophy, left bundle branch block and ST-inclusive left ventricular hypertrophy, and any other electrocardiogram abnormality (including right bundle branch block, atrial fibrillation, Q waves, premature ventricular contractions, and ST-segment depression).

In the exercise test population, age, inpatient status, presence of cardiovascular disease, cardiac risk factors (including history of hypertension, diabetes mellitus, family history of heart disease, and current smoking habit), age-adjusted exercise capacity, and black and Hispanic ethnicities were included in the proportional hazards analysis. We defined patients with cardiovascular disease as those with a history of myocardial infarction as determined through history or Q wave, coronary bypass surgery, coronary angioplasty, congestive heart failure, peripheral vascular disease, an abnormal exercise test response suggesting coronary artery disease (1.0-mm ST-segment depression, exercise-induced angina, or both), or pulmonary disease.

Comparative ethnic distributions are presented for comparison in Table 1. The ethnic distribution of patients referred for ECG and exercise testing was not statistically different, but minor statistical differences existed between those referred for ECG and the entire population of the VA Palo Alto Health Care System catchment area. The differences between VA health care users in general can be explained by local selection factors (ie, real estate value and cost of living).

Of the 41 087 men referred for ECG with available ethnic data, 31 928 (77.7%) were white, 5020 (12.2%) were black, 2454 (6.0%) were Hispanic, and 1685 (4.1%) were of other ethnicities. The demographics and prevalence of selected resting electrocardiogram abnormalities are described in Table 2. White patients were older and were more likely to be inpatients. The prevalence of any electrocardiogram abnormality (Q waves, bundle branch block, atrial fibrillation, premature ventricular contractions, ST depression or left ventricular hypertrophy with strain, or right ventricular hypertrophy) was also higher among whites compared with Hispanics and blacks. Hispanics had a higher body mass index than whites or blacks. Left ventricular hypertrophy with strain was more than twice as prevalent among blacks compared with whites or Hispanics.

Among the 6166 veterans referred for exercise testing with ethnic data available, there were 4530 whites (73.5%), 754 blacks (12.2%), 576 Hispanics (9.3%), and 306 (5.0%) of other ethnicities. The population characteristics of the 3 ethnic groups are provided in Table 3. Blacks referred for exercise testing were younger than whites and Hispanics. Whites were more likely to have had a myocardial infarction (by history or Q wave) and congestive heart failure than blacks and Hispanics. Digoxin use was more prevalent among whites. Blacks had more hypertension and were more likely to be receiving antihypertensive medications. Whites had higher rates of cardiovascular interventions compared with blacks and Hispanics. The percentage of inpatients did not differ between the 3 ethnic groups.

Table 3 presents the cardiac risk factor profiles in the ethnic groups in the exercise test population. Blacks had a greater prevalence of risk factors, including hypertension, hypercholesterolemia, and obesity, compared with the other ethnic groups. Diabetes mellitus was significantly more prevalent among Hispanics and blacks compared with whites. A history of smoking or smoking at the time of testing was more prevalent among blacks.

No major complications were encountered during testing, although nonsustained ventricular tachycardia occurred in 81 veterans (1.3%). Exercise test results for the 3 ethnicities are presented in Table 4. Whites had a higher prevalence of exercise-induced angina than Hispanics or blacks, and the test was terminated more frequently in whites because of angina compared with the other ethnicities. Exercise-induced ST depression (≥1.0-mm), the most common electrocardiogram abnormality, was significantly more frequent in whites. Abnormal exercise test–induced ST-segment depression was asymptomatic two thirds of the time in all 3 ethnicities.

Blacks attained significantly higher systolic and diastolic blood pressures and mean heart rates during maximal exercise compared with the other 2 ethnic groups. Age-adjusted exercise capacity, as reflected by the percentage of maximal METs achieved, was similar between ethnicities. Patients of all ethnicities gave a good effort as reflected in the mean Borg scale rating at maximal "very hard" effort (rating, 17).

The ECG population had a mean annual mortality of 3.5% during a 12-year follow-up (mean ± SD follow-up, 5.3 ± 4.0 years), and 10 258 (25.0%) died. The mean annual all-cause mortality was 3.7% in whites, 2.6% in blacks, and 2.8% in Hispanics. The Kaplan-Meier survival plot demonstrated that whites had significantly decreased survival compared with the other 2 ethnicities (P<.01) (Figure 1). The assumption of proportional hazards was assessed by plotting the Schoenfeld residuals for each noncensored individual.¹⁹ Because no trends were noted, the model was appropriate. After adjusting for differences in age, hospitalization status, and electrocardiogram abnormalities, Hispanics had a decreased risk of mortality compared with the other ethnicities (Table 5).

The mean annual mortality in the population referred for exercise testing was 3.2% during the follow-up (mean ± SD follow-up, 6.2 ± 3.7 years). This mortality rate was similar to that for the population referred only for resting ECG (3.5% per year). Those who died were significantly older and had a lower mean body mass index than survivors. The mean annual mortality of whites, blacks, and Hispanics was 3.4%, 2.9%, and 2.2%, respectively. The Kaplan-Meier survival plot (not shown) was similar to that in Figure 1.

At the time of exercise testing, whites had more comorbid conditions and had undergone more cardiac interventions than blacks. There was a 10% mean improvement in survival per MET increase in exercise capacity in all 3 ethnicities.

After adjusting for age, the proportional hazards analysis demonstrated that Hispanics had significantly decreased mortality rates compared with the other ethnicities in the ECG (hazard ratio, 0.81) and exercise test (hazard ratio, 0.76) populations (Table 5 and Table 6). These tables also include the results of the Cox proportional hazards regression analysis of the other variables that were significantly and independently associated with death.

Within the VA Palo Alto Medical Center, we found similar rates of major medical insurance coverage among whites (21.3%) and Hispanics (23.5%), with slightly fewer blacks having major medical coverage (18.1%, P = .001). However, few veterans use other health care providers if they are registered in VA clinics, which require no copayment for health services.

The effect of ethnicity on health care and outcomes remains a controversial issue because studies continue to produce conflicting results. Previous research suggests that black patients with coronary disease receive fewer cardiac interventions and have increased mortality rates compared with white patients.¹⁹ Studies^20,21 have also demonstrated that blacks have a poorer prognosis following myocardial infarction compared with whites. In contradiction to these findings, Keil et al²² reported that death from coronary disease is not increased among blacks. Despite a higher incidence of diabetes mellitus and hypertension^23,24 and a lower rate of cardiac procedure use among black patients following myocardial infarction,^24-26 investigations have found that these differences do not translate into increased mortality rates for blacks compared with whites. Further research suggests that ethnic minorities admitted to hospitals for various diagnoses have similar or improved survival compared with their white counterparts in VA²⁷ and non-VA hospitals.²⁸ Studies investigating differences in mortality rates between whites and Hispanics are also inconclusive. Pandey et al²⁹ and Wei et al³⁰ report equal or higher rates of cardiovascular mortality among Hispanics compared with whites, while others show that Hispanics have lower all-cause and cardiovascular mortality rates³¹ and a lower prevalence of myocardial infarction³² than whites.

Adding a unique dimension to the study of the effect of ethnicity on mortality, we have adjusted for baseline ECG results, clinical data, and exercise test findings. No prior study has had such data available, and it is important to adjust for these findings because cardiovascular diseases are the major cause of death in the veteran population. We found that Hispanics referred for ECG or exercise testing at the VA hospitals had improved survival compared with whites and that there were no differences in survival between blacks and whites.

Our study has several other important strengths. The VA Health Care System offers health care largely without financial considerations; therefore, patient financial status should not alter access to medical care. Moreover, financial status should not affect referral for ECG or exercise testing, theoretically eliminating referral bias of patients entering this study. The ethnic distribution of patients referred for ECG and exercise testing paralleled the ethnic demographics of patients receiving care at the VA Palo Alto Health Care System. Although we cannot rule out referral bias, the ethnic demographics of the referral populations are similar to the VA population. In addition, because most veterans are middle to low income earners,³³ any socioeconomic differences between ethnic groups are diminished within the VA.

Through the centralized VA computer system and a Windows-based relational database^12,13 in the exercise laboratory, we had detailed information on patient demographics, cardiac risk factors, comorbidities, and exercise test findings. Moreover, although quality-of-life issues were not addressed, a key component of quality of life, exercise capacity,³⁴ was measured at baseline, and no differences were found between ethnicities. The larger ECG group of more than 30 000 patients allowed us to compare mortality rates and strengthened our findings in the smaller exercise test group. Administrative databases generally do not have detailed clinical information⁸ concerning risk factors and test findings that was available for our study.^35,36

A possible explanation for similar or improved survival in minorities may be baseline differences in the severity of disease. Two studies^37,38 of angiographic data reported that blacks who underwent cardiac catheterization were more likely to have minimal disease than whites. Both studies included few blacks and examined selected populations. Among patients undergoing cardiac catheterization after presenting with acute chest pain, Johnson et al⁶ found that blacks had less severe angiographic disease than whites.

Our findings suggesting a lower prevalence of cardiac interventions among ethnic minorities are consistent with most previous research.⁶ It has been suggested that blacks have decreased access to cardiac procedures. This lack of access could not be explained by socioeconomic status, income, or insurance coverage. Other study findings support these results, including analyses performed at VA hospitals, where financial status should not affect access to care.^4,5 Leape et al³⁹ examined the use of cardiac procedures in ethnic minorities in 13 of 24 hospitals in New York that provide high-technology cardiac services. They found no variations in the rate of use by sex, ethnic group, or payer status among patients treated in hospitals that provide cardiac interventions. However, underuse was significantly greater in hospitals that do not provide these services, particularly among uninsured persons.

The reasons for this inequality of care are largely unknown. Minorities may be less likely to accept recommendations for cardiac procedures. For instance, blacks were less likely than whites to state that they would undergo procedures if their physician recommended them.^40,41 The Coronary Artery Surgery Study³⁸ also found that blacks were 10% more likely than whites to refuse recommendations for coronary artery bypass surgery. Perhaps this is explained by ethnic differences in mortality rates⁴² and survival following coronary artery bypass surgery.^43,44

Although our study was unique in its content, setting, and follow-up, its limitations and the inferences that can be made from the data are acknowledged. We were unable to provide information on cause of death and did not have information on cardiovascular procedures during the follow-up, so we could not remove patients from observation when these occurred. The data do not include clinical or disease severity measures that would affect clinical decision making and outcomes. Some information was self-reported and subject to recall bias. We had no way to assess referral bias except to note that the ethnic percentages referred for the 2 procedures and the ethnic distribution of VA users in general were similar. The differences between our catchment area and the national and state demographics are due to the high cost of property. Last, we had limited data on the use of other health care systems by our patients.

Future studies must address whether other measures of patient outcomes, such as symptoms and health-related quality of life, differ between ethnic groups. Our findings demonstrate that the health care provided to veterans referred for routine ECG or exercise testing at 2 California VA medical centers is associated with similar or improved survival for blacks and Hispanics compared with whites even after adjusting for differences in baseline cardiovascular status.

Corresponding author: Victor F. Froelicher, MD, Cardiology Division, VA Palo Alto Health Care System, 3801 Miranda Ave (Mail Stop 111C), Palo Alto, CA 94304 (e-mail: vicmd@aol.com).

Accepted for publication December 6, 2002.