Below is the uncorrected machine-read text of this chapter, intended to provide our own search engines and external engines with highly rich, chapter-representative searchable text of each book. Because it is UNCORRECTED material, please consider the following text as a useful but insufficient proxy for the authoritative book pages.
B Literature Review The study committee conducted an extensive review of literature on racial and ethnic disparities in healthcare (discussed in Chapter 1). In this appendix, summary tables of this literature are presented, along with cri- teria used in the conduct of this review. To assess the evidence regarding racial and ethnic differences in health care, the committee conducted literature searches via PUBMED and MEDLINE databases to identify studies examining racial and ethnic differences in medical care for a variety of disease categories and clini- cal services. Searches were performed using combinations of following keywords: â¢ Race, racial, ethnicity, ethnic, minority/ies, groups, African Ameri- can, Black, American Indian, Alaska Native, Native American, Asian, Pa- cific Islander, Hispanic, Latino. â¢ Differences, disparities, care. â¢ Cardiac, coronary, cancer, asthma, HIV, AIDS, pediatric, children, mental health, psychiatric, eye, ophthalmic, glaucoma, emergency, diabe- tes, renal, gall bladder, ICU, peripheral vascular, transplant, organ, cesar- ean, prenatal, hip, hypertension, injury, surgery/surgical, knee, pain, pro- cedure, treatment, diagnostic. This search yielded over 600 citations. To further examine this evi- dence base and address the study charge that called for an analysis of âthe 285
286 UNEQUAL TREATMENT extent of racial and ethnic differences in health care that are not otherwise attributable to known factors such as access to care,â only studies that provided some measure of control or adjustment for racial and ethnic dif- ferences in insurance status (e.g., ability to pay/insurance coverage or co- morbidities) were included in the literature review. Other âthresholdâ cri- teria included: â¢ Publication in past 10 years (1992-2002; this criterion was estab- lished because more recent studies tend to employ more rigorous research methods and present a more accurate assessment of contemporary pat- terns of variation in care); â¢ Publication in peer-reviewed journals; â¢ Elimination of studies focused on racial and ethnic differences in health status (except as it is affected by the quality of health care) and health care access, as well as publications that were editorials, letters, pub- lished in a foreign language, were non-empirical, or studies that controlled for race or ethnicity; and â¢ Inclusion only of studies whose primary purpose was to examine variation in medical care by race and ethnicity, contained original find- ings, and met generally established principles of scientific research (e.g., studies that stated a clear research question, provided a detailed descrip- tion of data sources, collection, and analysis methods, included samples large enough to permit statistical analysis, and employed appropriate sta- tistical measures). In addition, to ensure the comprehensiveness of the review, the com- mittee examined the reference lists of major review papers that summa- rize this literature (e.g., van Ryn, 2002; Geiger, this volume; Kressin and Petersen, 2001; Bonham, 2001; Sheifer, Escarce, and Schulman, 2000; Mayberry, Mili, and Ofili, 2000; Ford and Cooper, 1995). Articles not originally identified in the initial search were retrieved and analyzed for appropriateness of inclusion in the committee â s review. Finally, to ensure that the committeeâs search was not limited to studies with âpositiveâ findings of racial and ethnic differences in care, searches were conducted for studies that attempted to assess variations in care by patient socioeconomic status and geographic region. These studies were included if the researchers assessed racial or ethnic differences in care while controlling, as noted above, for patient access-related factors.
287 B: LITERATURE REVIEW To assess the quality of this evidence base, the committee ranked stud- ies on several criteria: â¢ Adequacy of control for insurance status (studies of patients cov- ered under the same health system or insurance plan were considered to be more rigorous than studies that merely assessed the availability of health insurance among the study population); â¢ Use of appropriate indicators for patient socioeconomic status (e.g., studies that measured patientsâ level of income, education, or other indi- cators of socioeconomic status); â¢ Analysis of clinical data, as opposed to administrative claims data (see limitations of administrative claims data noted below); â¢ Prospective or retrospective data collection (prospective studies were considered to be more rigorous than retrospective analyses); â¢ Appropriate control for patient co-morbid conditions; â¢ Appropriate control for racial differences in disease severity or stage of illness at presentation; â¢ Assessment of patientsâ appropriateness for procedures (e.g., stud- ies that provide primary diagnosis and include well-defined measures of disease status, as in studies of cardiovascular care that assess racial differ- ences in care following angiography) or that compare rates of service use relative to standardized, widely-accepted clinical guidelines; and â¢ Assessment of racial differences in rates of refusal or patient pref- erences for non-invasive treatment. Studies that met the committeeâs âthresholdâ criteria are summarized in Table B-1. As a âsecond levelâ analysis of the quality of evidence regarding ra- cial and ethnic disparities in cardiovascular care, the committee identified a subset of studies that permit a more detailed analysis of the relationship between patient race or ethnicity and quality of care, while considering potential confounding variables such as clinical differences in presenta- tion and disease severity. Several criteria were established to identify these studies, using generally accepted criteria of research rigor and quality. To begin, the committee identified only studies using clinical, as opposed to administrative data, for the reasons cited above. Secondly, the committee identified studies that provided appropriate controls for likely confound- ing variables, and/or employed other rigorous research methods. These
288 UNEQUAL TREATMENT criteria included the use of adequate control or adjustment for racial and ethnic differences in insurance status; prospective, rather than retrospec- tive data collection; adjustment for racial and ethnic differences in co- morbid conditions; adjustment for racial and ethnic differences in disease severity; comparison of rates of cardiovascular services relative to mea- sures of appropriateness; and assessment of patient outcomes. Several caveats should be noted in undertaking this approach. One, studies using clinical data allow researchers to better assess whether dis- parities in care exist and are significant after potential confounding fac- tors such as clinical variation and the appropriateness of intervention are taken into account, but these studies often are limited to small patient samples in one or only a few clinical settings, therefore sacrificing statisti- cal power and potentially underestimating the role of institutional vari- ables as contributing to healthcare disparities. Second, assessments of ra- cial and ethnic differences in patientsâ clinical outcomes following intervention must be made with caution. Patientsâ outcomes following medical intervention reflect a wide range of factors, some of which are unrelated to the intervention itself (e.g., the degree of social support avail- able to patients following treatment) and may vary systematically by race or ethnicity. In addition, a finding of no racial or ethnic differences in patient outcomes (e.g., survival) despite disparate rates of treatment should not be interpreted as demonstrating that disparities in the use of medical intervention are inconsequential. In such instances, researchers should ask whether equivalent rates of intervention might be associated with better patient outcomes among minorities. Finally, this second level of analysis should not be interpreted as suggesting that the larger litera- ture presented above is insufficient to draw conclusions regarding dis- parities in healthcare. Almost all of the individual studies reviewed ear- lier possess limitations, but the collective body of this evidence is robust. Despite these caveats, this second review afforded an opportunity to assess whether racial and ethnic disparities in care remain when racial differences in clinical presentation and other potentially confounding vari- ables are controlled. Studies were considered in this second review only if they met four of six criteria noted above, in addition to the âthresholdâ criteria that studies employ clinical databases. Thirteen studies were iden- tified through this process (see Table B-2). Of these, only two (Leape et al., 1999; Carlisle et al., 1999) found no evidence of racial and ethnic dispari- ties in care after adjustment for racial and ethnic differences in insurance status, co-morbid factors, disease severity, and other potential confounder
289 B: LITERATURE REVIEW as noted above. The remaining studies found racial and ethnic disparities in one or more cardiac procedures, following multivariate analysis. Al- most all studies found that adjustment for one or more confounding fac- tors reduced the magnitude of unadjusted racial and ethnic differences in care. Among the five studies that collected data prospectively, however, all found racial and ethnic disparities remained after adjustment for con- founding factors.
290 UNEQUAL TREATMENT TABLE B-1 Summary of Selected LiteratureâRacial and Ethnic Disparities in Health Care Analgesia Source Procedure/Illness Sample Analyses Todd, Deaton, Assessed racial differences in Retrospective cohort study of DâAdamo, and Goe, receipt of analgesia among 217 patients (127 African 2000 patients seen for extremity American, 90 white) seen in fractures in emergency an emergency department in departments. an urban hospital. Bernabei, Gambassi, Assessed adequacy of pain 13,625 cancer patients (12,038 Lapane et al., 1998 management among elderly white, 1,041 African Ameri- and minority cancer patients can, 163 Hispanic, 107 Asian, admitted to nursing homes. 276 American Indian) dis- charged from hospitals to any of 1,492 Medicare-certi- fied/Medicaid-certified nursing homes in five states.
291 B: LITERATURE REVIEW Analyses Findings Limitations Multiple logistic regressions Nearly three-fourths of white pa- -Moderate sample size. to predict use of analgesia tients (74%) received analgesia, -Racial/ethnic groups by race, controlling for time compared to 57% of African Ameri- other than white and since injury, total time in can patients. The crude risk of African American not the emergency department, receiving no analgesia was 66% sampled. payer status, and need for higher for black patients than white. -One site sampled. fracture reduction. After controlling for covariates, -Retrospective study. whites remained significantly more -Other relevant con- likely to receive analgesia (risk founds such as alco- ratio = 1.7, 95% CI 1.1 to 2.3). hol and drug use not considered. -Few racial/ethnic minority physicians in sample. Logistic regression to pre- More than a quarter of patients in -Small numbers in dict unresolved daily pain, daily pain (26%), as assessed by self- racial/ethnic groups. adjusting for gender, cogni- report and independent raters, -Retrospective, cross- tive status, communication received no pain medication. After sectional study. skills, and indicators of adjustment, African Americans had -Data set not specifi- disease severity (e.g., ex- 63% greater probability of being cally focused on pain. plicit terminal prognosis), untreated for pain relative to whites -Pain assessed by ob- being bedridden, number of (odds ratio = 1.63, 95% CI 1.18 to servational evaluation. diagnoses, and use of other 2.26). Older age, low cognitive -Family members medications. performance, and increased number involved in collection of other medications were also of information to associated with failure to receive varying degrees. any analgesic agent. -No data regarding analgesic dose or frequency of administration.
292 UNEQUAL TREATMENT TABLE B-1 Continued Analgesia Source Procedure/Illness Sample Analyses Cleeland, Gronin, Baez Assessed adequacy of pain 281 minority outpatients (106 et al., 1997 management among minority African American, 94 His- patients receiving care in panic, 16 other minority) with settings that primarily serve recurrent or metastatic cancer minorities vs. patients who at 9 university cancer centers, receive care in settings where 17 community hospitals and few minority patients are practices, and 4 centers that treated. primarily treat minority patients. Ng, Dimsdale, Rollnik, Assessed racial/ethnic differ- 454 (314 white, 37 Asian, 73 and Shapiro, 1996 ences in physicians prescrip- Hispanic, 30 African Ameri- tion of patient-controlled can) consecutive patients analgesia for post-operative receiving patient-controlled pain. analgesia in post-operative period.
293 B: LITERATURE REVIEW Analyses Findings Limitations Compared treatment of Sixty-five percent of patients who -Data regarding pain among this sample reported pain received inadequate race/ethnicity not with a larger, primarily pain medication. Patients treated in available for com- white sample from a previ- settings where the patient population parison group. ous study where partici- was primarily black or Hispanic and -Data collected pants were treated in set- those who were treated at university immediately after tings where fewer than 10% centers were more likely to receive data on the non- of patients were ethnic inadequate analgesia (77%) than those minority compari- minorities. Pain assessed by who received treatment in settings son group col- independent ratings of where patient population was prima- lected. patients and physicians. rily white (52%; p < 0.003). In addi- -No data collected Adequacy of analgesia tion, minority patients were more on ability to pay. estimated by widely ac- likely to be undermedicated for pain cepted measure of treat- than white patients (65% vs. 50%; p < ment of pain. 0.001), and were more likely to have the severity of their pain underesti- mated by physicians. Analysis of variance and No significant differences found in -Relatively small post-hoc LSD-tests using patient rating of pain or amount of numbers of African ethnicity as independent analgesia self-administered. Americans and Asians. variable. Dependent vari- -Sample located at ables include amount of Significant differences in the amount of one site. narcotic prescribed and narcotic prescribed among Asians, -Retrospective study. amount of narcotic self- blacks, Hispanics, and whites (F = -Analyses did not administered. 7.352; p < 0.01). Whites and African control for patient Americans were prescribed more size or primary narcotic than Hispanics and Asians. language. After adjustment for age, gender, pre- operative use of narcotics, health insurance, and pain site, ethnicity persisted as independent predictor of amount of narcotic prescribed.
294 UNEQUAL TREATMENT TABLE B-1 Continued Analgesia Source Procedure/Illness Sample Analyses Todd, Lee, and Assessed racial/ethnic differ- Prospective study of 207 pa- Hoffman, 1994 ences in physicianâs percep- tients (138 white, 69 Hispanic) tions of pain in patients with admitted to ED at UCLA Medi- isolated extremity trauma. cal Center between 1992-1993. Todd, Samaroo, and Assessed ethnic differences in 139 patients (108 white, 31 Hoffman, 1993 receipt of emergency depart- Hispanic) admitted to emer- ment analgesia for isolated gency department at UCLA. long-bone fractures. Patients with recorded alcohol or drug use excluded.
295 B: LITERATURE REVIEW Analyses Findings Limitations Analysis of Covariance to No differences found between non- -Patients enrolled evaluate influence of con- Hispanic and Hispanic patients in study primarily in founding variables on the patient pain assessment, physician early evening and relationship between ethnic- pain assessment, or disparity between weekends. ity and differences in pa- patient and physician pain assess- -Moderate samples tient and physician pain ment. Differences remained non- size. assessment. Independent significant after controlling for -Racial groups variables included occupa- confounds. other than Hispanic tional injury, injury loca- and white not tion, patient pain assess- sampled. ment, physician sex, injury -Single site sampled. type, insurance status, and patient ethnicity. Logistic regression to evalu- 55% of Hispanic patients and 26% of -Retrospective study. ate independent influence of white patients received no analgesic -No control for race/ethnicity on probability (crude relative risk = 2.12, 95% CI covariates such as of analgesic administration. 1.35 to 3.32, p = 0.003). After simulta- precise injury, pres- Independent variables neously controlling for covariates ence of translators. included race/ethnicity, Hispanic ethnicity was strongest -Single site. gender, language, insurance predictor of no analgesia (odds ratio = -Small sample size. status, occupational injury, 7.46, 95% CI 2.22 to 25.04, p < 0.01). -Small number of fracture reduction, time of Hispanics in sample. presentation, total time in -Racial/ethnic ED, hospital admission. groups other than white and Hispanic not sampled.
296 UNEQUAL TREATMENT TABLE B-1 Continued Asthma Analgesia Source Procedure/Illness Sample Analyses 5,062 patients (4,328 white, 734 Krishnan et al., 2001 Race/ethnicity and gender African-American) who partici- differences in consistency of pated in the Outcomes Manage- care with national asthma ment System Asthma Study guidelines within managed between 9/93 and 12/93. care organizations. 464 African-American and Zoratti, Havstad, Assessed racial/ethnic differ- 1,609 white patients treated Rodriguez et al., 1998 ences in treatment for asthma for asthma in a Southeast in a managed care setting. Michigan managed care system (27 ambulatory care clinics).
297 B: LITERATURE REVIEW Analyses Findings Limitations Multivariate logistic regres- After controlling for age, education, -Results may not sion to determine whether employment, and symptom frequency apply to patients race/ethnicity and sex were there were no significant race/ with mild asthma. associated with five indica- ethnicity or sex differences in the use -Bias in self-report tors of National Asthma of medication regimen consistent with data. Education and Prevention NAEPP recommendations for patients -Racial/ethnic Program (NAEPP) guide- with moderate or more severe asthma. groups other than lines (medication, self- white and African- management education, American not control of factors related to sampled. asthma severity, periodic assessment, and asthma specialist care). Regression analysis to African-American patients were more -Racial/ethnic predict use of services, likely than whites to access care in groups other than adjusting for age, gender, emergency rooms (p < 0.001), were African American marital status, and income hospitalized more often (p = 0.023), and white not (as assessed by average and were less likely to be seen by an assessed. income of patientsâ commu- asthma specialist (p = 0.027), after -Use of administra- nity of residence). controlling for income, marital status, tive database. gender, and age. Among only low- -Retrospective income patients, African Americans cross-sectional were more likely to be treated in study. emergency rooms than whites, al- -Number prescrip- though no significant differences were tions filled used as found in access to specialty care and estimate of actual hospitalization rates. After adjusting use. for age, gender, marital status and -No adjustment for income, African Americans were more co-morbidities. likely to use oral corticosteroids (p = 0.026) and were less likely to use inhaled anticholinergic medications (p = 0.016).
298 UNEQUAL TREATMENT TABLE B-1 Continued Cancer Analgesia Source Procedure/Illness Sample Analyses Elston Lafata, Cole Assessed sociodemographic 251 patients (157 white, 94 Johnson, Ben- differences in the receipt of minority [largely African Menachem, Morlock et colorectal cancer surveillance American]) treated for al., 2001 care. colorectal cancer in a man- aged care organization. Farley, Hines, Taylor Racial differences in cervical Retrospective examination of et al., 2001 cancer survival in military 1,553 patient records (65% health system. white, 10% African-American, 8% Filipino, 4% Korean, remaining percentages Japa- nese, Hawaiian, Indian, Asian, Pacific Islander, un- known, or other) from the Automated Central Tumor Registry for the U.S. Military Health Care System between 1988 and 1999. Patients in- cluded were diagnosed with invasive cervical carcinoma. Merrill, Merrill, and Receipt of surgery or radia- Data from 8,119 patients (86% Mayer, 2000 tion therapy among white white, 14% African-American) and African-American with invasive cervical cancer, women with cervical cancer. as obtained from 11 tumor registries in Surveillance, Epidemiology, and End Results (SEER) program.
299 B: LITERATURE REVIEW Analyses Findings Limitations Kaplan-Meier survival Within 18 months of treatment, over -Racial/ethnic analysis to determine cumu- half of the total cohort received a groups other than lative incidence of service colon examination (55%), nearly African American receipt; Cox Proportional three-fourths had received carcino- and white not Hazard models to quantify embryonic antigen (CEA) testing, and examined. the effects of baseline clini- nearly six in ten (59%) received meta- -Retrospective cal and sociodemographic static disease testing. Whites were study. characteristics on risk of more likely than African Americans, -Use of claims data. service receipt. Analyses however, to receive CEA testing adjusted for age, race, (RR = 1.47, 95% CI 1.12 to 2.14) and gender, site and stage of displayed a slight but non-significant original disease, type of trend toward higher rates of colonic treatment, comorbidity examination (RR = 1.43, 95% CI 0.94 index, estimated income. to 2.18). Survival analysis performed No significant difference between the -Small numbers in with Kaplan-Meier survival distribution of age, stage, grade or racial/ethnic curves and log rank tests to histology between African Americans minority groups. determine significant differ- and whites. No difference between -Retrospective ences. Cox proportional these groups found in type of treat- study. hazards regression to assess ment. Differences in five- and 10-year -Administrative factors influencing survival. survival rates were also not statisti- data. Data regarding age at diag- cally significant. nosis, histology, grade, stage, SES, treatment modality obtained. Logistic regression to pre- Overall, 8.03% of whites and 11.64% -Racial/ethnic dict receipt of therapy after of blacks did not receive either radia- groups other than adjusting for stage and tion therapy or surgery. For both African American grade of cancer, patient age, blacks and whites, the odds of not and white not nodal status, histology, and receiving treatment increased with examined. presence of multiple cancer older age and distant and unstaged -Administrative primaries. disease (vs. localized disease). Blacks data. were more likely to be diagnosed -Retrospective unstaged and were less likely to have study. localized disease; once stage was -No controls for accounted for, racial differences in hospital characteris- treatment status became insignificant. tics, appropriate- Among those not treated, blacks were ness, SES. more likely to have treatment not recommended than whites (53.68% vs. 40.32 %). Of those cases not
300 UNEQUAL TREATMENT TABLE B-1 Continued Cancer Analgesia Source Procedure/Illness Sample Analyses Bach, Cramer, Warren, Early stage lung cancer. 10,984 patients (10,124 white, and Begg, 1999 860 African Americans) age 65 and older with resectable stage I or stage II non-small- cell lung cancer. Patients resided in one of 10 study areas of the Surveillance, Epidemiology, and End Results (SEER) program. McMahon, Wolfe, Assessed use of diagnostic All Medicare Part B transac- Huan et al., 1999 and screening procedures tions in the state of Michigan among Medicare Part B from 1986 to 1989 in which eligible population. procedures were used to diagnose colorectal disease. Dominitz, Samsa, Assessed racial/ethnic differ- 3,176 patients (17.9% African Landsman, and ences in receipt of treatment American) with a new diag- Provenzale, 1998 and survival among patients nosis of colorectal cancer. with colorectal cancer in Veterans Administration (VA) health system.
301 B: LITERATURE REVIEW Analyses Findings Limitations receiving therapy, few were due to patient refusal (3.76% among whites, 5.88% among blacks). Kaplan-Meier method used Rate of surgery: 64% for black patients -Relatively small for constructing survival vs. 76.7% for white patients (p < 0.001). sample of African curves with log-rank statis- Five-year survival rate: 26.4% for black Americans. tic used for comparisons. patients vs. 34.1% for white patients -Racial/ethnic Cox proportional-hazards (p < 0.001). However, there was a groups other than method used to adjust for nonsignificant difference in survival white and African confounding variables. rates b/w black and white patients American not Analyses controlled for sex, who underwent surgery and similar examined. income, age, stage of dis- rates for those who did not. This -Retrospective ease, type of Medicaid suggests that lower survival rates study. insurance, and comorbidity. among black patients is largely ex- -Administrative plained by the lower rate of surgical data. treatment. Series of stepwise logistic Assessed contribution of patient age, -Racial/ethnic regression analyses to sex, race, urbanicity of patientsâ com- groups other than predict association between munity, per capita income of commu- white and African procedure utilization and nity, education level of community, and American not patient sociodemographic availability of physicians, internists, and examined. characteristics and resi- gastroenterologists per 100,000 popula- -Administrative dence characteristics. tion to prediction of diagnostic proce- data. dures. African Americans were more -Retrospective likely than whites to receive barium study. enema only (odds ratio = 1.38, 95% CI 1.34 to 1.41), were less likely to receive a combination of barium enema and sigmoidoscopy (odds ratio = 0.80, 95% CI 0.78 to 0.83), and were less likely to receive any colonoscopy (odds ratio = 0.83, 95% CI 0.81 to 0.85). Logistic regression to predict No significant racial differences found -Racial/ethnic likelihood of surgical resec- in rates of receipt of surgical resection groups other than tion, chemotherapy, or (70% among blacks, 73% among whites; African American radiation therapy, after odds ratio = 0.92, 95% CI 0.74 to 1.15), and white not adjusting for patient demo- chemotherapy (23% for both blacks and assessed. graphic characteristics, whites; odds ratio = 0.99, 95% CI 0.78 to -Administrative comorbidities, distant me- 1.24), or radiation therapy (17% among data. tastases, and tumor location. blacks, 16% among whites; odds ratio = -Lack of data on 1.10, 95% CI 0.85 to 1.43). Five-year SES. relative survival rates were similar for black and white patients (42% vs. 39% respectively, p = 0.16).
302 UNEQUAL TREATMENT TABLE B-1 Continued Cancer Analgesia Source Procedure/Illness Sample Analyses Howard, Penchansky, Assessed racial/ethnic differ- 246 women (89 African Ame- and Brown, 1998 ences in survival of breast rican, 157 white) who sought cancer. care for breast cancer in one of three health maintenance organizations (HMOs). Ball and Elixhauser, Colorectal cancer. 20,634 discharges b/w 1980 1996 and 1987 from 500 acute care hospitals in the U.S. Imperato, Nenner, and Assessed variation by race/ Pattern analysis of 4,154 Will, 1996 ethnicity in rates of radical Medicare claims for radical prostatectomy among male prostatectomy to treat pros-
303 B: LITERATURE REVIEW Analyses Findings Limitations Logistic regression to pre- No significant racial differences were -Relatively small dict stage of disease at time found in stage of disease, utilization sample. of diagnosis and Cox sur- of health services before diagnosis of -Racial/ethnic vival analysis to assess breast cancer, or receipt of breast groups other than determinants of survival. examination. African-American African American patients were more likely to die than and white not whites (30% vs. 18%, p < 0.04) and examined. experienced shorter average survival -Retrospective (1.63 years vs. 2.77 years, p < 0.024). review. Two percent of whites and eight percent of African Americans missed two or more appointments following diagnosis; after adjusting for the number of appointments made, Afri- can Americans were more likely than whites to miss appointments. Missed appointments and stage of diagnosis were strongly associated with sur- vival, and reduced the impact of race on survival. Logistic regression to pre- Black and white rates of inpatient -Racial/ethnic dict diagnostic subgroups, mortality were equivalent only for the groups other than procedure types, in-hospital most severely ill. Otherwise, odds of African American mortality. Semilogaraithmic inpatient mortality were 59% to 98% and white not ordinary least squares higher for black patients (odds ratio = examined. regression for length of 1.59 to 1.982, p < 0.05 to p < 0.01). -Use of discharge stay. data. Covariates: patient demo- Procedure type was equivalent only -Retrospective graphics, insurance status, for the sickest patients. Black patients study. clinical factors, and pro- with primary tumor and no evidence vider characteristics. of oncologic sequelae were 41% less likely than whites to receive a major colorectal therapeutic procedure (odds ratio = 0.59, p < 0.001). When metastasis was recorded black pa- tients with primary tumor were 27% less likely to received a major colorectal therapeutic procedure (odds ratio = 0.726, p < 0.05). Pattern analysis of rates of Rates of radical prostatectomy were -Rates for racial/ prostatectomy, relative to lower among African Americans than ethnic groups other incidence of prostate cancer among whites (b/w ratio ranged from than white and
304 UNEQUAL TREATMENT TABLE B-1 Continued Cancer Analgesia Source Procedure/Illness Sample Analyses Medicare patients in New tate cancer between 1991 and York state. 1993. Assessed variations in the use Data for 67,693 men (9.4% Harlan, Brawley, of radical prostatectomy and African American) with Pommerenke et al., radiation to treat prostate localized and regional cancer, 1995 cancer by geographic area, as obtained from Surveil- age, and race/ethnicity. lance, Epidemiology, and End Results (SEER) program database between 1984 and 1991. Assessed long-term survival 1,606 prostate cancer patients Optenberg, Thompson, of black and white prostate (7.5% African American, Friedrichs et al., 1995 cancer patients in Department 92.5% white) who were active of Defense (DoD) medical duty personnel, dependents, facilities. or retirees eligible for care in the military medical system.
305 B: LITERATURE REVIEW Analyses Findings Limitations and Medicare claims for 0.59 in 1991 to 0.86 in 1993; no confi- African American both black and white males. dence intervals provided). not examined. -Retrospective study. -Administrative data. -Analyses did not control for income/ SES, comorbidities or other potential confounds. Chi-square test of associa- Black men aged 50 to 69 years were -Racial/ethnic tion between race and less likely than similarly aged white groups other than receipt of treatment. Tests men to receive prostatectomy. For white and African for trends calculated using black and white men aged 70 to 79 American not Mantel-Haenszel test. years, rates of protatectomy were examined. similar in 1984, but became signifi- -Administrative cantly divergent by 1991, as a larger data. proportion of white men received the -Retrospective procedure (p < 0.01). In 1991, a signifi- study. cantly higher proportion of black men -Adjustment not aged 50 to 59 years received radiation. made for comor- For all age groups in 1991, twice as bidities, SES or many blacks as whites (12.5% vs. other potential 6.6%) received no treatment. confounds. Multiple life-table regres- Blacks presented at a significantly -Racial/ethnic sion analysis to determine if higher stage of cancer development groups other than stage and grade of cancer, than whites (26.4% of blacks present- white and African wait time, age or race affect ing with distant metastases compared American not patient survival. Cox pro- to 12.3% of whites, p < 0.001), and examined. portional hazard function demonstrated a greater percentage of -Administrative used to compute mortality recurrence (30.6% vs. 21.4%, p = 0.02). data. risk ratios for black and There were no significant racial differ- -Retrospective white patients. ences in wait time to receive treat- study. ment, and no significant differences were found in the type of treatment when stratified by stage of presenta- tion. Overall, stage, grade, and age were found to affect survival, but not race. When analyzed by stage, blacks demonstrated longer survival for distant metastatic disease (mortality risk ratio = 0.644, 95% CI 0.396 to 1.036).
306 UNEQUAL TREATMENT TABLE B-1 Continued Cardiovascular Disease Analgesia Source Procedure/Illness Sample Analyses Petersen et al., 2002 Assessed racial differences in Analysis of 606 black and treatment for AMI. 4,005 white VA patients with diagnosed AMI discharged from one of 81 VA hospitals. Bell and Hudson, 2001 Racial and gender differences Analysis of 379 records of in emergency room treatment patients (229 white, 150 of chest pain. African American) presenting to ER with chest pain during one calendar year at two county hospitals in North Carolina. Okelo et al., 2001 Rates of recommendation for Data reviewed for 938 con- coronary revascularization secutive cardiac when race/ethnicity were catheterizations in 882 pa- unknown by physicians. tients (26.5% African Ameri- can, 73.5% white) performed between 1993 and 1995. Cardiologists and cardiothoracic surgeons provided with all clinical and angiographic data without racial identifiers and were asked for revascularization recommendations.
307 B: LITERATURE REVIEW Analyses Findings Limitations Logistic regression to assess No differences between African- -Racial/ethnic use of guideline-based American and white patients in re- groups other than medications, invasive car- ceipt of beta blockers, but African white and African diac procedures, and all- Amercans were more likely to receive American not cause mortality at 30 days, aspirin and were less likely to receive examined. 1 year, and 3 years. thrombolytic therapy at time of ar- -Retrospective data rival and were less likely to receive collection. bypass surgery, even when only high- -Physician, hospital risk coronary anatomic subgroups characteristics not were assessed. No racial differences assessed. found in rates of refusal of invasive treatment. Logistic regression to assess Treadmill: no significant differences. -Racial/ethnic whether treadmill testing, groups other than cardiac catheterization (CC), CC: Whites more likely to receive white and African and echocardiogram (Echo) cardiac catheterization (adjusted odds American not were recommended or ratio = 2.8317, 95% CI 1.7833 to examined. performed. Analysis of 4.4963). -Relatively small covariance to assess wait sample. time to first EKG. Models Echo: African Americans more likely -Retrospective. tested main effects of clinic, to receive Echo (adjusted odds ratio = -Results from gender, race, and insurance, 0.5927, 95% CI 0.377 to 0.931). diagnostic proce- and interactions between dures (e.g., tread- gender and race and be- Time to first EKG: African-American mill stress tests) tween insurance and race. patients waited longer than whites for that may have Number of cardiovascular EKG. explained variance related co-morbid conditions in CC not available. also included in models. Revascularization recom- After adjustments, African Americans -Racial/ethnic mendations compared more likely to have a recommenda- groups other than between African-American tion for PTCA (odds ratio = 1.42, 95% African American and white patients and CI 0.96 to 2.11, p = 0 .08) and less and white not correlated with clinical data. likely to have recommendation for examined. Logistic regression analyses CABG (odds ratio = 0.59, 95% CI 0.37 -Physician, hospital performed for CABG and to 0.94, p = 0.02). characteristics not PTCA. Independent vari- assessed. ables included age, African- American ethnicity, co- morbid disease, LV dys- function, number of coro- nary arteries with significant stenosis, and involvement of specific arteries.
308 UNEQUAL TREATMENT TABLE B-1 Continued Cardiovascular Disease Analgesia Source Procedure/Illness Sample Analyses Schneider, Leape, Assess whether racial differ- Stratified weighted random Weissman et al., 2001 ences in cardiac revasculari- sample of 3,960 Medicare zation are due to âoveruseâ beneficiaries in 173 hospitals of the procedure in white (in five states) who under- patients. went coronary angiography in 1991 and 1992. Watson, Stein, Influence of race and gender Prospective study of 838 Dwamera et al., 2001 on use of invasive procedures patients (443 white men, 264 in patients with acute myo- white women, 79 African- cardial infarction (AMI). American men, 49 African- American women) with AMI seen between January 1994 and April 1995 in five com- munity hospitals in Michigan. Canto, Allison, Kiefe Reperfusion therapy for acute 26,575 Medicare patients et al., 2000 myocardial infarction (AMI). (25,044 white, 1,531 African American) meeting eligibil- ity criteria for reperfusion therapy.
309 B: LITERATURE REVIEW Analyses Findings Limitations RAND criteria used to Rates of inappropriate PTCA ranged -Retrospective determine proportion of from 4% to 24% among study states, study examining coronary artery bypass and 0% to 14% for CABG surgery. medical record and graft (CABG) and percuta- claims data. neous transluminal coro- White men had significantly higher -Racial/ethnic nary angioplasty (PTCA) adjusted odds than African American groups other than procedures that were ap- men of receiving inappropriate PTCA African American propriate, uncertain, or (odds ratio = 2.42, 95% CI 1.02 to not examined. inappropriate. 5.76). No significant differences were found among white women, African- Multivariable logistic regres- American women, and African- sion analysis to assess odds American men. Adjusting for of receiving inappropriate between-hospital effect of race and PTCA or inappropriate gender somewhat reduced higher CABG surgery. Analyses odds of inappropriate PTCA among controlled for age, income, white men. clinical characteristics, and state procedure performed. Inappropriate CABG surgery did not differ by race. Multiple logistic regression Rate of being offered CC (with white -Racial/ethnic to identify predictors of men as reference group), was 0.88 (95% groups other than cardiac catheterization (CC). CI 0.60 to 1.29, p = 0.502) for white African American Of those undergoing CC, women, 0.79 (95% CI 0.41 to 1.5, p = and white not analyses to predict coronary 0.465) for black men, and 1.14 (95% CI examined. artery bypass grafting 0.53 to 2.45, p = 0.733) for black women. -Small sample of (CABG), percutaneous African Americans. transluminal coronary For those receiving CC, the rate of -Single geographic angioplasty (PTCA), or being offered angioplasty was 1.22 location. atherectomy. Analyses (95% CI 0.75 to 1.98, p = 0.416) for -No controls for adjusted for age, hospital white women, 0.61 (95% CI 0.29 to appropriateness or of admission, insurance, 1.28, p = 0.192) for black men, and 0.4 SES. severity of AMI, and (95% CI 0.14 to 1.13, p = 0.084) for comorbidity. Coronary black women. The rate of being offered artery anatomy added as CABG was 0.47 (95% CI 0.24 to 0.89, covariate in analyses con- p = 0.021) for white women, 0.36 (95% ducted among patients CI 0.12 to 1.06, p = 0.065) for black receiving CC. men, and 0.37 (95% CI 0.11 to 1.28, p = 0.118) for black women. Bivariate and multivariate White men were most likely to receive -Study excluded analyses of prevalence ratios reperfusion therapy (59%), followed patients who were to predict use of reperfusion by white women (56%), black men not white or Afri- therapy by race and gender. (50%), and black women (44%). can American. Statistical adjustments for Prevalence ratios (after statistical -No controls for age, medical history, clinical adjustment): socioeconomic status.
310 UNEQUAL TREATMENT TABLE B-1 Continued Cardiovascular Disease Analgesia Source Procedure/Illness Sample Analyses Underuse and overuse of 356 patients (43% white, 27% Carlisle, Leape, Bickel, diagnostic testing for coro- African American, 19% Bell et al., 1999 nary artery disease. Latino, 9% Asian or Pacific Islander) presenting to ER in one of five Los Angeles area hospitals. Patients completed questionnaire asking whether they had received diagnostic testing for coronary artery disease. Patient medical records were also reviewed. Ethnic differences in use of 4,987 patients (3,152 white, Daumit, Hermann, cardiovascular procedures in 1,835 African American) with Coresh, and Powe, patients with end-stage renal end-state renal disease from 1999 disease as they transition to 303 dialysis facilities between Medicare health insurance. 1986 and 1987. Patients were followed for up to seven years. Data obtained from the Case Mix Severity Study of the US Renal Data System.
311 B: LITERATURE REVIEW Analyses Findings Limitations presentation, and hospital WW/WM â 1.00 (95% CI 0.98 to 1.03); -Retrospective characteristics. BW/BM â 1.00 (95% CI 0.89 to 1.13); cohort study. Logistic regression to assess BW/WM â 0.90 (95% CI 0.82 to 0.98); -Study limited to whether education, insur- BM/M â 0.85 (95% CI 0.78 to 0.93). patients presenting ance status, gender, age, Only level of education was associ- to ER. and race/ethnicity were ated with underuse, or inappropriate -Approximately independent predictors of use of diagnostic testing. Underuse 50% of potential underuse or overuse. more likely to occur among patients subjects did not without a college education (odds respond or could ratio = 2.2, 95% CI 1.0 to 4.4). not be contacted. -Issues of colinear- ity among educa- tion, insurance, and race/ethnicity. Logistic regression to assess After adjustment, odds of having a -No controls for effect of race on receipt of a cardiac procedure at baseline were hospital characteris- cardiovascular procedure at nearly three times greater for white tics and availability baseline. Covariates include patients than for African-American of procedures. age, type insurance at base- patients (odds ratio = 2.92, 95% CI -Data obtained line, type of employment, 2.04 to 4.18). from administrative employment status, marital records. status, region of country, During follow-up white patients were -Racial/ethnic coronary artery disease, 1.4 times more likely to have a proce- groups other than history of smoking, choles- dure (adjusted relative risk = 1.41, white and African terol level, triglyceride level, 95% CI 1.13 to 1.77). American not history diabetes, obesity, included. cerebrovascular disease, In patients with Medicare before end- congestive heart failure, stage renal disease, the baseline differ- history malignant condition, ence in procedure use was eliminated low serum albumin level, over follow up (odds ratio = 1.05, 95% and type of dialysis. CI 0.56 to 1.6). Logistic regression also used to identify receipt of proce- Among patients who already had dure during follow-up. Medicare at baseline, the adjusted Cox proportional hazards odds ratio of procedure use for white model used to assess time compared to African-American pa- to receipt of procedure tients was 3.0. At follow-up, no differ- during follow-up for white ence between ethnic groups seen in compared to African Ameri- procedures after hospitalization for can patients. myocardial infarction or coronary disease.
312 UNEQUAL TREATMENT TABLE B-1 Continued Cardiovascular Disease Analgesia Source Procedure/Illness Sample Analyses Gregory, Rhoads, Assess racial differences in 13,690 New Jersey residents Wilson et al., 1999 rates of cardiac procedures, (1,217 African American, relative to availability of 12,473 white) hospitalized hospital-based invasive car- with a primary diagnosis of diac services. AMI. Hannan, van Ryn, Coronary artery bypass graft 1,261 post-angiography pa- Burke et al., 1999 (CABG) surgery. tients (680 white non-Hispanic, 314 African American, 267 white Hispanic), stratified by race and gender, who would benefit from CABG in New York state, according to RAND appropriateness and necessity criteria. Patients identified and tracked for three months Data obtained from clinical data, telephone and mail surveys of patients and physicians, and information from NY Cardiac Surgery Reporting System. Leape, Hilborne, Bell et Assessed use of CABG or 631 patients (44% white, 27% al., 1999 PCTA for patients for whom African American, 29% His- revascularization procedures panic) at 13 New York City
313 B: LITERATURE REVIEW Analyses Findings Limitations Logistic regression to pre- For all patients, the likelihood of -Ethnic/racial dict receipt of catheteriza- receiving catherterization within 90 groups other than tion and PTCA/CABG, days of AMI was significantly greater African American after controlling for patient among those hospitalized in facilities and white not clinical and demographic that provided cardiac services. Blacks examined. factors and availability of were less likely to receive catheteriza- -Retrospective cardiac procedures in hospi- tion than whites (b/w odds ratio = cohort study. tal where patients were first 0.74 for those younger than age 65 -Use of hospital admitted. [95% CI 0.61 to 0.90], 0.68 for those records. age 65 years and older [95% CI 0.56 to -No controls for 0.83]) controlling for age, sex, health SES. insurance status (for those younger than age 65), anatomic location of primary infarct, co-morbidities, and the availability of cardiac services. Similarly, blacks were less likely than whites to receive revascularization procedures within 90 days of admis- sion (b/w odds ratio = 0.63 for those younger than age 65 [95% CI 0.52 to 0.76], 0.69 for those age 65 years and older [95% CI 0.54 to 0.86]), control- ling for patient demographic and clinical factors and availability of cardiac services. Stepwise logistic regression African-American and Hispanic -Results may not be to predict use of CABG patients were significantly less likely representative of within three months. Statisti- to undergo CABG than white non- NYS (in terms of cal adjustments for age, Hispanics. Odds ratios: access by race/ gender, vessels diseased, white/African-American â 0.64 (95% ethnicity and gen- risk status (low, medium, CI 0.47 to 0.87); white/Hispanic â 0.60 der in the state). high), type of insurance, and (95% CI 0.43 to 0.84). -No controls for other clinical characteristics. SES. Logistic regression to assess No significant variations found in rates -Moderate sample of revascularization among African- size. American patients, (72%), Hispanic patients (67%) and white patients (75%).
314 UNEQUAL TREATMENT TABLE B-1 Continued Cardiovascular Disease Analgesia Source Procedure/Illness Sample Analyses were deemed clinically hospitals who met RAND necessary. criteria for necessary revas- cularization. Data obtained by hospital record review. Scirica, Moliterno, Racial/ethnic differences in 2,948 (77% white, 14% black, Every, Anderson et al., care of patients with unstable 4% Hispanic, 1% Asian, 3% 1999 angina. unknown race/ethnicity) consecutive patients with unstable angina admitted to 35 U.S. hospitals in 1996 (GUARANTEE registry). Medical records were re- viewed and questionnaire was completed for each patient. Canto, Herman, Racial/ethnic differences in 275,046 consecutive AMI Williams, Sanderson presenting characteristics, patients (86% white, 3% et al., 1998 treatment, and outcomes in Hispanic, 1% Asian and patients with myocardial Pacific Islander, < 1% Native infarction. American) enrolled in the National Registry of Myocar- dial Infarction 2 from 1994 to 1996. African-American patients not included in analyses.
315 B: LITERATURE REVIEW Analyses Findings Limitations probability that a patient Rates of revascularization were signifi- -Retrospective would receive revasculari- cantly lower, however, among hospitals study. zation as a function of de- that did not provide revascularization -Data obtained by mographic characteristics services (and therefore had to refer record review. and type of hospital. patients to other hospitals) than those -No controls for that did provide revascularization (59% SES. to 76%, difference = 17% [95% CI 8% to 35%]). Logistic regression to assess Nonwhites had higher incidence of -Relatively small independent contribution of hypertension and diabetes. Cardiac number of minori- demographic, insurance, catheterization was performed less ties. and clinical factors in distin- often in nonwhites as compared to -Collapse of minori- guishing white from non- whites (36% vs. 53%, p = 0.001). In ties into one white patients. patients meeting criteria for appropri- category. ate catheterization (by AHRQ guide- -No controls for lines), fewer nonwhites underwent SES. the procedure (44% vs. 61%, p = 0.001) and among these fewer nonwhites had significant coronary stenosis (72% vs. 90%, p = 0.001). Angioplasty and CABG received equally often in white and nonwhite patients, among those catheterized who had indications for revascularization. Logistic regression to assess Hispanics were as likely as whites to -NRMI-2 not ran- factors predicting acute receive thrombolytic therapy. Asian domized sample of reperfusion strategies, and Pacific Islanders were less likely to patients. invasive cardiac procedures, receive this therapy (odds ratio = 0.84, -No available infor- and mortality. Variables 95% CI 0.72 to 0.99). Native Americans mation on SES. include demographics, more likely than whites to receive -Retrospective medical history, cardiac risk thrombolytic therapy (odds ratio = study. factors, chest pain, symp- 1.18, 95% CI 0.90 to 1.54). tom onset to hospital ar- rival, Killip class, pulse, All minority groups as likely as whites systolic blood pressure, to receive coronary arteriography. electrocardiogram, and Hispanics were as likely as whites to hospital characteristics. undergo revascularization procedures, however Asian and Pacific Islanders were less likely to undergo angioplasty (odds ratio = 0.82, 95% CI 0.64 to 1.04) and more likely to have bypass sur- gery (odds ratio = 1.23, 95% CI 0.96 to 1.57). Native Americans were less likely to undergo both angioplasty
316 UNEQUAL TREATMENT TABLE B-1 Continued Cardiovascular Disease Analgesia Source Procedure/Illness Sample Analyses Taylor, Canto, Racial/ethnic differences in Patients from National Regis- Sanderson, Rogers, management and outcome in try of Myocardial Infarction 2 and Hilbe, 1998 patients with Acute Myocar- (NRMI-2). 275,046 patients dial Infarction (AMI). included (86% white, 6% black). Laouri, Kravitz, French Assessed use of CABG and/or 671 patients (55% white, 21% et al., 1997 PTCA for patients for whom Latino, 12% African-Ameri- procedures are deemed clini- can) at six hospitals (four cally necessary following public and two academically coronary angiography. affiliated private hospitals) who met explicit clinical criteria for coronary revas- cularization. Data abstracted from medical records and from patient interviews. Peterson, Shaw, Assessed racial/ethnic differ- Prospective study of 12,402 DeLong et al., 1997 ences in use of coronary white and African-American angioplasty and bypass patients at Duke University surgery among patients with Medical Center (10.3% Afri-
317 B: LITERATURE REVIEW Analyses Findings Limitations (odds ratio = 0.72, 95% CI 0.50 to 1.05) and bypass surgery (odds ratio = 0.63, 95% CI 0.38 to 1.04) than whites. Mortality similar among whites, Hispanics, Asian and Pacific Island- ers, and Native Americans. Logistic regression to Black patients were less likely to assess variables indepen- receive intravenous thrombolytic -NRMI-2 not ran- dently predicting utiliza- therapy (odds ratio = 0.76, 95% CI domized sample of tion of acute reperfusion 0.71 to 0.80), coronary arteriography patients. strategies, invasive cardiac (odds ratio = 0.85, 95% CI 0.77 to -No available infor- procedures, and mortality. 0.95), and coronary artery bypass mation on SES. Variables included age, surgery (odds ratio = 0.66, 95% CI 0.58 -Retrospective race, sex, payer status, to 0.75). No significant differences study. history, chest pain, ST were found in hospital mortality. elevation, MI location and type, symptom onset to hospital arrival, Killip class, pulse, systolic BP, contraindications to throm- bolysis, census region, and hospital characteristics. Assessed underuse of African Americans were significantly coronary revascularization less likely than whites to undergo -Moderate sample relative to RAND/UCLA necessary CABG (b/w odds ratio = size. criteria for necessity of 0.49, 95% CI 0.23 to 0.99), and were -Retrospective revascularization proce- less likely to undergo necessary PTCA study. dure. Logistic regression (odds ratio = 0.20, 95% CI 0.06 to -No controls for analyses evaluated the 0.72). Patients at public hospitals were SES, or hospital effect of gender, ethnicity less likely to undergo PTCA than characteristics. and type of hospital on those at private hospitals (odds ratio CABG or PCTA, or any = 0.10, 95% CI 0.02 to 0.44). revascularization, control- ling for age, clinical pre- sentation, angiographic findings, and ejection fraction. Logistic regression models African Americans were 13% less to predict the likelihood likely than whites to undergo -Racial/ethnic that a patient would un- angioplasty (odds ratio = 0.87, 95% CI groups other than dergo angioplasty or 0.73 to 1.03) and 32% less likely to white and African
318 UNEQUAL TREATMENT TABLE B-1 Continued Cardiovascular Disease Analgesia Source Procedure/Illness Sample Analyses documented coronary dis- can American) with docu- ease. Also assessed whether mented coronary disease. differences were associated with differences in survival rates. Ramsey et al., 1997 Assessed gender and ethnic 1,228 Mexican-American and differences in receipt of white patients hospitalized percutaneous transluminal for myocardial infarction coronary angioplasty (PTCA) (MI). Data collection part of and aortocoronary bypass Corpus Christi Heart Project. surgery (ACBS). Sedlis, Fisher, Tice Assessed racial differences in 1,474 white and 322 African- et al., 1998 receipt of cardiac procedures American patients who had in a VA hospital. undergone catheterization and were likely candidates for surgery or angioplasty. Taylor, Meyer, Morse, Assessed rates of cardiovas- Abstracted chart reviews and Pearson, 1997 cular procedures by race in from 1,441 patients (1,208 white, 155 African American,
319 B: LITERATURE REVIEW Analyses Findings Limitations bypass surgery. Extension undergo bypass surgery (odds ratio = American not of life associated with by- 0.68, 95% CI 0.56 to 0.82). Racial examined. pass surgery calculated by differences were more marked among -Single site. use of proportional-hazards patients with severe disease (48% of -No information regression model. Risk African Americans with severe coro- about patient ratios for black and whites nary disease underwent surgery vs. preferences. compared after adjusting 65% of whites, p < 0.001). Analysis of -No controls for for base-line prognostic survival benefit of surgery also SES. factors. Independent vari- revealed racial differences; among ables included age, sex, patients expected to survive more severity of disease, other than one year, 42% of African Ameri- clinical and co-morbid cans underwent surgery, compared to factors, and insurance. 61% of whites (p < 0.001). Finally, the adjusted five-year mortality rate among patients revealed that African- American patients were 18% more likely than whites to die (odds ratio = 1.18, 95% CI 1.05 to 1.32). Logistic regression to pre- Among only patients who had re- -Single geographic dict receipt of services, after ceived catheterization to determine location. adjusting for age, sex, previ- extent of disease, Mexican Americans -No controls for ous diagnosis of coronary were less likely to receive PTCA, but SES, hospital heart disease, MI, diabetes not ACBS, than whites after adjusting characteristics. mellitus, hypertension, for clinical and demographic charac- occurrence of congestive teristics (odds ratio = 0.65, 95% CI 0.43 heart failure during MI, to 0.99). location and type of MI. Analyses were generated Therapeutic cardiac procedures (sur- -Racial/ethnic from surgical referral con- gery or PTCA) were offered more groups other than ference at VA hospital frequently for white patients (72.9%) African American between 1988 and 1996. than African-American patients and white not Racial differences in confer- (64.3%; odds ratio = 1.497, p = 0.0022). examined. ence recommendation and This difference could not be explained -Single site. patient compliance with by simple clinical differences between -Potential con- recommendations were the two groups. African-American founds such as SES analyzed using Fisherâs patients, however, were more likely not assessed. exact test. than whites to refuse invasive proce- dures (odds ratio = 2.026, 95% CI 1.311 to 3.130). Logistic regression to assess No differences found in rates of cath- -Retrospective eterization procedures between white study. and ânonwhiteâ patients during AMI -Potential con-
320 UNEQUAL TREATMENT TABLE B-1 Continued Cardiovascular Disease Analgesia Source Procedure/Illness Sample Analyses military health services 78 other) with principle or system. secondary diagnosis of AMI in 125 military hospitals. Weitzman, Cooper, Assessed rates of perfor- 5,462 patients (815 of these Chambless et al., 1997 mance of cardiac procedures African-American) in four in relation to gender, race, states (North Carolina, Mis- and geographic location. sissippi, Maryland, and Minnesota) hospitalized for myocardial infarction (MI). Allison, Kiefe, Centor Assess variations in use of Retrospective medical record et al., 1996 medications among African- review of 4,052 patients (3,542 American and white Medi- white, 510 African American) care patients hospitalized hospitalized in all acute care with Acute Myocardial In- hospitals in Alabama with farction (AMI). principle discharge diagnosis of AMI.
321 B: LITERATURE REVIEW Analyses Findings Limitations differences by patient race admission (odds ratio = 0.96, 95% CI founds such as SES, in rates of catheterization or 0.69 to 1.34) or between white and disease severity, revascularization proce- black patients (odds ratio = 1.19, 95% appropriateness not dures, controlling for age, CI 0.80 to 1.78). Similarly, no differ- assessed. gender, cardiovascular risk ences were found in rates of revas- factors, and clinical data cularization (PTCA or CABG) be- relevant to admission for tween white and ânonwhiteâ patients AMI. (odds ratio = 0.90, 95% CI 0.59 to 1.39) or between white and black patients (odds ratio = 1.11, 95% CI 0.65 to 1.89). No differences were found in mortality or rates of readmission within 180 days following initial discharge. However, white patients were significantly more likely than nonwhite patients to be considered for future catheterization (odds ratio = 1.77, 95% CI 1.20 to 2.61). Logistic regression to esti- After controlling for severity of MI -Racial/ethnic mate odds of having diag- and co-morbid conditions, blacks groups other than nostic and therapeutic proce- admitted to teaching hospitals were African American dures performed during an significantly less likely to receive and white not MI event by race, gender, PTCA (b/w odds ratio = 0.4, 95% CI assessed. and type of hospital. 0.2 to 0.6), CABG (b/w odds ratio = -Potential con- 0.4, 95% CI 0.2 to 0.9) or thrombolytic founds such as SES, therapy (b/w odds ratio = 0.5, 95% CI co-morbidities, 0.3 to 0.8). Similarly, blacks admitted appropriateness not to non-teaching hospitals were signifi- assessed. cantly less likely to receive PTCA (b/w odds ratio = 0.5, 95% CI 0.3 to 0.7), CABG (b/w odd ratio = 0.3, 95% CI 0.2 to 0.6) or thrombolytic therapy (b/w odds ratio = 0.5, 95% CI 0.3 to 0.7). -Racial/ethnic Logistic regression to After controlling for patient appropri- groups other than assess rate of receipt of ateness for therapy, age, gender, white and African thrombolysis, beta- clinical characteristics, and hospital American not andrenergic blockade and characteristics, white patients were examined. aspirin, controlling for more likely to receive thrombolytics -Relatively small patient age, gender, clinical than black patients (odds ratio = 0.51, sample of African factors, severity of illness, 95% CI 0.38 to 0.78). No differences Americans. algorithm-determined were found in receipt of beta-blockers -Retrospective study. candidacy for therapy, and (odds ratio = 1.18, 95% CI 0.91 to 1.53) -Data obtained hospital characteristics
322 UNEQUAL TREATMENT TABLE B-1 Continued Cardiovascular Disease Analgesia Source Procedure/Illness Sample Analyses Herholz et al., 1996 Assessed gender and ethnic Discharge data for 982 pa- differences in receipt of tients hospitalized for definite cardiovascular medications or possible MI; data are from on discharge from hospital the Corpus Christi Heart following myocardial infarc- Project. tion (MI). Blustein, Arons, and Assessed variations by race, 5,857 non-Medicare (less than Shea, 1995 payor, and gender in process 65 years of age) patients of care leading up to revascu- admitted to hospitals in larization procedures for California with a principal patients with cardiovascular diagnosis of acute myocardial disease. infarction (AMI). Carlisle et al., 1995 Assessed use of coronary 131,408 patients (89,781 white, artery angiography, bypass 16,509 African American, graft surgery, and angio- 19,218 Latino, and 5,900 Asian) plasty among Los Angeles discharged from L.A. County
323 B: LITERATURE REVIEW Analyses Findings Limitations (e.g., rural vs. urban, teach- or aspirin (odds ratio = 1.00, 95% CI through record ing vs. non-teaching). 0.81 to 1.24) by patient race. review. -No controls for SES. Logit regression to predict Mexican Americans received fewer -Single geographic receipt of medications by medications than whites (odds ratio = region. gender and ethnicity, after 0.62, 95% CI 0.33 to 1.15), even after -No controls for adjusting for age, diagnosis adjusting for clinical and demo- SES, hospital of diabetes mellitus, hyper- graphic characteristics. Mexican characteristics, tension, congestive heart Americans were less likely to receive appropriateness. failure, serum cholesterol almost all major medications, espe- level, and cigarette cially antiarrhythmics, anticoagulants, smoking. and lipid-lowering therapy. Series of chi square and Authors found differences in likeli- -Relatively small regression analyses to hood of receipt of procedures during number of determine likelihood of nearly every phase of treatment for minorities. receipt of services during different racial and payor groups. -Administrative prehospital, intrahospital Whites, those with private insurance, data, lack of clinical (duration of initial hospital- and those with more severe heart detail. ization), interhospital, and disease were more likely to gain initial -Retrospective posthospital (readmission admittance to hospitals providing study. for revascularization follow- revascularization services. Once hospi- ing initial hospitalization) talized, whites, males, those with phases. African-American private insurance, and those with more and Hispanic patients severe disease were more likely to grouped together as actually receive revascularization. âminorityâ due to small These same patterns were observed numbers. among those patients not initially admitted to hospitals offering revascu- larization but who later received revascularization upon re-admittance or transfer. In logistic regression analy- ses to assess odds of receiving revascularization during any admis- sion, whites were more likely to re- ceive revascularization (odds ratio = 1.49 [no CI reported]), as were the privately insured. Series of logistic regression African Americans were less likely -Retrospective. models to assess relation- than whites to receive bypass graft -Administrative ship between use of inva- (odds ratio = 0.62, 95% CI 0.56 to records used. sive procedures and ethnic- 0.69) and angioplasty (odds ratio = -Proxy used for ity, controlling for primary 0.80, 95% CI 0.72 to 0.88). Latinos co-morbidity and income.
324 UNEQUAL TREATMENT TABLE B-1 Continued Cardiovascular Disease Analgesia Source Procedure/Illness Sample Analyses County residents with pos- hospitals following angiogra- sible ischemic heart disease. phy, CABG, or angioplasty. National Hospital Discharge Survey records of 10,348 patients (9,289 white, 159 African American) hospital- ized with AMI. Giles et al., 1995 Assessed race and sex differ- ences in rate of receipt of catheterization, PTCA, or coronary artery bypass sur- gery (CABS). Maynard, Every, Implications of less intensive 420 black and 10,834 patients Martin, and Weaver, use of revascularization in hospitalized for acute myo- 1995 black patients on long-term cardial infarction in metro- survival. politan Seattle from 1988 to 1994.
325 B: LITERATURE REVIEW Analyses Findings Limitations diagnosis, age, gender, were less likely to receive angiogra- insurance type, income phy (odds ratio = 0.90, 95% CI 0.85 to (proxy), co-morbidities, and 0.95). Asian Americans did not differ differences among hospitals from whites in invasive cardiac proce- in volume of invasive dure rates, although all three ethnic procedures. groups were less likely to receive procedures than whites when hospital procedure volume was not controlled. Logistic regression analysis Significant differences by race and -Administrative adjusting for age, type of gender were found after statistical data. health insurance, hospital adjustment and patient matching -Retrospective. size and type, region, in- procedure. With white males as the -No controls for hospital mortality, and referent, black men were less likely to SES. hospital transfer rates to receive catherterization (odds ratio = -May only be able assess differences in rates of 0.67, 95% CI 0.51 to 0.87) or CABS to generalize to procedures by race. Analy- (odds ratio = 0.63, 95% CI 0.44 to patients with more ses also performed to match 0.90), while black women were less severe disease. individuals admitted to the like to receive catheterization (odds same hospital and who did ratio = 0.50, 95% CI 0.37 to 0.68), not undergo a procedure. PTCA (odds ratio = 0.42, 95% CI 0.23 Analyses limited to proce- to 0.76) or CABS (odds ratio = 0.37, dures occurring during 95% CI 0.22 to 0.62). Among only initial hospitalization. those patients who underwent cath- eterization (and therefore had access to a cardiologist), black women were less likely to receive subsequent PTCA or CABS. Logistic regression to assess No significant differences found in -Relatively small racial differences in age- proportion of black and white patients sample of African- adjusted hospital mortality receiving thrombolytic therapy or American patients. and use of revasculari- cardiac catheterization. After adjust- -Racial/ethnic zation. Log rank statistic ing for use of cardiac catheterization, groups other than used to determine differ- percent professionals in census block, African American ences in survival. history of prior coronary surgery, and white not history of angina, use of thrombolytic assessed. therapy, sex, and history of congestive -SES estimated by heart failure, black patients 40% less census blocks. likely to undergo revascularization (odds ratio = 0.60, 95% CI 0.45 to 0.81, p = 0.0008). After adjustment race was not associ- ated with long-term survival.
326 UNEQUAL TREATMENT TABLE B-1 Continued Analgesia Source Procedure/Illness Sample Analyses Peterson, Wright, Racial differences in proce- 33,641 (29,119 white, 4,522 Daley, and Thibault, dure use and survival follow- African American) male 1994 ing acute myocardial infarc- veterans discharged with tion (AMI) within diagnosis of AMI from Janu- Department of Veterans ary 1988 to December 1990. Affairs. Ayanian, Udvarhelyi, Assessed racial differences in 27,485 Medicare Part A en- Gatsonis et al., 1993 rates of coronary revascular- rollees (26,389 white, 1,096 ization following angiogra- African American) who phy and relationship of these underwent inpatient coronary differences to hospital angiography in 1987. characteristics. Whittle, Conigliaro, Racial differences in use of Retrospective study of Good, and Lofgren, cardiovascular procedures in 428,300 male veterans (74,570 1993 Department of Veterans African American, 353,730 Affairs. white) discharged from VA hospitals with diagnoses of cardiovascular disease or chest pain between 1987 and 1991.
327 B: LITERATURE REVIEW Analyses Findings Limitations Logistic regression to assess After adjustment, as compared to -Racial ethnic effect of race on use of car- white patient, African Americans 33% groups other than diac catheterization, coronary less likely to undergo cardiac white and African angioplasty, coronary bypass catheterizations within 90 days of AMI American not surgery, and overall coronary (odds ratio = 0.67, 95% CI 0.62 to 0.72); included. revascularization. Likelihood 54% less likely to undergo coronary -Administrative ratios calculated for 30-day, bypass surgery within 90 days of AMI database. 1-year, and 2-year survival. (odds ratio = 0.46, 95% CI 0.40 to 0.53), -Retrospective Analyses adjust for age, and 42% less likely to undergo angio- study. cardiac complications, num- plasty within 90 days of AMI (odds -No controls for ber of secondary diagnoses, ratio = 0.58, 95% CI 0.48 to 0.66). The SES. previous hospitalization, black/white ratio for any cardiac hospital location, on-site revascularization procedure within 90 availability of cardiac cath- days of AMI was 0.46 (95% CI 0.41 to eterization and bypass sur- 0.52). gery, and year of admission. African Americans more likely to survive 30 days following AMI compared to whites (adjusted odds ratio = 1.18, 95% CI 1.07 to 1.31). No differences found be- tween races for 1 or 2-year survival rates. Logistic regression analyses African Americans were less likely than -Racial/ethnic to predict revascularization, whites to receive a revascularization groups other than controlling for age, sex, procedure (w/b adjusted odds ratio = African American region, Medicaid eligibility, 1.78, 95% CI 1.56 to 2.03). Greater use of and white not principal diagnosis, second- revascularization occurred in public, examined. ary diagnoses, and hospital private, teaching, nonteaching, and -Relatively small characteristics. urban/suburban hospitals, and in sample of African- hospitals where revascularization American patients. procedures were available, as well as in -Administrative hospitals where such procedures were data set. not available, after controlling for -Retrospective patient demographic and clinical fac- study. tors. No significant black/white differ- ences in rates of revascularization were found in rural hospitals. Logistic regression to assess After adjustment, white patients more -Racial/ethnic association or race with use likely than African American patients groups other than of procedures controlling to undergo cardiac catheterization African American for diagnosis, region, age, (odds ratio = 1.38, 95% CI 1.34 to not examined. co-morbidity, marital status, 1.42), angioplasty (odds ratio = 1.50, -Retrospective year of diagnosis, whether 95% CI 1.38 to 1.64), and CABG (odds study of adminis- CABG performed at hospi- ratio = 2.22, 95% CI 2.09 to 2.36). trative data set. tal where diagnosis made. -No controls for admission practices.
328 UNEQUAL TREATMENT TABLE B-1 Continued Cardiovascular Disease Analgesia Source Procedure/Illness Sample Analyses Cerebrovascular Disease Mitchell, Ballard, Assessed rates of tests and Inpatient hospital records of Matchar et al., 2000 treatment for cerebrovascular 17,437 Medicare patients disease: noninvasive cere- (15,929 white and 1,508 brovascular tests, cerebral African American) with a angiography, carotid endart- principal diagnosis of tran- erectomy, anticoagulant sient ischemic attack (TIA). therapy, and probability of receiving care from a neurologist. Oddone, Horner, Racial differences in use of 803 patients (389 African Sloane et al., 1999 carotid artery imaging in American, 414 white) hospi- Veterans Affairs Medical talized in one of four VA Centers. Medical Centers between April 1991 and January 1995
329 B: LITERATURE REVIEW Analyses Findings Limitations Computed state age- and Nationally, CABG rate was 27.1 per -Some veterans in sex-adjusted rates of CABG 10,000 for whites, 7.6 per 10,000 for study obtained care for whites and African African Americans. Racial differences outside of VA. Americans and evaluated were greater in the Southeast, particu- relative to need for care (as larly in non-metropolitan areas. Cor- -Administrative indicated by myocardial relation of CABG rates was signifi- data set. infarction rate) and supply cantly associated with the density of -Racial/ethnic of physicians (as indicated thoracic surgeons and location in the groups other than by the number of thoracic Southeast for whites, but physician white and African surgeons and cardiologists availability and location was not American not per 10,000 persons). correlated with CAGB rates for Afri- examined. can Americans. -Retrospective study. -Limited informa- tion on demo- graphic factors. Logistic regression adjust- After adjusting for patient, illness, and -Racial/ethnic ing for comorbid illness provider characteristics, African Ameri- groups other than (including hypertension and cans were 83% as likely as whites to African American prior history of stroke), receive noninvasive cerebrovascular and white not ability to pay (proxy based testing (95% CI 0.73 to 0.93). Among examined. on dual Medicaid-Medicare those receiving noninvasive testing, -Retrospective eligibility and area of resi- African Americans were 54% as likely study. dence), and other clinical to receive cerebral angiography (95% CI -Administrative and demographic variables. 0.36 to 0.80), and among those receiving data. angiography, the odds of African Americans receiving carotid endarterec- tomy was 0.27 (95% CI 0.09 to 0.78). African Americans were 62% as likely to receive anticoagulant therapy, but this difference not statistically signifi- cant given small number of African- American subjects. African-American patients were 21% less likely to receive care from a neurologist (95% CI 0.69 to 0.90). Logistic regression to deter- African American patients were less -Retrospective mine adjusted odds ratios likely to have an imaging study of study reviewing for receiving any carotid their carotid arteries (22% vs. 45%, p = medical records. artery imaging. Models 0.001). Race remained an independent -Very small number adjust for age, comorbidity, predictor of imaging after adjusting of African Ameri-
330 UNEQUAL TREATMENT TABLE B-1 Continued Cerebrovascular Disease Analgesia Source Procedure/Illness Sample Analyses with ICD-9 diagnoses of either transient ischemic attack, ischemic stroke, or amaurosis fugax. Record review of clinical data. Childrenâs Health Care Weech-Maldonado Parentsâ ratings and reports Reponses for over 9,000 et al., 2001 of pediatric care under Med- children (842 Hispanic, 1,344 icaid Managed Care by race, African American, 131 Asian, ethnicity, and primary 330 American Indian, 6,329 language. white, 111 other) from the National Consumer Assess- ment of Health Plans Bench- marking Database 1.0 Data from 33 HMOs from Arkan- sas, Kansas, Minnesota, Oklahoma, Vermont, and Washington state.
331 B: LITERATURE REVIEW Analyses Findings Limitations for clinical factors (odds ratio = 1.50, cans received linical presentation, antici- 95% CI 1.06 to 2.13). procedure. pated operative risk, and Whites were significantly more likely -Study limited to hospital. to be assessed as appropriate candi- hospitalized dates for surgery using RAND criteria patients. (18% vs. 4%, p = 0.001) because of -No controls for higher prevalence of significant ca- SES. rotid artery stenosis. RR of carotid endarterectomy for whites compared to African Americans was 1.34 (95% CI 0.70 to 2.53). Ordinary least squares Compared with whites, Asian/other -No controls for regression to assess the reported worse care across several other SES character- effect of race/ethnicity, domains [getting needed care istics such as in- (Î² = -8.11, p < 0.05), timeliness of care Hispanic language, and come, occupation (Î² = -18.65, p < 0.001), provider com- Asian language on ratings -No examination of munication (Î² = -17.19, p < 0.001), and reports of care, control- clinical meaningful- staff helpfulness (Î² = -20.10, p < ling for parent age, gender, ness of differences 0.001), plan service (Î² = -10.95, p < education, and childâs in reports and health status. Care domains 0.001)]. English-speaking Asian par- ratings of care. examined include doctor/ ents did not differ significantly from -Mail and tele- nurse rating, health care whites on any reports of care. phone surveys, rating, health plan rating, Spanish-speaking Hispanic parents data did not iden- timeliness of care, provider reported more negative care than tify surveys admin- whites on timeliness of care (Î² = -9.24, communication, staff help- istered in English fulness, and plan service. p < 0.01), provider communication vs. Spanish. (Î² = -4.37, p < 0.05) staff helpfulness (Î² = -6.09, p < 0.05), and plan service (Î² = -6.93, p < 0.001). English-speaking Hispanic parents did not differ from whites on any reports of care. African-American parents scored lower than whites on reports of get- ting needed care (Î² = -3.52, p < 0.05), timeliness of care (Î² = -4.53, p < 0.01), and plan service (Î² = -4.29, p < 0.001). American Indians had worse reports of care than whites for getting needed care (Î² = -9.12, p < 0.05), timeliness of care (Î² = -3.52, p < 0.01), provider communication (Î² = -3.27, p < 0.05), and plan service (Î² = -4.12, p < 0.01).
332 UNEQUAL TREATMENT TABLE B-1 Continued Childrenâ Analgesias Health Care Source Procedure/Illness Sample Analyses Furth et al., 2000 Access to kidney transplant 3,284 patients < 20 years of list. age (1,122 black, 2,162 white) with ESRD who had first dialysis between January 1, 1988, and December 31, 1993. Hampers et al., 1999 Assess whether language Prospective investigation of barriers between patients and 2,467 patient visits to Emer- physicians were associated gency Department between with differences in diagnostic September and December testing and length of stay. 1997 (413 white, 557 African American, 1,284 Hispanic, 124 other, 89 NA). 286 families did not speak English, repre- senting a language barrier for the physician in 209 cases. Zito, Safer, dosReis, Psychotropic medication use. 99,217 African-American and Riddle, 1998 (60,868) and white (38,349) youths ages five through 14, who were Medicaid recipi- ents in the state of Maryland seen in ambulatory settings. Hahn, 1995 Use of prescription Two samples of children: 1) medications. ages one to five (n = 1,347), and 2) ages 6 to 17 (n = 2,155) who had at least one ambula-
333 B: LITERATURE REVIEW Analyses Findings Limitations Cox proportional hazard Controlling for confounders, black -Racial/ethnic analysis to examine inde- patients were 12% less likely than groups other than pendent effect of race on the white patients to be activated on the African American time from first dialysis for kidney transplant wait list (relative and white not ESRD until first activation hazard = 0.88, 95% CI 0.79 to 0.97). In examined. on cadaveric transplant addition, after controlling for con- -Administrative waitlist for index transplant founders, the relative hazard for black data. controlling for confounding patients in the lowest SES quartile -Retrospective study. factors (age, gender, cause being activated on the wait list was -Potential con- of ESRD, SES, incident year 0.84 (95% CI 0.70 to 1.01) compared to founds such as of ESRD, ESRD network, relative hazard of 1.0 (95% CI 0.8 to co-morbidities, facility characteristics). 1.3) for black patients in the highest appropriateness not SES quartile. examined. Mann-Whitney U tests used The presence of a language barrier -No independent or to compare total charges accounted for a $38 increase in family verification among groups. Analysis of charges for testing (F = 14.1, p < 0.001) of language barrier. covariance used to assess and 20 minute longer ED stay (F = 9.1, -No full control for predictors of total charges p = 0.003). complexity of cases and length of ED stay. -No controls for use Race/ethnicity, insurance of professional status, provider training, interpreter or ad patient care setting, and hoc interpreter triage category, patient age, -Single site patient vital signs, included in models to isolate effect of language barrier. Logistic regression to esti- Caucasians were twice as likely to -Racial/ethnic mate the probability of receive psychotropic prescriptions groups other than psychotropic medication compared with African Americans African American use as a function of race and after adjusting for geographic region and white not region. The effect of race (odds ratio = 1.97, 95% CI 1.84 to examined. controlling for region and 2.12). The interaction of race and -One geographic region was significant (Ï2 = 23.3, df = interaction of race and location. region were analyzed. 7, p < 0.001), such that the odds of -Administrative data. receiving psychotropic medications -Retrospective study. differed by geographic region (range -Potential confounds 1.23 to 2.60). such as income, service use, and provider specialties not assessed. Logistic and multiple re- For children ages one to five: -Administrative gression used to assess the 1) Black children (odds ratio = 0.532) data. probability of receiving a were half as likely to receive prescrip- prescription medication and tion medication compared with white
334 UNEQUAL TREATMENT TABLE B-1 Continued Childrenâ Analgesias Health Care Source Procedure/Illness Sample Analyses tory care visit in 1987. Data were obtained from the Household Component of the National Medical Expendi- ture Survey (NMES).
335 B: LITERATURE REVIEW Analyses Findings Limitations children (odds ratio = 1.0) (p < 0.001). Adding health factors to the model did not change relationships. How- ever, addition number of physician visits reduced differences, such that they were no longer significant. There was no difference in the probability of receiving medication for Hispanic children compared with white children. 2) After controlling for age, maternal education, insurance, poverty status, source of care, geographic location, health status, # bed days, # reduced activity days, and physician visits, black children received the fewest number of medications. The average number of medications for black children was 86.5% compared to that of white children, while Hispanic children averaged 94.1% compared to that of white children. For children ages six to 17: 1) Black (odds ratio = 0.536) and Hispanic (odds ratio = 0.621) children were less likely to receive any pre- scription medication compared to white (odds ratio = 1.0) children. The addition of health factors, and num- ber of physician visits did not change these relationships (odds ratio = 0.601, p < 0.001, odds ratio = 0.697, p < 0.01 respectively). 2) After controlling for age, maternal education, insurance, poverty status, source of care, geographic location, health status, # bed days, # reduced activity days, and physician visits, black children received the fewest number of medications. The average number of medications for black children was 89.7% compared to that of white children, and 92.1% for Hispanic children compared to that of white children.
336 UNEQUAL TREATMENT TABLE B-1 Continued Diabetes Analgesia Source Procedure/Illness Sample Analyses Chin, Zhang, and Assessed quality of care and 1,376 African-American and Merrell, 1998 resource utilization among white Medicare beneficiaries African-American and white with diabetes (14% African patients with diabetes. Americans). Emergency Services Lowe et al., 2001 Assessed racial differences in 15,578 African-American and denial of authorization for white patients who sought emergency department (ED) care in an urban hospital care by managed care emergency department. gatekeepers. Baker, Stevens, and Assessed racial differences in 1,049 patients (295 African Brook, 1996 emergency department use. American, 237 white, 517 Hispanic) registered for non- emergency medical problems in the Harbor-UCLA Medical Center Emergency Department.
337 B: LITERATURE REVIEW Analyses Findings Limitations Linear and logistic regres- African-American patients were less -Racial/ethnic groups sion to assess independent likely to have measurement of other than African contribution of race to glycosylated hemoglobin (adjusted American and white health status, quality of odds ratio = 0.65, 95% CI 0.48 to not examined. care, and resource utiliza- 0.88) lipid testing (odds ratio = 0.66, -Confounds such as tion, controlling for sex, 95% CI 0.48 to 0.89), ophthalmologi- hospital characteris- education, and age. Mea- cal visits (odds ratio = 0.72, 95% CI tics, appropriate- sures included patient 0.56 to 0.93), and influenza vaccina- ness, and comorbi- survey, ADA and RAND tions (odds ratio = 0.26, 95% CI 0.19 dities not examined. criteria for quality of to 0.36). care, and Medicare African-American patients were more reimbursement. likely to use the ED (39% vs. 29%, p < 0.01) and had fewer physician visits (8.4 vs. 9.7 visits per year, p < 0.05). In addition, African-American patients had higher reimbursement for home health services, however, once adjusting for case-mix variables race was not associated. Multiple logistic regression After adjusting for patientsâ age, -Racial groups to assess racial differences gender, day, and time of ED visit, other than African in authorization for emer- type of Managed Care Organization American and gency department services. (MCO) and triage category, African white not assessed. Americans were more likely to be -Single site. denied authorization for care (odds ratio = 1.52, 95% CI 1.18 to 1.94). Patients who were covered by a Medicaid MCO (odds ratio = 1.50, 95% CI 1.19 to 1.90) or those covered with MCOs with mixed Medicaid and commercial patient populations (odds ratio = 2.05, 95% CI 1.41 to 2.98) were more likely than those covered by purely commercial MCOs to be denied authorization for care. Logistic regression to assess 19% of African Americans, 13.2% of -Sample obtained at independent effect of race/ whites and 11.3% of Hispanic pa- one site, selective ethnicity on ED use. tients reported two or more previ- enrollment. ous ED visits (in preceding three -Cross-sectional months) (p = 0.01 across groups) survey. (unadjusted odds ratio 1.82 for
338 UNEQUAL TREATMENT TABLE B-1 Continued Emergency Analgesia Services Source Procedure/Illness Sample Analyses Eye Care Devgan, Yu, Kim, and Surgical treatment of glau- Retrospective cohort analysis Coleman, 2000 coma in African-American of 30,495 African-American Medicare beneficiaries. and 160,792 white patients over 65 years of age undergo- ing argon laser trabeculo- plasty or trabeculectomy surgery between 1991 and 1994. Wang, Javitt, and Glaucoma and cataract 642,048 Medicare beneficia- Tielsch, 1997 treatment. ries (606,069 white, 35,979 black) age 65 and older who used eye care services. Patients with physician- diagnosed glaucoma or cataract who underwent surgical treatment.
339 B: LITERATURE REVIEW Analyses Findings Limitations African Americans compared with Hispanics). After adjusting for age, insurance status, regular source of care, and transportation difficulties, ethnicity was not significantly associ- ated with two or more ED visits in the preceding three months (adjusted odds ratio for Hispanics compared with African Americans 1.48, 95% CI 0.95 to 2.3 and adjusted odds ratio for Hispanics compared with whites was 1.22, 95% CI 0.74 to 2.00). Age and sex adjusted rates For each age and age-sex subgroup, -Administrative of argon laser trabeculo- the rate of surgical procedures is data base. plasty and trabeculectomy higher in African Americans com- -Data does not surgery were obtained and pared to whites. The age-sex-adjusted contain information compared with surgery rate ratio was 2.14 (95% CI 2.11 to on beneficiaries rates expected based on 2.16). Assuming treatment should be who may be en- disease prevalence. performed in proportion to age-race rolled in HMOs or prevalence, African Americans under- VA hospitals. went glaucoma surgery at 47% below -Racial/ethnic expected rate (expected rate: 5.52 groups other than procedures per 1,000 person-year of African American enrollment, adjusted rate: 2.95 and white not procedures per 100 person-year analyzed. enrollment). Black-white relative risk of Black patients used eye care services -Administrative having a physician- at two-thirds the rate of white patients database. diagnosed condition and (age gender adjusted RR = 0.67, 95% -Differential pre- surgical treatment were CI 0.66 to 0.68). Black women were sentation for care compared to the expected 73% as likely to use services as white based on severity value based on population women, while black men were 56% as can not be ruled survey data for each specific likely to use services. Among users of out. disease. eye care services, black patients were -Other clinical 2.2 times more likely than whites to confounds may be diagnosed with glaucoma, after exist. adjusting for age and gender (RR = 2.17, 95% CI 2.12 to 2.22). In addition, among users of eye care services, blacks had lower than expected rates of treatment for glaucoma (observed RR = 3.2, 95% CI
340 UNEQUAL TREATMENT TABLE B-1 Continued Eye Care Analgesia Source Procedure/Illness Sample Analyses Gallbladder Disease Arozullah, Ferreira, Racial variation in rate of 16,181 patients (14,249 Cauca- Bennett et al., 1999 adoption of laparoscopic sian and 1,932 African Ameri- cholecystectomy procedure in can) diagnosed with gall Department of Veterans bladder or biliary disease Affairs Medical System. who underwent either Mortality and length of hos- open cholecystectomy or pital stay also examined. laparoscopic cholecystectomy. Data were collected through: a) record review of claims files, and b) prospectively compiled clinical data from records and interview, for the year before the new proce- dure was introduced and the first four years of use of the procedure (1991-1995).
341 B: LITERATURE REVIEW Analyses Findings Limitations 3.1 to 3.4 vs. expected RR of 4.3, 95% CI 3.5 to 5.4), but a higher treatment rate for cataract (RR = 1.2, 95% CI 1.2 to 1.3). Among patients with physi- cian diagnosed glaucoma and cata- ract, black patients were more likely to undergo surgical treatment for these diagnoses than white patients (RR = 1.5 for glaucoma, 95% CI 1.4 to 1.5; RR = 1.2 for cataract, 95% CI 1.2 to 1.3). Modified multiple logistic Claims data indicate that after control- -Administrative regression model to predict ling for confounding variables, Afri- data set. the use of laparoscopic can-American patients who under- -Racial/ethnic versus open cholecystec- went cholecystectomy were 25% less groups other than tomy. Predictors included likely as white patients to undergo African American race, age, marital status, the laparoscopic procedure (adjusted not examined. hospital geographic loca- odds ratio = 0.74, 95% CI 0.66 to 0.83). tion, co-morbid illnesses, The shortening of postoperative and year of surgery. To length of hospital stay (from 9 to < 4.5 examine mortality and days with new procedure) occurred in length of stay, multiple the first year for white patients and in logistic regression equations the fourth year for African-American used. Predictors included patients (p < 0.001). age, gender, marital status, coexisting medical condi- Clinical data indicate that after adjust- tion, geographic region, ment, African-American patients year of care, and type of were 0.68 times as likely to undergo cholecystectomy. the laparoscopic procedure (95% CI 0.55 to 0.84).
342 UNEQUAL TREATMENT TABLE B-1 Continued Analgesia HIV/AIDS Source Procedure/Illness Sample Analyses Shapiro, Morton, Assessed racial/ethnic, gen- Multistage probability sample McCaffrey et al., 1999 der, and other sociodemo- of 2,846 individuals, includ- graphic variations in care ing African-American and (number of care-seeking visits Hispanic patients, using data and use of protease inhibitors from the HIV Costs and [PI] or nonnucleoside reverse Services Utilization Study. transcriptase inhibitors [NNRTI]) for persons infected with HIV. Bennett, Horner, Assessed quality of care for Retrospective chart review of Weinstein et al., 1995 pneumocyctis carinii pneu- a cohort of 627 VA patients monia (PCP) among white, and 1,547 non-VA patients Hispanic and African- with treated or cytologically American patients with HIV confirmed PCP who were receiving care in either hospitalized from 1987 to Veterans Administration 1990. (VA) hospitals or non-VA systems. Moore, Stanton, Assessed use of anti- 838 African-American, His- Gopalan, and retroviral drugs and prophy- panic, and white patients Chaisson, 1994 lactic therapy to treat presenting at an urban HIV pneumocyctis carinii pneu- clinic from March 1990 monia (PCP) in an urban through December 1992. Data population infected with HIV. obtained through interview and record review with six- month follow-up.
343 B: LITERATURE REVIEW Analyses Findings Limitations Logistic regression to pre- Adjusting for insurance status, CD4 -Potential con- dict use of PI and NNRTI, cell count, sex, age, method of expo- founds such as prophylaxis against pneu- sure to HIV, and region of country, co-morbidities, SES mocyctis carinii pneumonia African-American and Hispanic not assessed. (PCP), use of antiretroviral patients were 24% less likely than medication, hospitaliza- whites to receive PI or NNRTI at initial tions, ambulatory visits, and assessment, although this disparity emergency department declined to 8% at the final assessment visits. stage, a difference that remained statistically significant (p = 0.016). On average, blacks waited 13.5 months to receive these medications, compared to 10.6 months for whites (p < 0.001). Logistic regression to pre- For all patients, regardless of the type of -Retrospective dict diagnostic procedures hospital in which they were treated, use study. (use and timing of broncho- of anti-PCP medications was initiated -No controls for scopy) and use and timing within two days of admission for 70% SES, co-morbidities. of PCP medications, con- to 77% of patients. Approximately 60% trolling for insurance status, of patients underwent a bronchoscopy age, sex, risk group status, at some point during hospitalization. severity of PCP illness at Black and Hispanic patients at non-VA admission, use of medica- hospitals were more likely to die during tions prior to admission, hospitalization, and were less likely to type of hospital, and hospi- undergo bronchoscopy in the first two tal volume of patients with days of admission. No racial differences AIDS. were found in use of bronchoscopy, receipt of anti-PCP medications within two days of admission, or mortality in VA hospitals. Logistic regression to pre- No racial differences were found in -Single site. dict receipt of antiviral the stage of HIV disease at the time of -Confounds such as agents or PCP prophylaxis, presentation. However, 63% of eli- comorbidities not adjusting for patient in- gible whites, but only 48% of eligible assessed. come, insurance status, blacks received antiretroviral therapy, mode of HIV transmission, and PCP prophylaxis was received by and place of residence. 82% of eligible whites and only 58% of eligible blacks. African-American patients were significantly less likely than whites to receive antiretroviral therapy (odds ratio = 0.59, 95% CI 0.38 to 0.93) or PCP prophylaxis (odds ratio = 0.27, 95% CI 0.13 to 0.56). Whites were more likely to report a usual source of care (59%) than Afri- can Americans (34%, p < 0.001).
344 UNEQUAL TREATMENT TABLE B-1 Continued Maternal Analgesia and Infant Health Source Procedure/Illness Sample Analyses Aron, Gordon, Cesarean delivery rates. 25,697 women (19,996 white, DiGiuseppe et al., 2000 5,701 nonwhite) with no prior history of cesarean delivery admitted to 21 northeast Ohio hospitals from January 1993 through June 1995. Data were obtained from Cleveland Health Quality Choice. Barfield, Wise, Rust Civilian vs. military outcomes 2,171,147 births for African- et al., 1996 in prenatal care utilization, American and white mothers birth weight distribution, and [79,154 in military hospitals fetal and neonatal mortality (16.2% AA), 2,091,993 in rates. civilian hospitals (9.5% AA)] recorded from 1981 to 1985 in the Maternal and Child Health database compiled by the Community and Organi- zation Research Institute of the University of California â Santa Barbara.
345 B: LITERATURE REVIEW Analyses Findings Limitations Nested (to account for Overall rates of cesarean delivery -Results may clustering of patients in were similar in white and nonwhite reflect regional individual hospitals and (over 90% African-American) patients. characteristics. provide more robust esti- After adjusting for clinical risk factors, -Retrospective mates of variance of group non-white women were more likely to study. effects) logistic regression deliver via cesarean (odds ratio = 1.34, -No assessment of used to yield odds ratios for 95% CI 1.14 to 1.57, p < 0.001). Analy- appropriateness cesarean delivery in non- sis also indicated that insurance status or necessity of white patients relative to independently influences use of cesarean. whites and for patients with cesarean delivery. government insurance or who were uninsured rela- tive to patients with com- mercial insurance. Analyses were adjusted for 39 risk factors. Relative risks and Mantel- Prenatal care utilization: utilization was -Racial/ethnic Haenszel Chi-square analy- lower for black patients than white groups other than ses for stratified compari- patients in both military (RR = 0.79, African American sons were calculated. 95% CI 0.75 to 0.82) and civilian (RR = and white not 0.51, 95% CI 0.50 to 0.52) populations. examined. However, the magnitude of the dis- -Administrative parity was lower in the military popu- data. lation (p < 0.001). -Retrospective study. Birth weight: for military and civilian -Observational groups black patients had higher rates study, no control of very low birth weight and moder- for insurance in ately low birth weight, however, rates civilian group, SES, were significantly lower in the mili- co-morbidities. tary group. For example in the very low-birth-weight category, the rate for black births was lower that the rate for black civilian births (RR = 0.68, 95% CI 0.56 to 0.82). For white pa- tients the military rates of very low birth weight (RR = 0.75, 95% CI 0.65 to 0.87) were also significantly lower than their civilian counterparts. Fetal and neonatal mortality: For military and civilian groups, mortality was significantly higher for black patients. While fetal mortality rates for white
346 UNEQUAL TREATMENT TABLE B-1 Continued Maternal Analgesia and Infant Health Source Procedure/Illness Sample Analyses Braveman, Egerter, Cesarean delivery rates. 217,461 singleton first live Edmonston, and births (15,529 African Ameri- Verdon, 1995 can, 19,142 foreign-born Asian, 62,303 foreign-born Latina, 26,802 U.S.-born Latina, 93,685 white) among women in California in 1991. Brett, Schoendorf, and Use of prenatal care tech- Births among non-Hispanic Kiely, 1994 nologies (ultrasonography, black and non-Hispanic white tocolysis, amniocentesis). women in 1990 (3.1 million available for ultrasonogra- phy, 3.2 million for tocolysis, 37,000 for amniocentesis). Data were obtained from the National Center for Health Statistics.
347 B: LITERATURE REVIEW Analyses Findings Limitations patients were similar for military and civilian groups, rates for black military groups were significantly lower than their civilian counterparts (RR = 0.80, 95% CI 0.65 to 0.99). Multiple logistic regression After adjusting for covariates (insur- -Data collected in to determine adjusted odds ance, personal, community, medical, single region. ratios of cesarean delivery and hospital characteristics), African- -Retrospective by race/ethnicity. American women were 24% more study. likely to undergo cesarean than whites (adjusted odds ratio = 1.24, 95% CI 1.18 to 1.31). U.S.-born Latinas were also at an elevated risk compared to whites (adjusted odds ratio = 1.07, 95% CI 1.03 to 1.12). Among women residing in 25% or more non-English speaking communities, who delivered high-birth weight babies or who gave birth at for- profit hospitals, cesarean delivery was more likely among nonwhites and was over 40% more likely among black women than white women (odds ratio = 1.51, 95% CI 1.20 to 1.89; odds ratio = 1.42, 95% CI 1.21 to 1.67; odds ratio = 1.42, 95% CI 1.20 to 1.68, respectively). Logistic regression was Amniocentesis was used substantially -Racial/ethnic used to estimate likelihood less frequently by black women (ad- groups other than of tocolysis and Mantel- justed RR = 0.58, 95% CI 0.56 to 0.60). African American Haenszel to estimate use of Ultrasonography was received by and white not ultrasonography and am- black women slightly less frequently examined. niocentesis. Confounders than white women (adjusted RR = -Administrative data. controlled for include: 0.88, 95% CI 0.87 to 0.88). Black -Retrospective study. maternal age, education, women with singleton births were -No controls for marital status, location of slightly more likely to receive hospital characteris- residence, birth order, tocolysis than white women (adjusted tics, many prenatal timing of first prenatal care RR = 1.06, 95% CI 1.04 to 1.09), al- care details (e.g., visit, and plural births. though the risk of idiopathic pre-term time of procedure), delivery is estimated to be three times regional differences higher in black women. in practices, appro- Women with plural births received priateness of tocolysis two thirds as often as white procedure. women (adjusted RR = 0.69, 95% CI 0.62 to 0.75).
348 UNEQUAL TREATMENT TABLE B-1 Continued Analgesia and Infant Health Maternal Source Procedure/Illness Sample Analyses Kogan, Kotelchuck, Self-reported receipt of 8,310 women (6,782 white Alexander, and prenatal care advice from non-Hispanic and 1,532 black Johnson, 1994 providers. women) who participated in the 1988 National Maternal and Infant Health Survey conducted by the National Center for Health Statistics. Mental Health Kales, Blow, Bingham Impact of race on mental Retrospective study of 23,718 et al., 2000 health care utilization among patients (859 Hispanic, 3,529 veterans. African American, 19,330 white) age 60 and older hospitalized for psychiatric diagnoses treated in Depart- ment of Veterans Affairs inpatient facilities in 1994. Melfi, Groghan, Antidepressant treatment. 13,065 Medicaid patients Hanna, and Robinson, diagnosed with depression 2000 treated between 1989-1994.
349 B: LITERATURE REVIEW Analyses Findings Limitations -Racial/ethnic Logistic regression to assess After adjustment for covariates, more groups other than contribution of race to white women reported receiving African American mothersâ report of receipt of advice for alcohol (odds ratio = 1.29, and white not advice or instructions dur- 95% CI 1.10 to 1.51) and smoking examined. ing any of their prenatal cessation (odds ratio = 1.20, 95% CI -Data self-report. visits on: breast-feeding, 1.01 to 1.39). Breast-feeding promo- alcohol consumption, to- tion just missed significance with a bacco, and use of illegal trend toward more advice for white drugs. Analyses controlled women. A significant interaction for age, marital status. between race and marital status emerged, such that black single women were 1.4 times more likely than single white women to not re- ceive advice on drug cessation, while there were no racial differences among married women. ANCOVA to test for group After adjustment, African-American -Administrative differences in inpatient patients had significantly fewer out- database. psychiatric variables. patient psychiatric visits (least- -Potential con- Covariates included age, squares means: H = 15.9 visits, AA = founds such as medical co-morbidity, 15.3 visits, W = 22.3 visits, W > AA, medication dosing/ psychiatric co-morbidity, p < 0.02). Similarly, African-American response, treatment and survival months. patients with substance abuse dis- compliance, illness Analyses also performed for orders had significantly more out- course, personal outpatient variable (out- patient psychiatric visits than white resources not patient visits). patients (least-squares means: H = measured. 19.4 visits, AA = 23.2 visits, and W = -Relatively few 13.2 visits, AA > W, p < 0.0001). Hispanics in sample. No significant differences found in inpatient care. Bivariate tests between 44% of whites and 27.8% blacks -Racial/ethnic those who did and did not received antidepressant treatment groups other than receive antidepressants and within 30 days of 1st indicator of African Americans between racial categories. depression (p < 0.001). and whites not Logistic regressions to Whites were more likely to receive assessed. examine determinants of antidepressants than black patients -Administrative receiving antidepressants. (odds ratio = 0.495, 95% CI 0.458 to database. Covariates included age, 0.536, p = 0.0001) and other/unknown -Retrospective gender, Medicaid eligibility racial category patients (odds ratio = study. status, year of initial depres- 0.749, 95% CI 0.627 to 0.880, p = 0.0006). sion, if initial care received Blacks were less likely than whites to
350 UNEQUAL TREATMENT TABLE B-1 Continued Mental Health Analgesia Source Procedure/Illness Sample Analyses Segal, Bola, and Prescription of antipsychotic 442 patients (256 white, 107 Watson, 1996 medications by physicians African American, 47 His- in psychiatric emergency ser- panic, 10 Asian, 22 âotherâ) vices. seen in psychiatric emergency rooms. Data were obtained through observation of evalu- ations and record review. Evaluators were primarily psychiatrists (80%) and white (88%). Chung, Mahler, and Inpatient psychiatric treatment. 164 adults (76 African Ameri- Kakuma, 1995 can, 88 white) admitted to acute inpatient setting with Axis I diagnosis of major mood or psychotic disorders.
351 B: LITERATURE REVIEW Analyses Findings Limitations from mental health pro- receive SSRIs (odds ratio = 0.844, 95% -Information not vider, number of comorbid CI 0.743 to 0.959, p = 0.0093) when available on sever- conditions. prior clinical research suggests that ity of depressive blacks are more susceptible than disorder. whites to side effects of Tricyclics and therefore should be more likely to receive SSRIs. Analysis of covariance More psychiatric medications were -Small number of models constructed using prescribed to African Americans than minorities. other patients (Î² = 0.99, p < 0.005). least-squares regression or -Sites all urban logistic regression to assess public hospitals in the influence of race on five African-American patients received single geographic more oral doses (Î² = 1.21, p = 0.02) prescription practice indica- area. and injections (Î² = 0.54, p = 0.04) of tors. Models controlled for -No controls for presence of psychotic disor- antipsychotic medications. The 24- SES, hospital char- der, severity of disturbance hour dosage of antipsychotic medica- acteristics. (GAS score), dangerousness, tion given to African Americans was psychiatric history, if physi- significantly higher than for other patients (Î² = 862, p < 0.001). cal restraints used, hours spent in the emergency The tendency to overmedicate African- service, clinicianâs efforts to American patients was lower when engage patient in treatment, clinicianâs efforts to engage the pa- if optimum time was spent tients in treatment were rated as being on the evaluation. higher. Models predicting number of medications, number of oral and injected antipsychotic and 24-hour dosage became non-significant. ANOVA and Logistic re- After controlling for diagnosis and -Relatively small gression to assess effects of SES, African-American patients had sample. race, diagnosis (psychotic shorter length of stay (F = 9.12, df = 1, -Single site. vs. nonpsychotic), and 150, p = 0.003). In addition, white -Retrospective socioeconomic status (insur- patients were 3.8 times more likely study. ance status) on treatment. than African-American patients to be -No assessment of Data were obtained through on one-to-one observational status diagnostic validity record review. (95% CI 1.6 to 8.9). Analysis of inter- between the two actions indicated that among high SES groups. patients, African Americans were 3.5 times more likely to receive urine drug screens, regardless of diagnosis (n = 109, 95% CI 1.2 to 10.1).
352 UNEQUAL TREATMENT TABLE B-1 Continued Mental Health Analgesia Source Procedure/Illness Sample Analyses Padgett, Patrick, Burns, 7,768 persons insured by Blue Use of inpatient mental and Schlesinger, 1994 Cross and Blue Shield health services. Associationâs Federal Em- ployees Plan in 1983, who had at least one inpatient psychiatric day and random sample of 5,000 nonusers of mental health services. Peripheral Vascular Disease Guadagnoli, Ayanian, Amputation and leg-sparing 19,236 Medicare patients who Gibbons et al., 1995 surgery for peripheral vascu- underwent amputation or lar disease of the lower leg-sparing surgery at 3,313 extremities. hospitals in the U.S.
353 B: LITERATURE REVIEW Analyses Findings Limitations Logistic regression devel- No significant differences were found -Administrative oped for each ethnic group among blacks, whites and Hispanics data. to predict probability of at in the probability of a psychiatric -Retrospective least one day of psychiatric hospitalization or in number of inpa- study. hospitalization and number tient psychiatric days. -No assessment of of inpatient days. Predictors diagnostic validity. included predisposing factors (education, family size, percentage of county black, Hispanic, or white), enabling factors (region of country, salary, high or low option selected for insur- ance coverage), and need factors (annual medical expenses, familyâs annual medical expenses, other family member receipt of inpatient psychiatric care. Logistic regression to assess Black patients were more likely to -Racial/ethnic odds of amputation and undergo all forms of amputation than groups other than surgery for black relative to were white patients (unadjusted odds African American white patients, controlling ratio = 1.47 to 2.24). White patients and white not for case-mix, region, and were twice (unadjusted odds ratio = examined. hospital characteristics. 0.51) as likely to undergo lower- -Administrative extremity arterial revascularization data. and almost three times (unadjusted -Retrospective odds ratio = 0.35) more likely to study. undergo angioplasty than black pa- -No controls for tients. potential confounds such as SES, disease Among patients with diabetes, black severity, appropri- patients were 58% more likely than ateness. white patients to undergo above the knee amputation (adjusted odds ratio = 1.58, 95% CI 1.32 to 1.90). Black patients who did not have diabetes were twice as likely to undergo the procedure (adjusted odds ratio = 2.13, 95% CI 1.87 to 2.41).
354 UNEQUAL TREATMENT TABLE B-1 Continued Peripheral Analgesia Vascular Disease Source Procedure/Illness Sample Analyses Pharmacy Morrison, Wallenstein, Differences in white and Random sample of 30% (347) Natale et al., 2000 nonwhite neighborhoods in of New York City pharma- pharmacy stocking of opioid cies. Pharmacists surveyed analgesics. via telephone. Physician Perceptions Thamer, Hwang, Racial and gender differences 271 nephrologists (72% white, Fink et al., 2001 in nephrologists recommen- 14% Asian, 5% African dations for renal transplanta- American) surveyed as part tion using hypothetical pa- of the Choices for Health tient scenarios. Outcomes in Caring for ESRD (CHOICE) Study. Survey administered between
355 B: LITERATURE REVIEW Analyses Findings Limitations Among patients with diabetes, blacks were 48% and 32% less likely to un- dergo percutaneous transluminal angioplasty (adjusted odds ratio = 0.52, 95% CI 0.40-0.67) and lower- extremity bypass surgery (adjusted odds ratio = 0.68, 95% CI 0.59 to 0.79), respectively. Among those who did not have diabetes, black patients were 71% less likely to undergo angioplasty (adjusted odds ratio = 0.29, 95% CI 0.23 to 0.37) and 44% less likely to undergo lower-extremity bypass surgery (adjusted odds ratio = 0.56, 95% CI 0.50 to 0.63). Overall, two-thirds of pharmacies that -No controls for Generalized linear model to did not carry any opioids were in differences in phar- assess relationship between predominantly nonwhite neighbor- macy supplies across racial/ethnic composition of hoods. After adjustment pharmacies neighborhoods. neighborhoods and opioid in predominantly nonwhite neighbor- -Sample from one supplies of pharmacies. hoods ( < 40% of residents white) site. Analyses controlled for were significantly less likely to have -Possible reporting proportion of elderly per- adequate opioid supplies than were errors by sons at census-block level pharmacies in predominantly white pharmacists. and crime rates at the pre- neighborhoods (at least 80% residents -Pharmacists only cinct level. white) (odds ratio = 0.15, 95% CI 0.07 questioned about to 0.31). Among 176 pharmacies with opioids recom- inadequate stock, reasons were as mended as appro- follows: 54%âlittle demand for medi- priate first-line cations, 44%âconcern about disposal, medications. 20%âfear of fraud and illicit drug use, 19% âfear of robbery, 7%âother (e.g., problems with reimbursement). Asian males less likely than white -Survey data in lieu Scenarios presented males to be recommended for trans- of treatment data. patientâs age, race (white, plantation (odds ratio = 0.46, (95% CI -Potential bias in African American, Asian), 0.24 to 0.91). Females were less likely response rate. gender, living situation, than males to be recommended (ad- -No controls for treatment compliance, justed odds ratio = 0.41, 95% CI 0.21 patient SES. diabetic status, residual to 0.79). No differences between renal function status, HIV African-American and white patients were found.
356 UNEQUAL TREATMENT TABLE B-1 Continued Physician Analgesia Perceptions Source Procedure/Illness Sample Analyses June 1997 to June 1998. Re- sponse rate 53%. Weisse, Sorum, Racial and gender differences Sanders, and Syat, 2001 in pain management. 111 surveyed primary care physicians from Northeast regions of U.S. who were presented vignettes depicting patients with medical com- plaints, two painful (kidney stone, back pain) and one control (sinusitis). Race and gender of fictitious patients varied. Questions following vignettes assessed physiciansâ aggressiveness in treating symptoms. Effect of race and SES on
357 B: LITERATURE REVIEW Analyses Findings Limitations status, weight, and cardiac ejection fraction. Respond- ing physicians asked if they would recommend trans- plantation given presence of certain criteria. Multiple logistic regression to assess independent effect of neph- rologist and patient factors on decision to recommend transplantation. Analyses adjust for patient and neu- rologist demographics, clinical characteristics, nephrologist training, and organizational affiliations. Analysis of variance to Kidney stone pain: Decision to treat -Small sample size. assess impact of patient with hydrocodone did not vary by -Convenience gender and race on treat- race. Among physicians who opted to sample. ment decision (hydro- treat with medication, dose of hydro- -Physicians in codone dosage). Physician codone selected did not differ by Northeast, limiting age and years in practice patient race (white = 308 mg, African generalizability. included as covariates. American = 271 mg), patient gender, -Approximately or physician gender. Interaction 50% of solicited between physician gender and patient physicians partici- race was found (F 1,85 = 9.65, p = pated. 0.003). Male physicians prescribed -No controls for higher doses to white patients than to physician prescrib- African Americans, while female ing habits. physicians prescribed higher doses to -Racial/ethnic African-American patients. groups other than white and African Back pain: Decision to treat with American not hydrocodone did not vary by race. investigated. Similarly, dose selected did not differ -Few racial/ethnic by patient race (white 188 = mg, minority physicians African American = 233 mg), patient in sample. gender, or physician gender. No interactions were observed. Sinus Infection: Decision to treat with antibiotic did not differ by patient race or gender. White patients were prescribed a longer course of antibiot- ics (X = 13.7 vs. 9.2 days, F1,87 = 4.90,
358 UNEQUAL TREATMENT TABLE B-1 Continued Physician Analgesia Perceptions Source Procedure/Illness Sample Analyses van Ryn and Burke, physician perceptions of 618 patient encounters 2000 patients. at eight New York state hospitals. Assessed physiciansâ recom- mendations for managing chest pain, using vignettes of âpatientsâ that varied only in gender and ethnicity. Patients Perceptions Doescher et al., 2000 Racial and ethnic differences 32,929 patients surveyed in patientsâ perceptions of through the Community their physicians (trust and Tracking Survey, a nationally satisfaction). representative sample sur- veyed 1996-1997.
359 B: LITERATURE REVIEW Analyses Findings Limitations p = 0.03) and were prescribed refills more often (Î§12 = 107 vs. 4.05, p = 0.04). Logistic regression used to Black patients rated less positively -Potential for social regress physician percep- than white patients on several dimen- desirability in tion variables on patient sions including physiciansâ assess- responses. race and SES, controlling for ment of patient intelligence (odds -Finding limited to ratio = 0.51, p â¤ 0.01), feelings of each other and patient age, one state and nar- sex, sickness, depression, affiliation toward the patient (odds row sample of ratio = 0.68, p â¤ 0 .05) and beliefs mastery, social assertive- patients. ness, as well as physician about patientâs likelihood of risk -Use of single-item behavior (odds ratio = 0.58, p â¤ 0.02) age, sex, race, and specialty. measures. and adherence with medical advice -Differences in care (odds ratio = 0.62, p â¤ 0.01). not measured. Logistic regression analysis Physicians were less likely to recom- -Representativeness to assess the effects of mend cardiac catheterization for of sample: partici- âpatientâ race and gender, women than men (odds ratio = 0.60, pants recruited at while controlling for physi- 95% CI 0.4 to 0.9) and African Ameri- national meeting. ciansâ assessment of the cans than whites (odds ratio = 0.60, -Hospital character- probability of coronary 95% CI 0.4 to 0.9). Analysis of race-sex istics where interaction revealed that African- physicianâs prac- American women were significantly ticed unknown. less likely to be referred for catheter- -Underemphasis of ization than white men (odds ratio = subgroup analysis. 0.4, 95% CI 0.2 to 0.7). Analyses adjusted for socio- After adjustment, patients from mi- -Racial/ethnic economic factors. nority groups reported less positive subgroups not perceptions of physicians than white assessed. patients on both scales. -Physician race/ ethnicity or other characteristics not assessed. -Potential for re- sponse bias.
360 UNEQUAL TREATMENT TABLE B-1 Continued Radiographs Analgesia Source Procedure/Illness Sample Analyses Selim, Gincke, Ren, Racial and ethnic differences 401 patients (315 white, 22 Deyo et al., 2001 in use of lumbar spine radio- African American, 4 non- graphs. white Hispanic, 1 âotherâ) with low back pain (LBP) receiving ambulatory care services in VA clinics in Boston area. Patients com- pleted Medical Outcome Study Short Form Health Survey (SF-36), LBP question- naire, comorbidity index, and straight leg raising (SLR) test. Rehabilitative Services Harada, Chun, Chui, Assessed sociodemographic Records of 187,900 hip frac- and Pakalniskis, 2000 and clinical characteristics ture patients (94% white, 4% associated with use of physi- African American, 3% cal therapy (PT) in acute âotherâ) derived from Medi- hospitals, skilled nursing care administrative databases. facilities, or both. Horner, Hoenig, Assessed racial differences in 2,497 African-American and Sloane et al., 1997 utilization of inpatient reha- white Medicare patients bilitative services among hospitalized following stroke elderly stroke patients. at any of 297 acute-care hospitals in five states.
361 B: LITERATURE REVIEW Analyses Findings Limitations Logistic regression to assess At higher levels of back pain, non- -Relatively small race, age, education, in- white patients received more spine sample. come, comorbidities, pain films than did white patients (74% -Small number of intensity, radiating leg pain, vs. 50%, p < 0.01). Among patients African-American SLR, 2 summary scores with positive straight leg raising test, and Hispanic par- from the SF-36 (physical nonwhite patients had more spine ticipants. component summary, films than white patients (23% vs. -Potential bias in mental component sum- 11%, p < 0.01). self-report data. mary) as predictors of -Nonwhite patients obtaining lumbar spine After controlling for clinical charac- combined in radiographs during 12 teristics, race was no longer an inde- analyses. months of follow-up. pendent predictor of lumbar spine -Generalizability of radiograph use. populationâ elderly male veter- ans in Boston area. Logistic regression to pre- African-American patients were less -Relatively few dict PT by pattern of use. likely than whites to receive acute minority patients. Independent variables physical therapy only (b/w odds -Administrative included age, gender, ratio = 0.81, 95% CI 0.73 to 0.89), data. comorbidity index, surgery were less likely to receive therapy in -Retrospective type, fracture type, urinary both acute care and skilled nursing study. incontinence, and hospital facilities (b/w odds ratio = 0.70, 95% -Analysis limited to characteristics. CI 0.65 to 0.76), and were more likely acute hospitalization. to receive no physical therapy at all (b/w odds ratio = 1.30, 95% CI 1.18 to 1.43). Logistic regression to pre- After adjusting for clinical and socio- -Administrative dict utilization of physical economic factors associated with use data. and occupational therapy of physical and occupational -Retrospective by race. therapy, no racial differences were study. found in the likelihood of use of therapy (RR = 1.06, 95% CI 0.89 to 1.27) or time to initiate therapy (African Americans = 6.6 days, whites = 7.4, p = 0.42). Similarly, no racial differences were found in length of physical or occupational therapy in days or as a proportion of hospital stay.
362 UNEQUAL TREATMENT TABLE B-1 Continued Rehabilitative Services Analgesia Source Procedure/Illness Sample Analyses Hoenig, Rubenstein, Racial and other sociodemo- 2,762 African-American and and Kahn, 1996 graphic and geographic white Medicare patients (9% differences in use of physical African American) treated in and occupational therapy in 297 randomly-selected hospi- elderly Medicare patients tals from five states. with acute hip fracture. Renal Care and Transplantation Ayanian, Cleary, Effect of patient preferences 1,392 patients (384 African- Weissman, and on access to renal American women, 354 white Epstein, 1999 transplantation. women, 337 African- American men, 317 white men) with end-stage renal disease who had recently begun to receive maintenance treatment with dialysis in Southern California, Alabama, Michigan, and the mid- Atlantic region of the U.S.
363 B: LITERATURE REVIEW Analyses Findings Limitations Multivariate logistic regres- After controlling for clinical factors, -Small number sion to predict utilization of African-American patients (odds ratio African Americans. physical or occupational = 1.56, 95% CI 1.04 to 2.34) and dual -Retrospective therapy by race, socio- eligible Medicare/Medicaid patients study. demographic variables, (odds ratio = 1.36, 95% CI 1.05 to 1.76) severity of hip fracture, were less likely to receive high-inten- geographic region, and sity physical or occupational therapy. other factors. Data obtained No racial differences were found in through record review. time to initiation of therapy. Measures included inter- African-American patients were -Racial/ethnic views and data from the slightly less likely than white patients groups other than renal networks and the to report wanting a kidney transplant African American United Network for Organ (76.3% African-American women vs. and white not Sharing. Logistic regression 79% of white women, p = 0.13; 80.7% examined. to estimate: 1) the adjusted African-American men vs. 85.5% -Potential bias in relative odds of referral for white men, p = 0.04). However, com- patient recall. evaluation at a transplant pared to preferences, African-Ameri- center; and 2) placement on can patients were much less likely a waiting list for a trans- than white patients to have been plant or receipt of trans- referred to a transplant center for plant within 18 months after evaluation (50.5% of African-Ameri- start of dialysis, for African- can women vs. 70.7% of white American and white men women; and 53.9% for African-Ameri- and women. Analyses can men vs. 76.2% for white men; p < control for patient prefer- 0.001 for each comparison), and to ence and expectations, have been placed on a waiting list or perceptions of care, region, to have received a transplant within age, education, income, 18 months after initiating dialysis insurance, employment, (31.9% African-American women vs. marital status, car owner- 56.5% for white women, and 35.3% for ship, type facility, cause of African-American men vs. 60.6% for renal failure, health status, white men, p < 0.001 for each com- and co-morbidities. parison).
364 UNEQUAL TREATMENT TABLE B-1 Continued Renal Care Analgesia and Transplantation Source Procedure/Illness Sample Analyses Kasiske, London, and Racial/ethnic differences in 41,596 patients registered Ellison, 1998 early placement on kidney with 238 UNOS centers on transplantation waiting list. the national OPTN kidney and kidney-pancreas waiting list between April 1, 1994, and June 30, 1996. Barker-Cummings, Use of peritoneal dialysis as 10,726 patents who began McClellan, Soucie, and initial treatment for end-stage treatment for end-stage renal Krisher, 1995 renal disease (ESRD). disease at dialysis centers in North Carolina, South Caro- lina, and Georgia and who reported to ESRD Network between January 1, 1989, and December 31, 1991. Use of services and proceduresâGeneral Jha, Shlipak, Hosmer Hospital mortality. 39,190 male patients (28,934 et al., 2001 white and 7,575 black) admitted to 147 VA hospitals nation-wide for one of six diagnoses (pneumonia, angina, congestive heart failure, chronic obstructive pulmonary disease, diabetes, chronic renal failure).
365 B: LITERATURE REVIEW Analyses Findings Limitations Logistic regression to assess White patients more likely to be -Retrospective patient and center charac- placed on waiting list before vs. after study utilizing teristics on listing before initiating maintenance dialysis than administrative data. dialysis or registration after non-white patients. Independent -Analyses did not being placed on mainte- predictors of listing before dialysis include measures nance dialysis. included being African American for hospital charac- (odds ratio = 0.465, p < 0.001, refer- teristics of appro- ence: white), Hispanic (odds ratio = priateness. 0.588, p < 0.001, reference: white) and Asian/other (odds ratio = 0.548, p < 0.001, reference: white), in addition to factors including age, prior transplant, level of education, employment sta- tus, insurance status, receiving insu- lin, listed for kidney-pancreas vs. kidney only, and listed in a center with high volume. Logistic regression (back- African Americans were 57% less -Racial/ethnic ward stepwise procedure) likely than whites to be initially groups other than to assess relationship be- treated with peritoneal dialysis (odds African American tween ethnicity and initial ratio = 0.43, 95% CI 0.39 to 0.47). After and white not dialysis modality, con- controlling for confounding character- assessed. trolling for patient istics (age, education, social support, -Potential con- characteristics. home ownership, functional status, founds such as albumin level, hypertension, history hospital characteris- of MI, peripheral neuropathy, and tics, appropriate- comorbid diabetes) the odds ratio of ness not examined. initial treatment for African Ameri- cans compared with whites was 0.45 (95% CI 0.38 to 0.52). -Racial/ethnic Principle outcome was Mortality at 30 days was 4.5% in black groups other than mortality at 30 days. Sec- patients and 5.8% in white patients African American ondary outcomes were in- (RR = 0.77, 95% CI 0.69 to 0.87, p = and white not hospital and 60-day mortal- 0.001). Mortality for black patients examined. ity. Analysis included was lower for each of the six diag- -Administrative logistic regression for inpa- noses. Adjustments for patient and data. tient mortality and Cox hospital characteristics had a small -Retrospective Proportional hazard models effect (RR = 0.75, 95% CI 0.66 to 0.85, p study. for 30-day and 6-month < 0.001). Black patients also had lower -Confounders such mortality to estimate the in-hospital and 6-month mortality. as illness severity,
366 UNEQUAL TREATMENT TABLE B-1 Continued Use of services and proceduresâGeneral Analgesia Source Procedure/Illness Sample Analyses Tai-Seale, Freund, and Effect of mandatory enroll- Data from Medicaid eligibil- LoSasso, 2001 ment in managed care (MC) ity, claims, and MC encounter on service use among African data from two counties in one American compared to white state where one county im- Medicaid beneficiaries. plemented âfreedom-of- choiceâ waiver enrolling its Medicaid beneficiaries in MC, and one county not involved in the waiver. In the waiver county, 3,490 adults and 3,414 children from pre-period (12 months prior to enrollment); 4,082 adults and 3,834 chil- dren in post-period. In non- waiver county, 2,087 adults and 2,093 children in pre- period and 1,200 adults and 1,200 children in post-period. Approximately half sample in each group was African American. Andrews and Ethnic differences in receipt Data from 1.7 million (88% Elixhauser, 2000 of major therapeutic proce- white, 12% Hispanic) hospital dures during hospitalization. discharges. Data from 1993 discharge abstracts from Healthcare Cost and Utiliza- tion Project State Inpatient Database for California, Florida, and New York. Weinick, Zuvekas, and Racial and ethnic magnitude Data from three national Cohn, 2000 of disparities in use of health databases (1977 National care services from 1977 to Medical Care Expenditure 1996. Survey, 1987 National Medi- cal Expenditure Survey, 1996 Medical Expenditure Panel Survey).
367 B: LITERATURE REVIEW Analyses Findings Limitations independent association of admissions prac- race with mortality. tices not assessed. Count data models adjusted African-American beneficiaries had -Racial/ethnic for nonrandom selection fewer visits to physicians than white groups other than within difference-in differ- beneficiaries after mandatory enroll- African American ence (DD) econometric ment. This held for both adults and white not approaches. (DD = -1.937, p < 0.01) and children assessed. Services assessed include (DD = -0.813, p < 0.01). No differences -Use of administra- physician visits, emergency found for inpatient admissions. Afri- tive data. department visits, and in- can-American children had a signifi- -Using different patient admissions. Differ- cant increase in use of emergency samples in pre- and ence-in difference method rooms (DD = 0.116, p < 0.01). post-waiver periods. used to identify the pro- -Data from two gram effect of mandatory In analyses controlling for racial differ- counties in one enrollment in managed care ences in trends of service use that were state. on use of services. unrelated to managed care, but may -Disproportional have biased difference-in-difference enrollment of estimates, results indicate that African- African Americans American adults (DD = -2.463, p < 0.01) in HMOs. and children (DD = -1.098, p < 0.01) had lower levels of relative service use. Increases in emergency department visits for African-American children not evident. Decrease inpatient service use found for African-American adults (DD = -0.039, p < 0.05). Logistic regression to assess Hispanics less likely than non-His- -Administrative effect of ethnicity on likeli- panics to receive major procedures for database. hood of receiving therapeutic 38% of 63 conditions and more likely -Data could not procedure for 63 conditions. to receive procedures for 6.3% of examine differences Analyses controlled for age, conditions. between Hispanic gender, disease severity, subgroups. health insurance, income of patientâs community, and hospital characteristics. Outcomes analyzed in- In 1996, blacks were 2.1 percentage -Administrative cluded usual source of care, points more likely than whites to lack data bases. probability of having at least a usual source of care (p < 0.10) and -Retrospective one ambulatory care (AC) Hispanics were 9.9 percentage points study. visit, and average number of more likely than whites to lack a -Need and appro- visits for those indicating usual source of care (p < 0.001). Dis- priateness of ser- AC services. Other variables parities increased from 1977 to 1996, vices not examined.
368 UNEQUAL TREATMENT TABLE B-1 Continued Use of services and proceduresâGeneral Analgesia Source Procedure/Illness Sample Analyses White-Means, 2000 Use of services (paid care- Data are from the National giver, therapist, mental Long Term Care Survey. health, dentist, foot doctor, 527 black and 4,007 white optometrist, chiropractor, ER disabled elderly Medicare visit, doctor visits, prescrip- recipients. tion medications) by disabled elderly. Khandker and Simoni- Prescription drug utilization. 487,922 black and 341,274 Wastila, 1998 white Georgia Medicaid enrollees in 1992. 76% of black and 84% of white en- rollees received prescriptions through Medicaid on an outpatient basis. Harris, Andrews, and Influence of race (African Discharge abstract data on Elixhauser, 1997 American and white) and 1,727,086 discharges (87.9% gender on likelihood of hav- white, 12.1% African Ameri- can, 63.6% female, 36.4%
369 B: LITERATURE REVIEW Analyses Findings Limitations examined included insur- particularly among Hispanics. Ad- ance coverage, family justed analyses indicate that the income, age, sex, marital disparity for Hispanics increased by status, education, health 6.5 percentage points (p < 0.01). The status, region of country, disparity for blacks decreased 3.2 and residence in or outside percentage points (p < 0.05) during of metropolitan area. Used this time period. regression-based difference- indifference approach to 50-75% of disparities would remain if examine change in dispari- disparities in income and insurance ties over time, controlling for coverage were eliminated. variables listed above. Regression analysis to Given similar medical conditions, -Racial/ethnic estimate relative influence black patients are less likely to use groups other than of health conditions and services, particularly prescription African American financial resources on racial medications and physician services. and white not patterns of community Black patients who live in rural areas, examined. long-term care services. small cities, and western states or who -Administrative Models include measures of have more joint and breathing prob- data. medical conditions and lems are more likely to use services. -Retrospective disabilities, income, insur- Differences in personal attributes (i.e., study. ance status, regional and income, health) do not fully explain rural residence, whether racial differences in use of prescrip- unpaid caregivers provide tions and physician services. in-home services, and sociodemographic charac- teristics (gender, education). Model estimating black- Black children used 2.7 fewer pre- -Racial/ethnic white differences in use and scriptions compared to white chil- groups other than level of use of prescription dren. Black adults used 4.9 fewer African American drugs controlling for age, prescriptions, and black elders used and white not sex, and Medicaid eligibility 6.3 fewer prescriptions than white examined. characteristics. elders (all significant at the 99% level). -Administrative data. White Medicaid enrollees had higher -Retrospective study. use and spending than black enrollees -SES and clinical across most high-volume therapeutic factors not examined drug categories. as potential confounds. Logistic regression to assess African Americans were less likely -Racial/ethnic independent effect of race than whites to receive major therapeu- groups other than and gender on likelihood of tic procedures in 37 of 77 (48.1%) white and African having a major procedure conditions. They were more likely
370 UNEQUAL TREATMENT TABLE B-1 Continued Use of services and proceduresâGeneral Analgesia Source Procedure/Illness Sample Analyses ing a major therapeutic or male) from the Hospital diagnostic procedure. Cost and Utilization Project (HCUP-2) for 1986. Hospitals include national sample of 469 facilities. Giacomini, 1996 Gender and ethnic differences Retrospective analysis of data in hospital-based procedure on 7,249 hospital discharges utilization. in California between 1989 and 1990.
371 B: LITERATURE REVIEW Analyses Findings Limitations (identified using ICD-9- than whites to receive a major thera- American not CM codes). Analyses peutic procedure in 9.1% of condi- examined. controlled for influence of tions. There was no significant -Retrospective personal (age, expected difference in 42.8% of disease catego- study. pay source, indicators of ries (alpha = 0.05). Similarly, African -Administrative clinical condition) and Americans were less likely to receive data. hospital-level characteris- a major diagnostic, without therapeu- tics (e.g., bed size, public tic, procedure in 20.8% of conditions, ownership, teaching hospi- more likely to receive diagnostic tal, urban location). procedure in 13% of disease catego- ries. There were no significant differ- ences between races in 66.2% of categories. Females were less likely than males to receive major therapeutic procedures for 32 of 62 (52%) conditions. Females were less likely to receive a major diagnostic, without therapeutic, pro- cedure in 26% of conditions. Patterns emerged with respect to conditions for which there were race and gender differences. For example, African Americans had lower rates than whites and women had lower rates than men for many trauma categories. Logistic regression to White patients were more likely than -Administrative estimate likelihood of African Americans to receive kidney data. obtaining procedure as transplantation (odds ratio = 3.05, -Retrospective function of ethnicity and 95% CI 2.27 to 4.17), defibrillator study. gender. Analyses con- implant (odds ratio = 2.86, 95% CI -Potential con- trolled for insurance sta- 1.28 to 6.25), CABG (odds ratio = 2.44, founds including tus, age, principal diagno- 95% CI 2.08 to 2.78), endarterectomy measures of SES, sis, and number of (odds ratio = 2.27, 95% CI 1.41 to appropriateness of co-morbidities. Odds ratios 3.70), and angioplasty (odds ratio = services, hospital calculated for following 2.00, 95% CI 1.79 to 2.22). characteristics not procedures: heart trans- assessed. plant, kidney transplant, Whites were more likely than Latino extracorporeal shockwave patients to receive angioplasty (odds lithotripsy, hip replace- ratio = 1.72, 95% CI 1.56 to 2.22), ment, carotid endarterec- kidney transplantation (odds ratio = tomy, CABG, PTCA, pace- 1.58, 95% CI 1.20 to 2.08), and CABG
372 UNEQUAL TREATMENT TABLE B-1 Continued Use of services and proceduresâGeneral Analgesia Source Procedure/Illness Sample Analyses Gornick, Eggers, Reilly et Assessed racial differences in mortal- 26.3 million Medi- al., 1996 ity and use of services among a Medi- care beneficiaries care population. (24.2 million whites, 2.1 million African Americans) aged 65 years or older. Phillips, Hamel, Teno et al., Assessed racial differences in use of: 9,105 hospitalized 1996 operation, dialysis, pulmonary artery adults (79% white, catheterization, endoscopy, bronchos- 16% African Ameri- copy, and hospital charges. can, 3% Hispanic, 1% Asian) in five geographically diverse teaching hospitals, with one of nine illnesses associated with average 6-month mortality of 50%. Data collected through chart review and inter- views with patients and physicians.
373 B: LITERATURE REVIEW Analyses Findings Limitations maker implant, and auto- (odds ratio = 1.49, 95% CI 1.35 to matic cardioverter-defibril- 1.67). lator implant. Whites were more likely than Asian patients to receive endarterectomy (odds ratio = 2.08, 95% CI 1.18 to 3.85) and angioplasty (odds ratio = 1.30, 95% CI 1.15 to 1.47). Asians were more likely than whites to receive hip replacement (odds ratio = 0.47, 95% CI 0.29 to 0.77). Malesâ odds of receiving most proce- dures exceeded those of females. Multiple regression to B/w differences found in: -Racial/ethnic predict utilization rates by mortality: 1.19 men (p < 0.001), 1.16 groups other than race-specific median in- women (p < 0.001) African American come, age, gender, and hospital discharges: 1.14, p < 0.001 and white not interaction of race and ambulatory care visits: 0.89, p < 0.001 examined. income. bilateral orchiectomy: 2.45, p < 0.001 -Administrative data. amputations of lower limbs: 3.64, p < -Retrospective study. 0.001 -Factors such as Adjusting for differences in income clinical, hospital reduced differences, but not characteristics not significantly. assessed as poten- tial confounds. Logistic regression to assess Black patients utilized significantly -Highly selective independent effect of race fewer resources than patients of other sample. on procedure use, control- races (odds ratio = 0.70, 95% CI 0.6 to -Data on SES vari- ling for age, gender, educa- 0.81). The median adjusted difference ables not available tion, income, type insur- in hospital cost was $2,805 lower for for all subjects. ance, severity of illness, black patients (95% CI $1,672 to $3,883 functional status, study site, less). Results remained significant and other confounding after adjusting for physicianâs percep- variables tions of patientsâ prognosis.
374 UNEQUAL TREATMENT TABLE B-1 Continued Use of services and proceduresâGeneral Analgesia Source Procedure/Illness Sample Analyses Wilson, May, and Assessed racial differences in Records of nearly 300,000 Kelly, 1994 receipt of total knee arthro- Medicare recipients who plasty among older adults underwent total knee arthro- with osteoarthritis. plasty between 1980 and 1988. Escarce, Epstein, Racial differences in use of 1986 physician claims data for Colby, and Schwartz et medical procedures among 1,204,022 Medicare enrollees al., 1993 Medicare enrollees. (1,109,954 whites and 94,068 African Americans). Indi- viduals enrolled in HMOs excluded. Vaccination Schneider et al., 2001 Magnitude of racial differ- Data from 1996 Medicare ences in influenza vaccination Current Beneficiary Survey. in managed care vs. fee-for- 13,674 Medicare beneficiaries service insurance. (12,414 white, 1,260 African American).
375 B: LITERATURE REVIEW Analyses Findings Limitations Natural logarithm transfor- The prevalence of symptomatic os- -Racial/ethnic groups mation method to estimate teoarthritis of the knee was lower other than African confidence intervals for among whites than blacks, although American and white white-to-black ratios of this difference was non-significant. not examined. rates of total knee replace- African Americans, however, were -Administrative data. ment. less likely than whites to receive total -Retrospective study. knee arthroplasty (odds ratios ranged -Clinical, SES, from 1.5 to 2.0 for women, 3.0 to 5.1 hospital factors, for men). This disparity persisted at appropriateness not each of five levels of income strata. explored as con- founds. Mantel-Haenszel method to Whites more likely than African -Racial/ethnic calculate white-black rela- Americans to receive 23 of 32 services groups other than tive risks, adjusting for age (white-black RR > 1.0, p < 0.05). For African American and sex. example, whites were 1.5 to 2.0 times and white not as likely to receive eight of the study assessed. services, 2.0 to 3.0 times as likely to -Administrative receive three of the services, and more data. than 3.0 times as likely to receive -Retrospective coronary bypass, coronary angio- study. plasty, and carotid endarterectomy. -Potential con- founds such as SES African Americans were more likely and clinical and than whites to receive seven services hospital characteris- (white-black RR < 1.0, p < 0.05). For tics not assessed. example, African Americans more than 1.5 times as likely to receive laser trabeculoplasty, glaucoma surgery, and retinal photocoagulation. Percentage of respondents Both whites and African Americans -Racial/ethnic (adjusting for SES, clinical had higher rates of vaccination under groups other than comorbidities, and care- managed care, however racial dispar- African American seeking attitudes) who ity was not reduced under managed and white not received vaccination and care. examined. magnitude of racial dispar- -Potential bias in ity in vaccination was After adjustment, the racial disparity self-report data. calculated, comparing in fee for service was 24.9% (95% CI patients with managed care. 19.6% to 30.1%). The disparity in managed care was 18.6% (95% CI 9.8% to 27.4%). Both disparities were statistically significant, however the
376 UNEQUAL TREATMENT TABLE B-1 Continued Analgesia Use of services and proceduresâGeneral Source Procedure/Illness Sample Analyses Womenâs Health Brown, Perez-Stable, Hormone Replacement 8,986 women (50% white, Whitaker, Posner et al., Therapy (HRT). 20.2% Asian, 14.7% African 1999 American, 8.6% Latina, 6.3% Soviet immigrant) seen in the general internal medicine, family medicine, and gynecol- ogy practices at UCSF between January 1, 1992, and November 30, 1995. Marsh, Brett, and Hormone replacement 25,203 sampled visits made by Miller, 1999 therapy (HRT). women (age 45-64, 16.4% by black and 83.6% by white women). Data were obtained from the National Health Care survey.
377 B: LITERATURE REVIEW Analyses Findings Limitations and those with fee-for- absolute percentage point difference service insurance. in racial disparity between the man- aged care and fee-for-service groups (6.3%, 95% CI -4.6% to 17.2%) was not. Logistic regression was Compared to white women, all other -Single site. used to calculate odds of groups were less likely to be pre- -Retrospective prescribing HRT for each scribed HRT after adjusting for age, review. ethnic group using whites income, diabetes, hypertension, CHD, -Data not available as the reference group. and osteoporosis. Asians (odds ratio = on variables such Predictor variables were 0.56, 95% CI 0.49 to 0.64), African as education, age, income, and clinical Americans (odds ratio = 0.70, 95% CI menopausal symp- diagnosis. 0.60 to 0.81)), Latinas (odds ratio = toms, hysterectomy 0.70, 95% CI 0.58 to 0.84), and Soviet status, etc. immigrants (odds ratio = 0.14, 95% CI -Physician recom- 0.10 to 0.20) were each less likely to mendations or receive a prescription for HRT than patient characteris- were white women. Women with tics not assessed. osteoporosis were also more likely to receive HRT. Logistic regression used to While physician visit rates were equal -Racial/ethnic examine whether any previ- for black and white women, the rate groups other than ously identified racial dif- of visits per year in which HRT was African American ferences in HRT could be prescribed to white women (odds and white not attributed to known con- ratio = 0.38, 95% CI 0.32 to 0.45) was examined. founders (age, source of more than twice the rate for black -Retrospective payment for visit, drugs women (odds ratio = 0.17, 95% CI 0.12 study. other than HRT, whether to 0.23) in this age group. -Limited informa- physician had previously tion on patient seen patient, physician or characteristics. clinic specialty type, site of care, region of practice, obesity, duration of visit, physician sex).
378 UNEQUAL TREATMENT TABLE B-1 Continued Analgesia Health Womenâs Source Procedure/Illness Sample Analyses Burns, McCarthy, Mammography. 3,187,116 women (7% black, Freund, Marwill et al., 93% white) ages 65 and older 1996 receiving Medicare who re- sided in one of the following states, Alabama, Arizona, Connecticut, Georgia, Kansas, New Jersey, Oklahoma, Penn- sylvania, Oregon, or Washing- ton. Women had received bilateral mammography. Data were obtained from HCFA database for 1990.
379 B: LITERATURE REVIEW Analyses Findings Limitations Logistic regression to pre- In every state, at each primary care -Racial/ethnic dict mammography use visit level (one, two, or three or more groups other than according to age, number of visits) black women had mammogra- African American primary care visits, income, phy less often than white women and white not state of residence for black (even across income levels). Age, examined. and white women in each income, and state adjusted logistic -Administrative state. models reveal that among white data. women, primary care use has a sig- -Retrospective nificant effect on use of mammogra- study. phy: for one visit odds ratio = 2.73, 95% CI 2.70 to 2.77, for two visits odds ratio = 3.98, 95% CI 3.93 to 4.03, for three or more visits odds ratio = 4.62, CI 4.58 to 4.67. Results for black women reveal an analogous, but weaker effect: for one visit odds ratio = 1.77, CI 1.67 to 1.87, for two visits odds ratio = 2.49, CI 2.36 to 2.63, for three or more visits odds ratio = 3.15, CI 3.04 to 3.25.
380 UNEQUAL TREATMENT TABLE B-2 Selected Studies Exerting Control Over Key Clinical Characteristics Type Prospective/ Adjust for: Disease Author Year of Data Insurance Retrospective Comorbidities? Severity Petersen 2002 Clinical VA healthcare Retrospective Yes et al. system Conigliaro 2000 Clinical VA healthcare Retrospective Yes et al. system Carlisle et al. 1999 Clinical Statistical adjust- Retrospective No records ment for type of and ED insurance logs Daumit et al. 1999 Clinical ESRD Medicare Prospective Yes Hannan et al. 1999 Clinical Statistical adjust- Prospective Yes ment for type of insurance Leape et al. 1999 Clinical Statistical adjust- Retrospective No and labora- ment for type of tory data insurance from medi- cal records Scirica et al. 1999 Clinical Statistical adjust- Prospective Yes ment for type of insurance Canto et al. 1998 Clinical Statistical adjust- Retrospective Yes ment for payor status
381 B: LITERATURE REVIEW ust for: Disease Assessed Find orbidities? Severity Approriateness Outcomes? Disparities? Yes Yes Yes â no overall Yes, black patients with AMI were differences in equally likely as whites to receive mortality found. beta-blockers, more likely than whites to receive aspirin, but were less likely to receive thrombolytic therapy at time of arrival and were less likely to receive bypass surgery, even when only high-risk coronary anatomic subgroups were assessed. No racial differ- ences in refusal rates for invasive treatment. Yes Yes No Yes, especially when CABG was deemed ânecessary.â No Yes No No, only lack of post-high school education was significant predic- tor of underuse. Yes Yes Yes Yes, but diminished with insur- ance eligibility. Yes Yes No Yes, African-American patients less like to undergo CABG than whites, considering RAND criteria. Yes Yes No No significant racial or ethnic differences after accounting for hospital type and necessity of revascularization. No Yes No Yes, among patients meeting criteria for appropriate catheter- ization, fewer nonwhites received catheterization. Yes No Yes Non-African-American minorities less likely to receive beta-blocker TX at discharge, but as likely to receive intravenous thrombolytic therapy (except Asian/Pacific Islanders) and undergo coronary arteriography and revasculariza- tion procedures as whites. No differences in hospital mortality.
382 UNEQUAL TREATMENT TABLE B-2 Continued Type Prospective/ Adjust for: Disease Author Year of Data Insurance Retrospective Comorbidities? Severity Taylor et al. 1998 Clinical Statistical adjust- Retrospective Yes ment for payor status Laouri et al. 1997 Clinical Not assessed, but Retrospective Yes and labora- patients sampled with patient tory data from both public follow-up from medi- (where patients cal records are likely insured) and private hospi- tals (patients likely uninsured). Maynard et al. 1997 Clinical Statistical adjust- Prospective Yes ment for payment by Medicaid Peterson et al. 1997 Clinical Statistical adjust- Prospective Yes data ment for type of insurance Taylor et al. 1997 Clinical Statistical adjust- Prospective Yes data ment for payment type of insurance
383 B: LITERATURE REVIEW ust for: Disease Assessed Find orbidities? Severity Approriateness Outcomes? Disparities? Yes No Yes Yes, African Americans less likely to receive intravenous throm- bolytic therapy, coronary arteriog- raphy, and CABG than whites. No differences in hospital mortality. Yes Yes No Yes, significant underuse of revas- cularization procedures among African Americans and patients at public hospitals. Yes No Yes Despite less intensive use of revas- cularization procedures in African Americans, long-term survival after AMI was similar to whites. Yes Yes Yes African Americans less likely than whites to receive bypass surgery, but no differences found in angio- plasty. Differences in treatment most pronounced among patients with severe disease. Differences in treatment associated with lower survival among African Americans. Yes Yes Yes African Americans less likely than whites to receive bypass surgery, but no differences found in angio- plasaty. Differences in treatment most pronounced among patients with severe disease. Differences in treatment associated with lower survival among African Americans.