Abstract
Background
Critics have raised concerns that specialty cardiac hospitals exacerbate racial disparities in cardiovascular care, but empirical data are limited.
Methods and Results
We used administrative data from the Medicare Provider and Analysis Review (MEDPAR) Part A and Provider-of-Service (POS) files from 2002 to 2005. Multivariable logistic regression models were constructed to examine the likelihood of black Medicare patients being admitted to a cardiac hospital for coronary revascularization when compared with white patients within the same healthcare referral region (HRR) after accounting for geographic proximity to the nearest hospitals, procedural acuity and co-morbidities. We identified 35,309 patients who underwent coronary artery bypass grafting (CABG) in 18 HRRs and 94,525 patients who underwent percutaneous coronary intervention (PCI) in 20 HRRs with cardiac hospitals performing these procedures. Patients at cardiac hospitals were more likely to be men and white, and to have less co-morbidity than those at general hospitals. The likelihood of black patients undergoing coronary revascularization at a cardiac hospital was significantly lower for CABG (adjusted odds ratio [OR], 0.67; P=0.01) and PCI (adjusted OR, 0.63; P<0.0001). However, this relationship was substantially attenuated in black patients living in close proximity (i.e., within 10 miles) to cardiac hospitals (adjusted OR for CABG, 0.95; p=0.75; adjusted OR for PCI, 0.78; P=0.01).
Conclusions
Black patients were significantly less likely to be admitted at cardiac hospitals for coronary revascularization. Precise reasons for these findings are unclear, but suggest complex associations between race and geography in decisions about where to receive care.
Keywords: Racial disparities, coronary revascularization, specialty hospitals
Specialty hospitals typically provide care to narrow patient populations and for a limited number of medical procedures, with the majority of hospitals focused on cardiovascular or orthopedic conditions.1 Proponents of specialty hospitals argue that these facilities provide higher quality healthcare and greater cost-efficiency by concentrating physician skills and hospital resources on the needs and services of a specific patient population.2 Critics claim that specialty hospitals, which are largely physician-owned, “game” the hospital reimbursement system by primarily treating low-risk and well-insured patients and providing less uncompensated care.3 Critics have even raised accusations that, by focusing on these types of patients, specialty hospitals may exacerbate existing racial disparities and hospital segregation.
In 2005, the Medicare Payment Advisory Commission (MedPAC) released a report that documented a 60% lower proportion of black patients treated at specialty cardiac hospitals when compared with peer general hospitals in the same healthcare market (3.6% of Medicare discharges versus 9.6%). This finding drew the attention of policy-makers, resulting in an open letter from three U.S. congressmen to the Centers for Medicare & Medicaid Services (CMS).4 Although the MedPAC report’s findings are highly provocative, the approach used in that analysis had important limitations. First, the analysis did not account for distances that black and white patients lived from specialty hospitals, despite the critical role of geographic location in hospital choice.5 Second, it was unclear the extent to which these differences in racial distributions across hospitals may have been explained by differences in procedural acuity and patient co-morbidity, both of which were not adjusted for in the analysis but could contribute to hospital choice. These issues are critical for placing the contentious results from the MedPAC report in proper context.
Accordingly, the purpose of this study was to re-examine the question of whether black patients were less likely to undergo coronary revascularization at cardiac hospitals when compared with white patients. We focused on cardiac hospitals since these facilities are responsible for the bulk of Medicare payments to specialty hospitals and because of the large literature on existing disparities in coronary revascularization. Importantly, we used multivariable analyses to account for geographic proximity to the nearest cardiac hospital as well as for differences in procedural acuity and co-morbidities while assessing the likelihood that a black patient would be treated at a cardiac hospital.
Methods
Data Sources and Study Population
For these analyses, Medicare Provider and Analysis Review (MEDPAR) Part A and Provider-of-Service (POS) files from 2002 to 2005 were obtained from CMS. Part A files include data on acute-care hospitalizations for fee-for-service Medicare beneficiaries. POS files contain data on hospital providers including facility characteristics and Zip code locations.
We used the International Classification of Diseases, Ninth Revision, Clinical Modification (ICD-9-CM) procedural codes to identify all patients who underwent coronary revascularization with coronary artery bypass grafting (CABG) (ICD-9-CM procedural codes, 36.10-36.19) or percutaneous coronary intervention (PCI) (ICD-9-CM procedural codes, 36.01, 36.02, 36.05-36.07, 36.09). We limited our analysis to patients 65 years or older identified by race as white or black given the small number of individuals in other racial and ethnic categories. Furthermore, identifying race has been most reliable for black and white patients.6 We also excluded patients (1) enrolled in a Medicare managed care plan, (2) treated outside of the U.S., or (3) admitted from institutionalized settings, like a skilled nursing facility, or transferred from another acute-care facility. The last group – which accounted for 4635 patients undergoing CABG and 10,583 patients undergoing PCI - was excluded because these patients often differ substantially from other patients both in terms of demographics and complexity (although their inclusion did not substantially impact on our overall results).7 Finally, we limited our analysis to patients treated in hospitals that performed at least 5 CABGs or PCIs during the last year of the study period. Patients who underwent both CABG and PCI during the study period (n=2346) were included in both study populations during the analysis given the relatively long study period and the likelihood of repeat revascularizations. However, we did perform a sensitivity analysis after excluding these patients and found nearly identical results (which are available from the authors).
Specialty Hospital and Market Identification
We categorized all hospitals that performed coronary revascularization into cardiac or peer general hospitals using an approach similar to the General Accounting Office (GAO) and others.8 Specifically, we constructed a cardiac specialty index based on the percentage of cardiac-to-total admissions in Medicare beneficiaries in 2002 and 2003; thus, a higher percentage correlated with a greater degree of hospital cardiac specialization. From this cardiac specialty index, we reviewed the top 100 facilities and selected those that: 1) had proprietary or corporate ownership, 2) did not provide obstetrical or pediatric services, and 3) were identified by CMS as a “physician-owned specialty hospital” during their recent national survey.9 This selection process was meant to exclude physician-owned general hospitals or cardiac hospitals without physician-ownership. Data on additional services available at these hospitals were obtained from the American Hospital Association Annual Survey, American Hospital Directory and online hospital websites.10,11
We used Hospital Referral Regions (HRRs) from the Dartmouth Atlas of Health Care to identify competing peer general hospitals within the same healthcare market as one or more cardiac hospitals.12 HRRs are large geographic units representing distinct markets for tertiary care that were developed by studying patterns of hospital utilization for major cardiac surgery and neurological surgery among Medicare beneficiaries in the mid-1990s. Based on their Zip code, patients and hospitals were assigned to 1 of 306 HRRs. We identified 19 HRRs with at least one cardiac hospital performing CABG and 20 HRRs with at least 1 cardiac hospital performing PCI. For analyses related to CABG, we excluded 1 HRR (Sioux Falls, SD) because it included 0 black patients during the study period. Thus, the final sample contained 18 HRRs for CABG and 20 HRRs for PCI.
Statistical Analysis
Univariate analyses were performed to compare the characteristics of patients admitted to cardiac and peer general hospitals – including age, sex, race, procedural acuity, and the presence of specific co-morbidities – using Student t-tests for continuous variables and chi-squared tests for categorical variables. We then used multivariable logistic regression models to evaluate the likelihood of a black patient undergoing coronary revascularization at a cardiac hospital as compared with peer general hospitals in the same HRR after adjusting for baseline differences between patients treated at each type of facility. In separate regression models for CABG and PCI, the dependent variable was “admission to a specialty cardiac hospital” and the key independent variable of interest was an indicator variable representing black race with a reference group of white race.
To account for the relationship between distance to a specific hospital and admission to that hospital in our models, we included as a patient-level covariate in these models a measure of differential distance as a continuous variable. Differential distance was calculated for each patient using population-based centroids of the five digit zip code that was associated with each patient’s residential address and the address for the nearest cardiac and peer general hospital within the same HRR.13 As such, it specifically refers to the difference in distances between the nearest cardiac hospital and the nearest peer general hospital with a positive differential distance indicating that the patient lived closer to a peer general hospital while a negative differential distance suggested that a patient lived closer to a cardiac hospital. For purposes of our analyses, inclusion of differential distance allowed us to examine whether cardiac hospitals were more (or less) likely to admit black patients after accounting for the relationship between hospital location and the neighborhoods where blacks and whites may live.
Other covariates included in the models were: age (65-69, 70-74, 75-79, 80-84, 85-89, 90 or older), gender, procedural acuity (elective, urgent and emergent), and co-morbidities. Co-morbidities were assessed using the Quan coding algorithm, a recently described method for assessing co-morbidities from ICD-9-CM diagnostic codes. Recent data suggest that the Quan coding algorithm may outperform previous approaches for determining the prevalence of co-morbidities in administrative data like the Charlson co-morbidity score.14 In the final models for CABG and PCI, we included co-morbidities that had both clinical validity and a P-value of <0.05, with the intent to account for patient differences in important co-morbidities across hospitals. (A full list of these covariates that were included for CABG and PCI are available in Tables 2 and 3.)
Table 2.
Characteristics of CABG patients at cardiac and peer general hospitals.
| Cardiac Hospitals N = 7978 | Peer General Hospitals N=27,331 | P-value | |
|---|---|---|---|
| Gender (%) | |||
| Female | 2647 (33.18) | 9405 (34.41) | 0.0410 |
| Race (%) | |||
| Black | 179 (2.24) | 1037 (3.79) | <.0001 |
| Age (%) | |||
| 65-69 | 1409 (17.66) | 4954 (18.13) | 0.0616 |
| 70-74 | 2457 (30.80) | 8508 (31.13) | |
| 75-79 | 2252 (28.23) | 7797 (28.53) | |
| 80-84 | 1411 (17.69) | 4654 (17.03) | |
| 85-89 | 413 (5.18) | 1268 (4.64) | |
| 90+ | 36 (0.45) | 150 (0.55) | |
| Admission Type (%) | |||
| Elective | 5382 (67.46) | 14987 (54.84) | <.0001 |
| Urgent | 1657 (20.77) | 5471 (20.02) | |
| Emergent | 905 (11.34) | 6845 (25.04) | |
| Admission Source (%) | |||
| Emergency Room | 597 (7.48) | 6820 (24.95) | <.0001 |
| Comorbidities | |||
| Hypertension Uncomplicated | 4482 (56.18) | 15299 (55.98) | 0.7482 |
| Congestive Heart Failure | 1644 (20.61) | 6854 (25.08) | <.0001 |
| Myocardial Infarction | 1562 (19.58) | 7644 (27.97) | <.0001 |
| Cerebrovascular Disease | 1078 (13.51) | 2457 (8.99) | <.0001 |
| Peripheral Vascular Disease | 1151 (14.43) | 3131 (11.46) | <.0001 |
| Diabetes mellitus | 168 (2.11) | 826 (3.02) | <.0001 |
| COPD | 1650 (20.68) | 6152 (22.51) | 0.0005 |
| Valvular Heart Disease | 2228 (27.93) | 6716 (24.57) | <.0001 |
| Fluid Disorder | 1328 (16.65) | 4146 (15.17) | 0.0014 |
| Tumor no metastasis | 60 (0.75) | 407 (1.49) | <.0001 |
| Blood Loss Anemia | 29 (0.36) | 298 (1.09) | <.0001 |
| Weight Loss | 65 (0.81) | 401 (1.47) | <.0001 |
| Psychosis | 52 (0.65) | 119 (0.44) | 0.0143 |
| Alcohol Abuse | 86 (1.08) | 244 (0.89) | 0.1304 |
| Dementia | 17 (0.21) | 40 (0.15) | 0.1915 |
| Differential Distance, miles (SD)* | 13.26 (24.39) | 16.17 (25.16) | <.0001 |
Source: CMS Data: 2004 to 2006
Distance to the nearest cardiac hospital minus the distance to the nearest peer general hospital; greater positive values imply the patient lives closer to a peer general hospital
Table 3.
Characteristics of PCI patients at cardiac and peer general hospitals.
| Cardiac Hospitals N = 18,844 | Peer General Hospitals N=75,681 | P-value | |
|---|---|---|---|
| Gender (%) | |||
| Female | 7571 (40.18) | 33289 (43.99) | <.0001 |
| Race (%) | |||
| Black | 523 (2.78) | 4021 (5.31) | <.0001 |
| Age (%) | |||
| 65-69 | 4176 (22.16) | 17829 (23.56) | 0.6596 |
| 70-74 | 5237 (27.79) | 19635 (25.94) | |
| 75-79 | 4554 (24.17) | 18059 (23.86) | |
| 80-84 | 3205 (17.01) | 13095 (17.30) | |
| 85-89 | 1384 (7.34) | 5755 (7.60) | |
| 90+ | 288 (1.53) | 1308 (1.73) | |
| Admission Type (%) | |||
| Elective | 11965 (63.50) | 32891 (43.46) | <.0001 |
| Urgent | 3572 (18.96) | 15175 (20.05) | |
| Emergent | 3251 (17.25) | 27535 (36.38) | |
| Admission Source (%) | |||
| Emergency Room | 2383 (12.65) | 27976 (36.97) | <.0001 |
| Other | 16461 (87.35) | 47705 (63.03) | |
| Comorbidities | |||
| Hypertension Complicated | 832 (4.42) | 5024 (6.64) | <.0001 |
| Hypertension Uncomplicated | 11825 (62.75) | 43177 (57.05) | <.0001 |
| Congestive Heart Failure | 2778 (14.74) | 15179 (20.06) | <.0001 |
| Cerebrovascular Disease | 1286 (6.82) | 3538 (4.67) | <.0001 |
| Peripheral Vascular Disease | 2974 (15.78) | 8316 (10.99) | <.0001 |
| Diabetic Complicated | 249 (1.32) | 1844 (2.44) | <.0001 |
| COPD | 2787 (14.79) | 13127 (17.35) | <.0001 |
| Emphysema or Bronchitis | 293 (1.55) | 2077 (2.74) | <.0001 |
| Pulmonary Circulatory Disease | 443 (2.35) | 1464 (1.93) | 0.0003 |
| Valvular Disease | 2567 (13.62) | 8384 (11.08) | <.0001 |
| Fluid Disorder | 721 (3.83) | 5258 (6.95) | <.0001 |
| Renal Failure | 483 (2.56) | 3376 (4.46) | <.0001 |
| Renal Disease | 485 (2.57) | 3430 (4.53) | <.0001 |
| Mild Liver Disease | 41 (0.22) | 338 (0.45) | <.0001 |
| Lymphoma | 38 (0.20) | 287 (0.38) | 0.0002 |
| Tumor no Metastasis | 147 (0.78) | 1287 (1.70) | <.0001 |
| Blood Loss Anemia | 72 (0.38) | 652 (0.86) | <.0001 |
| Weight Loss | 31 (0.16) | 358 (0.47) | <.0001 |
| Neurological Disease | 290 (1.54) | 1699 (2.24) | <.0001 |
| Paralysis | 15 (0.08) | 135 (0.18) | 0.0023 |
| Psychosis | 22 (0.12) | 192 (0.25) | 0.0004 |
| Alcohol Abuse | 113 (0.6) | 438 (0.58) | 0.7358 |
| Drug Abuse | 15 (0.08) | 186 (0.25) | <.0001 |
| Dementia | 48 (0.25) | 386 (0.51) | <.0001 |
| Differential Distance, miles (SD)* | 19.85 (33.21) | 22.86 (32.63) | <.0001 |
Source: CMS Data: 2004 to 2006
Distance to the nearest cardiac hospital minus the distance to the nearest peer general hospital; greater positive values imply the patient lives closer to a peer general hospital
All models were adjusted for potential clustering effects by HRR using random effects models with the healthcare market included as a random intercept, with the coefficient for race representing the marginal coefficient over all HRRs. We explored potential interactions between race and gender given prior evidence of its potential importance in the recommendations for invasive cardiac procedures.15 Additional interactions between race and differential distance were also explored. Finally, we performed sensitivity analyses that examined patterns of admission to cardiac hospitals in (1) patients undergoing elective revascularization, (2) patients who resided within 10 miles of a cardiac hospital, and (3) patients with differential distances within 10 miles. All analyses were performed using SAS Version 9.1.3 (Research Triangle Institute, NC). P-values of <0.05 were considered significant and all statistical tests were 2-sided. C-statistics for the full model and models used during sensitivity analyses for both CABG and PCI ranged from 0.77 to 0.80. This research proposal was approved by the University of Iowa Institutional Review Board.
Results
We identified 18 HRRs that contained 20 cardiac hospitals and 102 peer general hospitals performing CABG; in contrast, we identified 20 HRRs that contained 22 cardiac hospitals and 167 peer general hospitals performing PCI. Most HRRs had a single cardiac hospital with 2 HRRs having 2 cardiac hospitals. The number of peer general hospitals in each of the HRRs varied, ranging from 2 to 14 for CABG and 2 to 24 for PCI (Table 1). In addition, the geographic distribution of these HRRs across the U.S., their proportion of black Medicare enrollees and their population-based rates of CABG and PCI during the study period are displayed in Table 1. The proportion of black Medicare enrollees varied across HRRs from <1% in Sioux Falls, SD to nearly 30% in Lafayette, LA. Substantial variation in the use of coronary revascularization was also noted in population-based rates for both CABG and PCI.
Table 1.
Hospital referral regions with cardiac hospitals between 2002 and 2005.
| HRR | HRR Name | Cardiac Hospitals (n) | Peer General Hospitals (n) | % (SD) of blacks in population* | No. of CABG per 1000 mean, (SD) | No. of PCI per 1000 mean, (SD) | ||
|---|---|---|---|---|---|---|---|---|
| CABG | PCI | CABG | PCI | |||||
| 12 | PHOENIX, AZ | 1 | 1 | 14 | 24 | 3.33 (0.02) | 4.3 (0.4) | 13.8 (0.3) |
| 15 | TUCSON, AZ | 1 | 1 | 3 | 7 | 2.71 (0.00) | 4.0 (0.7) | 10.7 (1.1) |
| 19 | LITTLE ROCK, AR | 1 | 1 | 10 | 11 | 19.79 (0.10) | 6.9 (0.8) | 15.8 (2.4) |
| 25 | BAKERSFIELD, CA | 1 | 1 | 3 | 3 | 4.28 (0.04) | 5.8 (1.0) | 15.3 (0.9) |
| 43 | FRESNO, CA | 1 | 1 | 2 | 3 | 5.09 (0.06) | 4.9 (0.2) | 12.3 (2.1) |
| 183 | INDIANAPOLIS, IN | 2 | 2 | 7 | 12 | 9.56 (0.08) | 5.6 (0.3) | 10.7 (0.2) |
| 201 | WICHITA, KS | 2 | 2 | 5 | 6 | 4.19 (0.03) | 7.6 (0.2) | 14.8 (0.6) |
| 213 | LAFAYETTE, LA | 1 | 1 | 3 | 7 | 27.52 (0.05) | 6.0 (0.7) | 27.9 (3.5) |
| 220 | SLIDELL, LA | 1 | 1 | 2 | 2 | 11.12 (0.05) | 5.8 (2.2) | 19.7 (3.7) |
| 277 | LINCOLN, NE | 1 | 1 | 2 | 3 | 1.56 (0.06) | 8.2 (0.6) | 14.9 (1.1) |
| 293 | ALBUQUERQUE, NM | 1 | 1 | 3 | 5 | 1.33 (0.02) | 3.3 (0.14) | 6.4 (0.1) |
| 330 | DAYTON, OH | 1 | 1 | 7 | 7 | 12.82 (0.09) | 6.8 (0.4) | 14.7 (2.2) |
| 339 | OKLAHOMA CITY, OK | 1 | 1 | 9 | 12 | 7.99 (0.06) | 6.1 (0.2) | 18.0 (2.5) |
| 340 | TULSA, OK | 1 | 1 | 5 | 7 | 7.37 (0.04) | 4.6 (0.3) | 11.7 (1.4) |
| 371 | SIOUX FALLS, SD | 1 | 1 | 2 | 2 | 0.74 (0.06) | 4.7 (0.6) | 10.4 (0.8) |
| 385 | AUSTIN, TX | 1 | 1 | 5 | 6 | 6.96 (0.12) | 4.2 (0.5) | 13.3 (0.5) |
| 391 | DALLAS, TX | NA | 1 | NA | 24 | 13.85 (0.04) | 5.2 (0.7) | 10.4 (0.2) |
| 400 | LUBBOCK, TX | 1 | 1 | 2 | 3 | 5.72 (0.08) | 7.1 (0.1) | 13.7 (0.8) |
| 412 | SAN ANTONIO, TX | 1 | 1 | 6 | 6 | 4.98 (0.05) | 5.0 (0.8) | 7.8 (1.0) |
| 451 | MILWAUKEE, WI | 1 | 1 | 14 | 17 | 12.07 (0.14) | 5.9 (0.8) | 14.7 (1.4) |
Source: CMS data 2004 to 2006
Mean and SD of the percentage of blacks in the population of the HRR between 2002 and 2005 extrapolated from the 2000 U.S. Census Survey.
HRR=hospital referral region; CABG=coronary artery bypass grafting; PCI=percutaneous coronary intervention; SD=standard deviation
We identified 35,309 patients who underwent CABG and 94,525 patients who underwent PCI in these HRRs. Patient characteristics for those who underwent CABG and PCI at cardiac and peer general hospitals are displayed in Tables 2 and 3. Patients at cardiac hospitals were more likely to be men and white for both CABG and PCI. Patients at cardiac hospitals were also more likely to undergo coronary revascularization as an elective procedure and less likely to be admitted from the emergency department. In general, patients at cardiac hospitals also had lower rates of co-morbidities when compared with patients at peer general hospitals with the exception of more cerebrovascular, peripheral vascular and valvular heart disease. Finally, the differential distance between the nearest cardiac and peer general hospitals for both CABG and PCI patients was smaller for patients treated at cardiac hospitals suggesting that patients admitted at these facilities lived relatively closer to them. We also noted that the mean and median distances to cardiac hospitals were 9.3 and 20.8 miles, respectively, for black patients and 23.4 and 38.4 miles, respectively, for white patients.
Among those who underwent CABG, 1216 (3.4%) were black patients and 34,093 (96.6%) were white patients. Among the black CABG patients, 179 (14.7%) were admitted to cardiac hospitals. In contrast, 7799 (22.9%) white CABG patients were admitted to cardiac hospitals. Among those who underwent PCI, 4544 (4.8%) were black patients and 89,981 (95.2%) were white patients. As with CABG, a smaller proportion of black PCI patients were treated at cardiac hospitals when compared with white patients: 523 (11.5%) black PCI patients were admitted to cardiac hospitals while 18,321 (20.4%) white patients were admitted to these facilities. In unadjusted analyses, the odds of black patients undergoing coronary revascularization at a cardiac hospital as compared to whites was 0.64 for CABG (95% CI, 0.54 to 0.76; P<0.0001) and 0.62 for PCI (95% CI, 0.56 to 0.68; P<0.0001).
Table 4 shows the likelihood of black patients undergoing coronary revascularization at cardiac hospitals when compared with white patients before and after multivariable analyses. Accounting for differential distance in addition to age, gender, procedural acuity and co-morbidities did not substantially influence the overall results (Table 4). The adjusted odds ratio for black patients undergoing coronary revascularization at a cardiac hospital was 0.67 (95% CI, 0.49 to 0.92; P=0.013) for CABG and 0.63 (95% CI, 0.53 to 0.74; P<0.0001) for PCI. There were no significant interactions noted between race and gender or between race and differential distance. Also, results were similar when we examined those patients undergoing elective revascularization only, which made up of 57.7% and 47.5% of the CABG and PCI patients, respectively.
Table 4.
Likelihood of black patients undergoing coronary revascularization at cardiac hospitals relative to white patients.
| CABG | PCI | |||
|---|---|---|---|---|
| Odds Ratio (95% CI) | P-Value | Odds Ratio (95% CI) | P-Value | |
| Unadjusted OR | 0.64 (0.54, 0.76) | <0.0001 | 0.62 (0.56, 0.68) | <0.0001 |
| Adjusted OR* | 0.67 (0.49, 0.92) | 0.0133 | 0.63 (0.53, 0.74) | <0.0001 |
| Adjusted OR* in elective cases only | 0.64 (0.49, 0.85) | 0.0017 | 0.75 (0.63, 0.89) | 0.0011 |
Adjusted for age, gender, differential distance, procedural acuity and co-morbidities CABG=coronary artery bypass grafting; PCI=percutaneous coronary intervention
Table 5 summarizes the results of several sensitivity analyses. Although our main results were largely unchanged in the subset of patients who lived relatively equidistant between specialty and general hospitals (differential distances less than 10 miles), we did note a significant interaction between race and distance in the group undergoing PCI (adjusted odds ratio for black patients being treated at a cardiac hospital with each additional mile, 0.94; P=0.006). Furthermore, in the subset of CABG and PCI patients who resided within 10 miles of a cardiac hospital, we found that black patients were no less likely than white patients to receive CABG at a cardiac hospital. Although black patients continued to be less likely to receive PCI at cardiac hospitals, the extent of this relationship was diminished. In this sensitivity analyses, a strong interaction effect between race and distance was again noted (adjusted odds ratio for black patients undergoing CABG at a cardiac hospital with each additional mile, 0.90; P=0.018; and adjusted odds ratio for black patients undergoing PCI at a cardiac hospital with each additional mile, 0.92; P=0.001). In general, these interaction effects suggested that differences in the likelihood of coronary revascularization at cardiac hospitals increased between black and white patients as they lived farther away from these facilities.
Table 5.
Sensitivity analyses examining the likelihood of black patients undergoing coronary revascularization at cardiac hospitals relative to white patients.
| N | CABG Odds Ratio (95% CI) | P-Value | N | PCI Odds Ratio (95% CI) | P-Value | |
|---|---|---|---|---|---|---|
| Adjusted OR*: | ||||||
| Differential distance, <10 miles | 22,671 | 0.60 (0.49, 0.74) | 0.0003 | 52,136 | 0.58 (0.48, 0.70)† | <.0001 |
| Adjusted OR* | ||||||
| Distance to nearest cardiac hospital, <10 miles | 10,716 | 0.95 (0.69, 1.31)† | 0.7522 | 28,616 | 0.78 (0.64, 0.94)† | 0.0095 |
Adjusted for age, gender, differential distance, procedural acuity and co-morbidities CABG=coronary artery bypass grafting; PCI=percutaneous coronary intervention
Interaction effect between race and distance, P-value<0.05
Discussion
We found that black patients in the Medicare population were significantly less likely to undergo coronary revascularization at cardiac hospitals when compared with white patients, even after accounting for differential distance to cardiac hospitals and peer general hospitals, procedural acuity and co-morbidities. However, this relationship was substantially attenuated in black patients living in close proximity to cardiac hospitals. These results were generally consistent for both CABG and PCI, and add new information to the previous literature on differences between patients treated at specialty and peer general hospitals.16 Most importantly, they suggest that racial differences that exist in the types of patients treated at cardiac hospitals are real and not easily accounted for by confounding factors like geographic proximity.
Cardiac hospitals have generally been shown to have better outcomes when compared with peer general hospitals for coronary revascularization, acute myocardial infarction and heart failure, although differences are frequently small and not always consistent across facilities.17,18 Reasons for these improved outcomes at cardiac hospitals remain largely unclear, but potential explanations have included: (1) a concentration of clinical expertise at a facility, (2) an ability to provide more comprehensive services, and (3) an improvement in provider control over hospital operations.18
Another possible reason for better outcomes at cardiac hospitals, however, may be due to careful patient selection at these facilities. Prior studies have suggested that patients treated at specialty hospitals are younger and have fewer co-morbidities when compared with those treated at peer general hospitals.16 Critics suggest that the role of physician-ownership plays an important role in this regard. Physician-ownership is thought to lead to ‘cherry-picking’ of healthier patients that are often more profitable to treat, leaving peer general hospitals to care for more severely-ill individuals.19 It is also logical, though unproven, that developers of specialty hospitals might place these facilities further away from underserved areas or communities in an effort to maximize profits. Thus, reduced access to cardiac hospitals and a greater burden of illness may partly explain why black patients were less likely to undergo coronary revascularization at cardiac hospitals.
However, we found that black patients actually lived closer to cardiac hospitals on average than white patients. This may relate to the fact that specialty hospitals tend to be located in close proximity to urban centers with higher concentrations of blacks. Our finding that black patients were significantly less likely to undergo coronary revascularization at cardiac hospitals was consistent even after adjusting for geographic proximity, procedural acuity and co-morbidities. This naturally raises questions as to why blacks are less likely to be admitted to cardiac hospitals. In particular, it is unclear to what extent our findings are driven by provider or patient selection or even facility level policies.20
Interestingly, in an important subgroup of patients who lived in close proximity to cardiac hospitals (i.e., <10 miles), we found that the relationship between race and the selection of a cardiac hospital was non-significant among CABG patients and attenuated in PCI patients. This finding implies to us less evidence for overt racism at cardiac hospitals and more evidence favoring the role of patient access or preference. It may be, for example, that black patients and their families have fewer resources to travel long distances for major procedures, limiting the choice of hospitals for their procedures. This speculation is supported by the significant interaction effects we noted between race and distance in this subgroup of patients. Alternatively, it may be that black and white patients living in close proximity to cardiac hospitals select hospitals in a similar manner due to more comparable socioeconomic factors.
Our findings should be interpreted in the context of the following limitations. First we relied on administrative data. All the complex clinical and socioeconomic factors possibly contributing to variation in hospital selection among patients were not accounted for, resulting in the potential for residual confounding. This also limited our ability to more carefully distinguish between racial subgroups of patients, including those with multiple racial or ethnic backgrounds. Yet use of administrative data did allow for an exhaustive examination of patterns of utilization for coronary revascularization using a population-based approach. This is unlikely to occur from more clinically-rich data sources where participation is typically voluntary. Second, we only focused on cardiac hospitals and the Medicare population. We also were not able to examine whether supplemental insurance has an impact on the use of revascularization at different hospitals. However, cardiac hospitals make up a large proportion of inpatient costs for Medicare and are relevant for policy-makers. Additional studies examining specialty hospitals in orthopedic or other surgical care may be valuable in evaluating similar questions in younger and uninsured patients. Third, the limited number of cardiac hospitals currently operating in the U.S. restricted our ability to examine the role of regional differences in our findings.
In summary, we found that black patients were significantly less likely to be admitted at cardiac hospitals for coronary revascularization when compared with white patients. However, we also found that the relationship between race and the selection of a cardiac hospital was less significant among CABG patients and attenuated in PCI patients when we considered only those patients living in close proximity to cardiac hospitals. Precise reasons for these findings are unclear, but suggest complex associations between race and geography in decisions about where to receive care. These potentially include differences in patient access and preference for cardiac hospitals, as well as their familiarity with the availability of specific services and facilities. As the number of cardiac hospitals continues to grow in the U.S., an improved understanding of why black patients were less likely to use cardiac hospitals may provide important insights into strategies for minimizing racial disparities.
Acknowledgments
Funding Support This project was supported by grant from the Agency for Healthcare Research and Quality (1R01HS015571-01A1) and the National Heart, Lung and Blood Institute (R01 HL085347-01A1). Dr. Cram is supported by a K23 career development award (RR01997201) from the NCRR at the NIH and the Robert Wood Johnson Physician Faculty Scholars Program.
Role of the Sponsor: The agencies and foundations that funded this work were not involved in the design and conduct of the study; in data management or analysis; or in manuscript preparation.
Footnotes
Financial Disclosure:None.
Disclaimer: The content of this publication does not necessarily reflect the views or policies of the Department of Health and Human Services, the Department of Veterans Affairs, or the Agency for Healthcare Research and Quality, nor does mention of trade names, commercial products, or organizations imply endorsement by the U.S. Government. The authors assume full responsibility for the accuracy and completeness of the ideas presented.
References
- 1.Iglehart JK. The emergence of physician-owned specialty hospitals. N Engl J Med. 2005;352:78–84. doi: 10.1056/NEJMhpr043631. [DOI] [PubMed] [Google Scholar]
- 2.Herzlinger RE. Specialization and its discontents: the pernicious impact of regulations against specialization and physician ownership on the US healthcare system. Circulation. 2004;109:2376–8. doi: 10.1161/01.CIR.0000130782.33860.E0. [DOI] [PubMed] [Google Scholar]
- 3.Voelker R. Specialty hospitals generate revenue and controversy. JAMA. 2003;289:409–410. doi: 10.1001/jama.289.4.409. [DOI] [PubMed] [Google Scholar]
- 4.Rangel C, Lewis J, Stark P. Letter to Mark McClellan, Centers for Medicare and Medicaid Services. Submitted on July 28, 2005. Available online at: http://www.house.gov/stark/news/109th/letters/mcclellan.pdf.
- 5.Norton EC, Staiger DO. How hospital ownership affects access to care for the uninsured. Rand J Econ. 1994;25:171–85. [PubMed] [Google Scholar]
- 6.Arday SL, Arday DR, Monroe S, Zhang J. HCFA’s racial and ethnic data: current accuracy and recent improvements. Health Care Financ Rev. 2000;21:107–16. [PMC free article] [PubMed] [Google Scholar]
- 7.Rosenberg AL, Hofer TP, Strachan C, Watts CM, Hayward RA. Accepting critically ill transfer patients: adverse effect on a referral center’s outcome and benchmark measures. Ann Intern Med. 2003;138:882–90. doi: 10.7326/0003-4819-138-11-200306030-00009. [DOI] [PubMed] [Google Scholar]
- 8.Nallamothu BK, Rogers MA, Chernew ME, Krumholz HM, Eagle KA, Birkmeyer JD. Opening of specialty cardiac hospitals and use of coronary revascularization in Medicare beneficiaries. JAMA. 2007;297:962–8. doi: 10.1001/jama.297.9.962. [DOI] [PubMed] [Google Scholar]
- 9.Final report to Congress implementing the DRA provision affecting specialty hospitals. Centers for Medicare & Medicaid Services Web site. 2006 August; Available online at: http://www.cms.hhs.gov/apps/media/press/release.asp?counter=1941.
- 10.The Annual Survey of Hospitals Database: Documentation for 2003 Data. Chicago, IL: American Hospital Association Press; 2003. [Google Scholar]
- 11.American Hospital Directory. Available online at: http://www.ahd.com/freesearch.php3.
- 12.Wennberg DE, Birkmeyer JD, editors. The Dartmouth Atlas of Cardiovascular Health Care. Chicago: American Hospital Association Press; 1999. [PubMed] [Google Scholar]
- 13.McClellan M, McNeil BJ, Newhouse JP. Does more intensive treatment of acute myocardial infarction in the elderly reduce mortality? Analysis using instrumental variables. JAMA. 1994;272:859–66. [PubMed] [Google Scholar]
- 14.Quan H, Sundararajan V, Halfon P, Fong A, Burnand B, Luthi JC, Saunders LD, Beck CA, Feasby TE, Ghali WA. Coding algorithms for defining comorbidities in ICD-9-CM and ICD-10 administrative data. Med Care. 2005;43:1130–9. doi: 10.1097/01.mlr.0000182534.19832.83. [DOI] [PubMed] [Google Scholar]
- 15.Schulman KA, Berlin JA, Harless W, Kerner JF, Sistrunk S, Gersh BJ, Dube R, Taleghani CK, Burke JE, Williams S, Eisenberg JM, Escarce JJ. The effect of race and sex on physicians’ recommendations for cardiac catheterization. N Engl J Med. 1999;340:618–26. doi: 10.1056/NEJM199902253400806. [DOI] [PubMed] [Google Scholar]
- 16.Specialty hospitals: information on national market share, physician ownership, and patients served. Washington, DC: General Accounting Office; Apr, 2003. GAO-04-167. [Google Scholar]
- 17.Cram P, Rosenthal GE, Vaughan-Sarrazin MS. Cardiac revascularization in specialty and general hospitals. N Engl J Med. 2005;352:1454–62. doi: 10.1056/NEJMsa042325. [DOI] [PubMed] [Google Scholar]
- 18.Nallamothu BK, Wang Y, Cram P, Birkmeyer JD, Ross JS, Normand SL, Krumholz HM. Acute myocardial infarction and congestive heart failure outcomes at specialty cardiac hospitals. Circulation. 2007;116:2280–7. doi: 10.1161/CIRCULATIONAHA.107.709220. [DOI] [PubMed] [Google Scholar]
- 19.Casalino LP, Devers KJ, Brewster LR. Focused factories? Physician-owned specialty facilities. Health Aff (Millwood) 2003;22:56–67. doi: 10.1377/hlthaff.22.6.56. [DOI] [PubMed] [Google Scholar]
- 20.Weiner SJ, Laporte M, Abrams RI, Moswin A, Warnecke R. Rationing access to care to the medically uninsured: the role of bureaucratic front-line discretion at large healthcare institutions. Med Care. 2004;42:306–12. doi: 10.1097/01.mlr.0000118706.49341.49. [DOI] [PubMed] [Google Scholar]