Blacks have a higher prevalence of congestive heart failure (CHF) than do whites (1). Hospitalizations for CHF, as well as readmissions after an initial hospitalization, have also been reported to be more common in black patients (2- 4). However, some of the studies evaluating patients hospitalized with CHF have demonstrated better short-term survival in black patients compared with white patients (2,5- 7). This survival advantage has been observed for both in-hospital mortality and mortality up to one year after discharge. The reasons for this paradox of increased hospitalizations in the face of a better survival are not clear. One reason suggested has been differential access to medical care by race. It has been postulated that reduced access to outpatient care in blacks (because of a greater proportion being uninsured and unable to pay for care) may result in a higher number of hospitalizations for lesser disease severity (2). None of these studies has been able to link the use of outpatient care with hospital use and mortality to explore this possibility (2,5,7).

The Veterans Health Administration (VA) healthcare system offers a unique opportunity to address this issue. VA healthcare utilization databases can be linked with vital status data to create patient-specific longitudinal inpatient, outpatient, and mortality records. In addition, the VA healthcare system is a nationwide, federally funded “equal-access” healthcare system. Access to care is not restricted by insurance status or an inability to pay for care. Physicians who provide care in this system are salaried and lack financial incentives to provide or deny care to patients (8).

The main objectives of our study were twofold. The first was to determine racial differences in short-term and intermediate-term mortality in a large, diverse cohort of patients hospitalized for CHF in the VA healthcare system. The second was to examine racial differences in the patterns of healthcare utilization in these patients after the index hospitalization.

We performed a retrospective cohort study of veterans hospitalized with CHF. We identified all veterans who were admitted to 153 acute-care VA hospitals nationwide with a principal diagnosis of CHF (diagnosis related group [DRG] = 127) between October 1, 1997, and September 30, 1999. To control for stage of CHF we included only those patients who had a first admission for CHF, as defined by no VA hospitalizations for CHF in the prior two years. We excluded patients for whom race could not be identified in the database (n = 300) and restricted our analyses to black and non-Hispanic white patients (n = 21,994). For evaluation of short-term mortality we used the entire cohort of 21,994 patients. However, to examine healthcare utilization and intermediate-term mortality following discharge, the cohort included only patients discharged alive after the index admission (n = 21,003).

We linked three national VA databases: the VA Patient Treatment Files (PTF), which contain abstracts on all patients discharged from VA hospitals; the VA Outpatient Clinic files, which contain records for all services provided to outpatients by VA staff; and the VA Beneficiary Identification and Records Locator Subsystem death file. The VA death file contains death records regardless of the place of death and captures information on 90% to 95% of deceased veterans (9). In addition, death was identified from the PTF. All the databases contain encrypted patient identifiers and can be linked to create person-specific longitudinal utilization and vital status records.

Both short-term and intermediate-term mortality were outcome measures. Short-term mortality was defined as mortality within 30 days of the date of the index admission. Intermediate-term mortality was measured at one and two years following discharge in the cohort of patients that were discharged alive after the index admission.

The VA healthcare utilization was measured for the year following discharge. Rates of medical clinic visits and urgent care/emergency room visits were used as measures of outpatient care. The rate of all-cause readmissions and bed-days of care as well as the rate of medical readmissions were used as measures of inpatient care (10- 11). The all-cause readmission rate is the average number of discharges that a patient had in the year following discharge. The bed-day rate is the average number of days spent in the hospital for any admissions per patient-year of follow-up after discharge. Similar rates were calculated for medical readmissions, that is, all admissions excluding those with a DRG for psychiatry or surgery. Both the bed-day rate and the readmission rate measure the intensity of hospital use in a cohort. Avoiding hospitalizations or shortening hospitalizations will both reduce the bed-days of care, but only the former will reduce the readmission rate. Therefore, the bed-day rate will be the more sensitive indicator of changes in hospital use in a system. Medical clinic visits occurred when a patient received outpatient care in a general medical or medical subspecialty clinic.

Person time was used as the denominator for the rates. All rates were calculated per patient-year of follow-up. In the year after discharge, the number of days spent in the hospital was deducted from the days “at risk” for readmission and outpatient visits.

The Student t test was used to test for baseline differences in continuous variables and the chi-square test was used for categorical variables. Unadjusted survival curves were generated by the Kaplan-Meier method, and survival was compared between groups using the log-rank test. Logistic regression was used to calculate the risk-adjusted odds of mortality for black patients relative to white patients. To account for correlations among patients treated at the same hospital and for similar practice patterns among hospitals located in a particular geographic area, we used hierarchical mixed models. Patients were nested within hospitals and hospitals were nested within Census Bureau divisions. Risk adjustment was performed by controlling simultaneously for covariates that included demographic variables (age, gender, income, and marital status, with marital status and income used as measures of social support) and comorbidities. Using International Classification of Diseases-Ninth Revision-Clinical Modification (ICD-9-CM) diagnosis codes, a panel of clinicians selected 21 comorbid conditions a priori as potentially important in influencing mortality and utilization in patients with CHF. The list of all comorbidities and the ICD-9-CM codes used to define the comorbidities is provided in the Appendix. A log transformation of the utilization measures was performed, as the distribution of the number of clinic visits and hospitalizations was right-skewed. Again, hierarchical mixed models, as described earlier, were used to calculate risk-adjusted rates of utilization per patient-year of follow-up and to compare these rates between black and white patients. All values of p < 0.01 were considered statistically significant. Statistical analysis was performed using SAS software 8.2 (SAS Institute, Cary, North Carolina).

The cohort included 4,901 (22.3%) black patients and 17,093 (77.7%) white patients. (Table le1) displays the baseline characteristics in the two groups. More than 98% of the cohort members were men. Black patients were on average four years younger than white patients. A higher proportion of black patients had an annual income <$10,000 and were unmarried compared with white patients. Furthermore, black patients were more likely to have a history of hypertension and renal insufficiency, whereas white patients were more likely to have ischemic heart disease, chronic lung disease, diabetes, peripheral vascular disease, and cerebrovascular disease. Overall, the burden of comorbidities was higher in white patients (Figure 1). A greater proportion of white patients (49%) had ≥5 comorbidities as compared with black patients (41.4%), and a greater proportion of black patients (31.2%) had ≤2 comorbidities as compared with white patients (24.2%, p < 0.0005).

During the index admission, 160 (3.3%) black patients and 831 (4.9%) white patients died (p < 0.0005). At 30 days after admission, the crude mortality was 227 (4.6%) in blacks and 1,253 (7.3%) in whites (p < 0.0005). At one year after discharge, there were 949 deaths (20%) in black patients and 4,190 deaths (25.8%) in white patients (p < 0.0005), and at two years there were 1,444 deaths (30.5%) and 6,087 deaths (37.5%), respectively (p < 0.001). (Figure 2) shows the Kaplan-Meier survival curves in black and white patients.

As shown in (Table le2), the risk-adjusted mortality at all time points was also significantly lower in black patients when compared with white patients. The risk-adjusted odds ratio (OR) for mortality at 30 days after admission for black compared with white patients was 0.7 (95% CI 0.6 to 0.82). At one and two years after discharge, the ORs were 0.82 (95% CI 0.75 to 0.9) and 0.84 (95% CI 0.78 to 0.91), respectively. The survival advantage was consistent across subgroups of patients by age, presence of hypertension, presence of coronary artery disease, and presence of diabetes (Table le3).

Risk-adjusted rates of healthcare utilization in the year following discharge are shown in (Table le4). After discharge, on average, black patients received a lesser number of medical outpatient clinic visits (8.6 ± 0.4) as compared with white patients (9.6 ± 0.3, p < 0.001), though this difference was relatively small. In contrast, black patients received somewhat more of their outpatient care through the emergency room/urgent care clinic, with an average of 0.4 more urgent care/emergency room (ER) visits in black patients compared with white patients, per patient-year of follow up (p < 0.001). However, the readmission rates for all-cause readmissions as well as for medical readmissions, and the bed days of care per patient-year of follow-up were similar in black and white patients (Table le4).

In the VA healthcare system, a system designed to provide equal access to healthcare, irrespective of the ability to pay, we found significant differences in short-term and intermediate-term mortality by race, but relatively small differences in patterns of healthcare utilization, after a hospitalization for CHF. Our results demonstrate a lower short-term and intermediate-term mortality in black patients with a first hospitalization for CHF as compared with white patients. One explanation for this finding is that black patients may have a lesser overall disease burden or be at an earlier stage of disease progression at the time of admission compared with white patients. In our study, black patients were younger and had fewer comorbidities compared with white patients, consistent with a lesser overall burden of disease. We have attempted to account for differences in disease burden by risk-adjustment of the outcomes for differences in age and several comorbidities, as well as by the method of selection of our cohort. Because the first hospitalization for a chronic condition often signifies that the disease has become more severe, we selected our cohort using an index admission that was the first admission for CHF in a two-year period. In addition, we examined mortality across various subgroups and found that the survival advantage was consistent in both younger and older black patients as well as in those with coronary artery disease, diabetes, or hypertension. In spite of these measures, we may have been unable to account for all differences in clinical severity, and it is possible that the differences in mortality may be attributable to unmeasured residual confounding.

An alternative explanation for the higher mortality in whites would be that the care received by white patients was systematically inferior to that received by black patients. However, extensive data in the literature on racial disparities in healthcare would not support that explanation. For example, two previous studies that evaluated patients admitted for CHF to non-VA hospitals showed worse quality of care for black patients compared with white patients (12- 13). We were unable to directly evaluate racial differences in the process of care and medication use for the patients. However, findings from recent studies within the VA have suggested that the process of care delivered to hospitalized VA patients may not differ significantly by race (14- 15). Most recently, Gordon et al. (14) evaluated process of hospital care by race in 2,582 patients discharged from 12 VA hospitals. They found that inpatient treatment and discharge care were similar among black, white, and Hispanic patients admitted with one of three medical diagnoses, including CHF. In addition, Petersen et al. (15) recently showed that at the time of discharge from the VA, prescription of medications in patients with acute myocardial infarction, including beta-blockers, was similar for black and white patients. In addition, a recent study of Medicare beneficiaries has shown that the quality of care did not differ by race, including prescription of angiotensin-converting enzyme inhibitors or angiotensin receptor blockers at discharge after a heart failure (HF) hospitalization (16). In our study, an indirect parameter to judge process of care during the hospitalization is the average length of stay during an admission. If a significant number of black patients were being discharged prematurely, it would be expected that the average number of days spent in the hospital for the same number of admissions would be lower in blacks compared with whites, but this was not found to be the case. No published studies have evaluated process of care and adherence to guidelines by race in outpatients with CHF within the VA system. This remains an area for future research. Other explanations for better survival in black patients, including better social support or better socioeconomic status, are also not supported by the findings of our study or those of others (17- 18).

Furthermore, the etiology, pathophysiology, cardiovascular adaptation to cardiac injury, and manifestations of CHF may be different in black patients compared with white patients, leading to a relatively higher need for hospitalization at an earlier stage in the disease. For example, multiple studies including ours have shown that blacks with CHF are younger at presentation and are more likely to have hypertension and left ventricular hypertrophy with a lower frequency of coronary artery disease (19- 22). Hypertensive heart disease associated with a greater susceptibility to sodium retention in black patients may make this ethnic group more prone to sudden symptomatic pulmonary congestion with resultant use of hospitalization at earlier stages of disease progression (7), but with better survival after that index hospitalization. In addition, it may be expected that hypertensive heart disease is associated with a higher proportion of diastolic HF in black patients and may therefore have a better prognosis. However, review of prior studies of patient cohorts with CHF, which include both diastolic and systolic HF and which have documented the racial distribution of patients, have not demonstrated a higher proportion of black patients as having diastolic HF compared with white patients, and one study has even demonstrated a higher percentage of systolic HF in black compared with white patients (23- 30).

The finding that that the black patient cohort had better survival, appears to be in contrast to the accepted belief that black patients with CHF present at a later course in their disease and have a worse prognosis compared with white patients (1,31). However, most studies or statistics documenting poorer outcomes in blacks with CHF have not been restricted to hospitalized patients, but have included the overall population with HF, either in clinical trials or in the community (1,19). It should be emphasized that studies like ours that demonstrate better outcomes in black patients have studied hospitalized patients. Earlier non-VA studies have suggested that black patients hospitalized with CHF may represent a group with a lesser severity of illness owing to different patterns of hospitalization by race (2,6- 7). Black patients may have had reduced access to outpatient care through lack of insurance or an inability to pay for medical care, and they may therefore receive a higher proportion of their medical care in an in-hospital setting (2,6). However, this explanation is not supported by our study. The VA system is a financially “equal-access” healthcare system. In addition, in our study, black patients did not receive a larger share of their care in a hospitalized setting, as demonstrated by a similar rate of readmissions and similar number of days spent in the hospital for medical care by race.

The findings of lower short-term mortality after admission in blacks have been shown in a previous study of patients hospitalized in VA hospitals with a number of different medical diagnoses, including CHF (5). Jha et al. (5) found a lower short-term mortality (at 30 days and 6 months) in hospitalized black VA beneficiaries. Our study adds to their study by evaluating longer term survival, as well as by examining patterns of healthcare utilization after the index admission. Our study demonstrated that the survival advantage in black patients in the short-term is maintained over time, at least for two years of follow-up.

In a setting designed to provide financially equitable access to care, we found that both black and white patients with CHF received relatively high-intensity outpatient care. In the year following discharge, even though black patients had a fewer number of medical clinic visits as compared with white patients, the difference was relatively small. This appeared to have been offset by a slightly higher number of visits to the ER/urgent care by the black patients. Furthermore, within the VA setting, the overall outpatient care (outpatient care and urgent care/emergency room visits) received by black patients appeared to be successful in keeping the pattern of readmissions similar to the white patients. As noted earlier, this is in contrast to the non-VA setting, where most studies have shown an increased number of rehospitalizations in black patients with CHF (2- 4). The overall increased access to care in the outpatient setting may have reduced the racial differences in in-hospital use of services observed in non-VA studies. The reasons for the small differences observed in patterns of outpatient service utilization with the somewhat lower rates of medical clinic visits and higher rates of urgent care/ER visits could not be addressed by our study. They may include physician, healthcare system, or patient factors such as racial disparities or bias in the scheduling of follow-up visits or nonadherence to scheduled visits. Lesser social support, as demonstrated by a higher proportion of black patients being unmarried and having a lower income, may make black patients less able to adhere to regularly scheduled visits. In addition, one should also consider racial differences in patient-physician interaction and differences in the way racially diverse groups of patients accept the recommendations of physicians as a possible explanation for our findings (32- 35). Interventions designed to make physicians more sensitive to cultural diversity and bias and to influence the patient-physician interaction may be important in improving and eliminating racial disparities in the patterns of healthcare utilization in both VA and non-VA settings (36).

The present study was based on retrospective analyses of VA administrative databases. Such data do not include sufficient detail on clinical severity of illness in CHF. However, in another study using detailed clinical data obtained by chart review, we examined outcomes in black and white patients admitted to a tertiary care VA hospital over a similar time period. Analysis of that data showed that black race was an independent predictor of improved survival after hospitalization during follow-up of up to 36 months (29). In addition, it is important to point out that we only measured healthcare utilization within the VA. Veterans age 65 years and older are eligible to receive care either in the VA system or in the private sector under Medicare. Of note, 80% of the patients in our cohort are ≥65 years. To examine this issue, our co-authors (C.M.A., N.J.P., N.P.W.) have evaluated VA and Medicare (non-VA) bed-days of care rates in fiscal years 1997 and 1998, in a national cohort of CHF patients ≥65 years who had an index admission at a VA hospital for CHF. They found that non-VA, Medicare-reimbursed bed-days of care constituted 19% and 21% of bed-days of care per patient-year for patients in their cohort during 1997 and 1998, respectively (37). When analyzed by race for fiscal years 1997 and 1998, Medicare bed-days accounted for 21% and 23% of the bed-days of care per patient-year in black patients and 19% and 21% of the bed-days of care per patient-year in white patients, respectively (Dr. Carol Ashton, unpublished data, May 2003). Therefore, the data do not support excessive non-VA, Medicare-reimbursed hospital use by white veterans as compared with black patients with CHF. However, we cannot rule out differential utilization of non-VA, Medicare-reimbursed outpatient services by race in this patient population. Database studies linking VA files to Medicare outpatient utilization files are needed to address this issue. In addition, we do not have data on non-VA use in patients under 65 years of age, but such patients form less than 20% of our cohort. Also, these patients generally have very limited access to non-VA care, as the great majority of VA beneficiaries are underinsured or uninsured (38).

Another limitation of this study is that men accounted for almost 99% of patients in the cohort, consistent with the demographics of hospitalized VA beneficiaries. The results, therefore, cannot be generalized to women, especially because there are gender differences in HF outcomes that may interact with race (39).

In the VA healthcare system, a system designed to provide financially “equal access” to care for all enrolled patients, the racial gap in patterns of healthcare utilization in patients with CHF is small. The observation of better survival in black patients after a hospitalization for CHF is not readily explained by differences in healthcare utilization and needs further evaluation.

The authors gratefully acknowledge the indefatigable programming assistance of Meei Ku-Goto and Diana Urbauer. We thank Dr. Douglas Mann for his support and guidance.

Appendix 1, online

For a list of all the comorbidities and the ICD-9-CM codes used to define the comorbidities, please see the March 3, 2004, issue of JACC at http://www.cardiosource.com/jacc.html.