The Role of Racial and Ethnic Data Collection in Eliminating Disparities in Health Care
Allen Fremont and Nicole Lurie*
Racial and ethnic disparities in health have been extensively documented. While the causes are both numerous and diverse, disparities in health care have been shown to play a substantial role. A recent Institute of Medicine report (IOM, 2003) exhaustively catalogued disparities in care and concluded that important differences were present even among groups that were similarly insured. Many observers now conclude that eliminating racial and ethnic disparities in health and health care is a central issue in overall efforts to improve quality (IOM, 2001; Bierman et al., 2002). Consistent with this view, Healthy People 2010 specified elimination of such disparities as one of its two overarching goals (U.S. Department of Health and Human Services, 2000b).
Making progress toward the goal of eliminating disparities will require widespread, reliable, and consistent data about the racial and ethnic characteristics of the U.S. population. This information is needed to identify the nature and extent of disparities, to target quality improvement efforts, and to monitor progress. Tracking the racial and ethnic composition and changing health care needs of different populations is vital if our health care system, which includes both public health and the delivery of personal health care services, is to fulfill its essential functions. Measurement, reporting, and benchmarking are critical to improving care.
Despite widespread public perception that the federal government and the private sector collect vast amounts of data, the availability of racial and ethnic data in the health care system itself is quite limited. A variety of government sources include data on race and ethnicity, but the utility of these data is constrained by ongoing problems with reliability, completeness, and lack of comparability across data sources. With only a few exceptions, private insurers and health plans do not maintain data on the race or ethnicity of their enrollees.
In this paper, we provide a framework for describing the role of racial and ethnic data in supporting essential functions of the health system. We first illustrate the value of racial ethnic data collection by describing ways such information can be used to reduce disparities, particularly with respect to the quality of care. We describe how data on primary language and socioeconomic status can complement racial and ethnic information. We then assess current sources of racial and ethnic information and the challenges inherent in collecting it. We conclude with a series of recommendations for enhancing the availability and use of data in the public and private sectors.
Throughout our discussion, we emphasize the federal role in data collection. However, since only about half of the minority population in the United States receives care in a public-sector system (e.g., Medicare, Medicaid, Department of Veterans Affairs, Department of Defense), the importance of private-sector and other government efforts should not be overlooked. Nevertheless, successful fulfillment of the federal role will be essential to facilitate state, local, and private-sector initiatives in public health, service delivery, and research.
THE CHALLENGE OF IDENTIFYING HEALTH DISPARITIES
The United States is becoming increasingly diverse. White Americans currently constitute nearly 70 percent of the population. However, by 2050, persons of color will make up nearly half of the population. In some states, such as California, this transition has already occurred, and the proportion of California’s population that is Hispanic is expected to grow dramatically in the next decade. Asian/Pacific Islanders still constitute only a small proportion of the U.S. population, but they currently have the largest rate of population growth (U.S. Bureau of the Census, 1990). These demographic shifts have profound implications for health and health care in this country because minority populations experience a disproportionate burden of health problems.
Overall, African Americans continue to have some of the worst health outcomes. However, discussion of health disparities among racial and ethnic minorities must move well beyond comparisons of African Americans
and whites. Indeed, there is considerable variation in health status among all of the major racial and ethnic groups including whites, African Americans, Hispanics, Asian/Pacific Islanders, and Native Americans/Alaska Natives. For example, while rates of diabetes are disproportionately high among African Americans, American Indians, and Hispanics, the prevalence of diabetes among Asians is less than that for whites (National Center for Health Statistics, 2001). There also can be considerable variation within racial and ethnic subgroups. For example, although Hispanics experience lower overall mortality rates than whites, Puerto Ricans have higher infant mortality rates than whites (National Center for Health Statistics, 2000). Some racial and ethnic subgroups have increased burdens of specific diseases. For instance, Vietnamese American women have cervical cancer mortality rates many times higher than those for other Asian and white women (IOM, 1999).
At present, the sources of such disparities remain unclear, but a wide range of explanatory factors have been suggested, including sociocultural, socioeconomic, behavioral, and biological risk factors, and environmental living conditions (Robert and House, 2000; Fremont and Bird, 2000; Williams, 1999). For example, minority populations as a whole tend to have lower socioeconomic status (SES) than other groups, and low SES is associated with poorer health, independent of race or ethnicity (Gornick, 2002). It is also generally agreed that differences in access to care, including preventive services, and racial and ethnic differences in the quality of care obtained contribute to observed disparities in health. In some minority groups and subgroups the prevalence of various conditions is especially high. Thus, the benefits of improved care for these groups may be substantially more than for others.
The challenge of understanding variations in health between and among racial and ethnic groups is further heightened as more Americans are of mixed racial and ethnic backgrounds. Although only a small proportion of respondents identified themselves as belonging to more than one racial and ethnic group on the latest census, the number of individuals in this group is expected to increase. The Office of Management and Budget (OMB, 1977) has issued guidance and developed a way to bridge the changes that should help examinations of changes over time.
THE ROLE OF RACIAL AND ETHNIC DATA IN SUPPORTING THE ESSENTIAL FUNCTIONS OF THE HEALTH CARE SYSTEM
The health system serves many important functions, but for the purposes of this paper we focus on three, with a particular emphasis on the last: ensuring the health of the population, ensuring equitable access to care, and ensuring quality of care. Admittedly, the system does not perform
optimally in any of these areas, but it performs especially poorly in each of these areas for minority racial and ethnic populations. Data on race and ethnicity are therefore essential for improving performance for each of these functions. Documenting the extent and types of problems and identifying populations at particular risk is a crucial first step to improving performance. When available, these data convey critical information to both providers and policymakers.
Ensuring the Health of the Population
The ability to provide consistent and reliable epidemiological data on the incidence and prevalence of various health conditions and related risk factors among different racial and ethnic populations is essential to ensuring the health of the population. It also supports the rationale for allocating health care resources and developing appropriate public health interventions. For example, examination of data on race and ethnicity revealed that rates of HIV infection were rising more rapidly among African Americans and Hispanics than in any other racial and ethnic group (Shapiro et al., 1999).
Targeting Risk Factors
Risks at the individual or community level such as smoking, unsafe sexual practices, or environmental exposures are irrefutable contributors to poor health outcomes. Racial and ethnic data in federally conducted health surveys such as the CDC Behavioral Risk Factor Surveillance System (BRFSS) and others enable public health officials to better characterize the distribution of such risk factors among different racial and ethnic groups and identify emerging problems.
Access to Care
Access to care is a prerequisite for entering and staying in the health care system. Available racial and ethnic data have been used to document important differences in access between racial and ethnic groups. For example, Hispanics are substantially more likely to be uninsured than whites, and African Americans are more likely to have public insurance than whites (Collins, Hall, and Neuhaus, 1999; Hoffman and Pohl, 2000). Among blacks and whites, rates of insurance were relatively constant during the 2 decades between 1977 and 1996, but during the same period the proportion of Hispanics who were uninsured increased substantially (Weinick, Zuvekas, and Cohen, 2000).
Even when minority individuals have health insurance, they are more likely to experience barriers to care and are less likely to utilize certain types of services. For example, minority patients are less likely to report having a regular source of care (Collins, Tenney, and Hughes, 2002; Doty and Ives, 2002). Conversely, they are more likely to use the emergency room or to be hospitalized for ambulatory care-sensitive conditions such as congestive heart failure (IOM, 2003). Such utilization may reflect poorer care. Finally, because minority patients overall tend to have greater actual need for services, apparently equivalent care between racial and ethnic groups may signify underutilization by minority patients if case-mix issues are not taken into account (Lurie, 2002).
Quality of Care
The IOM report Crossing the Quality Chasm highlighted considerable gaps between current standards of care and the quality of care that patients actually receive. These gaps were particularly pronounced for racial and ethnic minorities. Many observers now conclude that eliminating racial and ethnic disparities is a core issue in improving quality of care (IOM, 2001).
Numerous studies have documented racial and ethnic disparities in care (see IOM, 2003, for an exhaustive review). For example, using RAND/ UCLA appropriateness criteria, Laouri and colleagues (1997) showed that African Americans were only half as likely to undergo a needed coronary artery bypass graft and one-fifth as likely to undergo a percutanerous transluminal coronary angioplasty. Similarly, Ayanian and colleagues (1999) have shown that African Americans with end stage renal disease were considerably less likely to receive a referral for a renal transplant than comparable whites, even when patient preference was taken into account.
Several recent studies have also shown racial and ethnic disparities in performance on Health Employer Data and Information Set (HEDIS) process measures, which are widely used in managed-care settings (Schneider, Zaslavsky, and Epstein, 2002; Fremont et al., 2002; Virnig et al., 2002). All three studies showed that black patients were substantially less likely than whites to receive indicated care such as an annual hemoglobin A1c test in diabetics. Virnig and colleagues (2002) also documented disparities between other racial and ethnic groups. Racial and SES disparities were also observed for several intermediate outcome measures including control of lipids and hemoglobin A1c in diabetics and blood pressure in hypertensive enrollees (Fremont et al., 2002).
An increasing number of studies have assessed disparities in quality of care by using surveys or qualitative methods to elicit patient reports about their care. Such studies have documented significant differences in how patients from different racial and ethnic groups experience the care they
receive and the kinds of barriers they encounter in accessing it. For example, in a recent Commonwealth Fund study, Asian Americans were least likely to report that their doctors understand their backgrounds and values (48 percent) compared to Hispanics (61 percent), African Americans (57 percent) and whites (58 percent) (Collins, Tenney, and Hughes, 2002). Asians were also least likely to report a great deal of confidence in their doctor. Hispanics, regardless of language skills, were more likely than other patients to report having difficulty communicating with and understanding their doctor (33 percent Hispanics and 16 percent whites) (Doty and Ives, 2002). African Americans were nearly twice as likely as whites to report being treated with disrespect during a recent visit (Collins, Tenney, and Hughes, 2002). These sorts of findings have stimulated public and private provider efforts to ensure culturally competent care and to provide language-appropriate services for their patients (IOM, 2002).
USING DATA ON RACE AND ETHNICITY TO REDUCE DISPARITIES IN HEALTH AND HEALTH CARE
The discussion above has focused on ways in which available racial and ethnic data can support essential functions of the health system by identifying populations at risk for particular conditions or with special needs, and documenting disparities in access to and quality of specific types of care. However, simply collecting racial and ethnic data and describing disparities in health and health care does not automatically lead to reductions in disparities. This information needs to be used in ways that stimulate development and implementation of efforts to effectively eliminate disparities. Thus, in this section we highlight some potential uses of racial and ethnic data to promote improved health and health care in minority populations.
Refining Public Health Initiatives and Enhancing Access to Care
Knowing which racial and ethnic population groups are most at risk can help more effectively target public health efforts. For example, documentation of high rates of HIV among African Americans and Hispanics has helped stimulate the development of federal programs that target minority groups at high risk, particularly those in low-income communities (Shapiro et al., 1999). In many instances public health efforts, such as educational campaigns, may not work equally well across different racial and ethnic groups. For instance, an apparently successful mass media campaign to educate the public about the importance of placing infants on their side or back to reduce risk of sudden infant death syndrome was subsequently shown to be far less effective among black mothers than white mothers (Malloy, 1998) largely because the educational messages were not
appropriately focused on black women. Such evaluations require racial and ethnic data and are essential to refining public health efforts. At the very least they can reinforce the need to tailor public health efforts to meet the needs of different racial and ethnic groups.
Racial and ethnic data can also be used to facilitate programs designed to improve access to care. For example, African Americans and Hispanics with cancer typically face substantial barriers to obtaining and completing treatment. Their rates of participation in clinical trials involving state-of-the-art treatment protocols are especially low (National Cancer Institute, 2003). Such data have prompted the development of “Patient Navigator” programs in which culturally and language-appropriate individuals with special training are matched with patients at risk to educate them about screening and prevention measures or guide them through the treatment process once they are diagnosed. Although further evaluation is needed, such programs show promise for reducing disparities in access to cancer care and outcomes (U.S. House of Representatives, 2002).
Improving the Quality of Care
An important national strategy for improving quality of care has been the promotion of accountability for quality (IOM, 2001; Berwick, 1998). Measurement and reporting are essential components of this strategy. For example, widely used quality monitoring programs such as the National Committee for Quality Assurance (NCQA) Health Employer Data and Information Set (HEDIS) have been shown to improve performance on key quality measures among participating health plans (NCQA, 1999).
Currently, however, most health plan quality improvement efforts, including the NCQA program, are not focused on reducing racial and ethnic disparities in care (Fiscella et al., 2000). Several studies have shown that managed care alone is an insufficient mechanism to eliminate disparities in care. Thus, routine reporting of widely used quality measures separately by race and ethnicity provides an excellent opportunity to identify disparities within health plans and to apply quality improvement principles to reduce them. Performance measures are typically reported as averages across all eligible patients but are not broken down by racial and ethnic group. Consequently, important disparities within and between plans are not recognized or addressed. A number of recent studies have shown that it is feasible to obtain and use information on enrollee race and ethnicity with selected HEDIS measures to detect racial and ethnic and socioeconomic disparities (Schneider, Zaslavsky, and Epstein, 2002; Virnig et al., 2002; Fremont et al., 2002; Nerenz et al., 2002).
Since the publication of the IOM’s report To Err Is Human (2000), reducing medical errors and improving patient safety have emerged as ma-
jor areas of focus for policymakers, providers, and researchers. Though research in this area is still new, early studies show differences in the extent and nature of medical errors and problems with safety experienced among patients from different racial and ethnic groups (Burstin, 1993). Thus, just as with quality of care in general, efforts to identify and eliminate problems with patient safety can be enhanced by taking into account possible racial and ethnic differences of patients at risk.
Stimulating Value Purchasing
As reflected in the forthcoming National Quality Report, many policy makers believe that encouraging consumers and employers to base purchasing decisions on quality of care will ultimately lead to better quality and lower costs. Experts believe that information about quality is most useful to consumers when the information pertains to care received by people like themselves. Thus, reporting measures of care for specific racial and ethnic subgroups can strengthen the ability of consumers—both individuals and employers who purchase care on their behalf—to rationally choose health care providers.
In addition, individuals and advocacy groups can use racial and ethnic data in negotiations with providers, health departments, and elected officials to hold them accountable for results and to develop additional programs and policies to address disparities. In this vein, such data may ultimately have uses in domains outside the personal delivery system. For example, in some communities, examining data on diabetes prevalence and outcomes for blacks and Hispanics has revealed the need for more community-based opportunities for safe exercise. Advocacy groups have used this information in working with local officials to build walking paths and recreational facilities.
Similarly, employers and other purchasers can use racial and ethnic data to ensure that they are getting good value for their premiums. For example, after learning of the IOM report on disparities, the benefits specialist at a large national employer became concerned that her workforce, which is largely minority, may not be receiving care of appropriate quality (or quantity). She now systematically queries plans during renewal negotiations and monitors actions they are taking to reduce disparities.
Understanding the Underlying Causes of Disparities and What to Do about Them
The routine collection of data on patient race and ethnicity can also help researchers disentangle factors underlying health care disparities (IOM, 2003). Recent research efforts such as the AHRQ-sponsored Excellence
Centers to Eliminate Ethnic/Racial Disparities (EXCEED) are beginning to clarify causal factors and effective interventions. However, the development of such knowledge is likely to proceed slowly without the availability of additional racial and ethnic data with which to engage a wider group of researchers. (Also see the discussion of language preference and socioeconomic factors below.) Since causal factors and effective interventions may vary across settings, the availability of such data will also be crucial to enabling quality improvement teams to identify specific factors underlying disparities and appropriate interventions in their respective organizations. Provider organizations can enhance their own efforts by sharing best practices for reducing disparities.
As previously discussed, access to care and an array of community-level factors also have important influences on health. Expanding the availability of racial and ethnic data is a clear prerequisite both to a better understanding of how such factors affect different groups and to the development and evaluations of interventions to address them.
Ensuring Compliance with Civil Rights Law
Routine monitoring of access, use of services, and key processes and outcomes of care by race and ethnicity is essential to ensuring compliance with civil rights laws and detecting evidence of discrimination. Whether these practices are intentional or not, whether they are at the level of an individual practitioner or due to system-level problems, they can produce harmful outcomes (IOM, 2003). Title VI of the Civil Rights Act of 1964 (U.S. Office for Civil Rights, 2000) and related statutes were intended to ensure that patients from different racial and ethnic subgroups have equal access to quality care. However, enforcement of these basic rights by the Office for Civil Rights and other entities is made far more difficult without standardized, readily available data on race and ethnicity to monitor the care that different subgroups receive (Smith, 1999).
LANGUAGE PREFERENCE AND SOCIOECONOMIC FACTORS
Categorizing individuals in racial and ethnic categories has helped to identify health disparities. However, many anthropologists and other social scientists view these categories as relatively crude and inaccurate tools for understanding differences between groups (IOM, 2002). Indeed, substantial variation in important characteristics between and within racial and ethnic groups may have as much or more effect on health and the quality of care than race or ethnicity per se. Two such characteristics are language preference and SES.
Many individuals experience language barriers ranging from no English proficiency to limited proficiency in speaking, reading, or comprehending English (IOM, 2003). For example, it is estimated that more than one in four Asian/Pacific Islanders and Hispanics live in households where no adolescent or adult speaks English “very well” (U.S. Bureau of the Census, 1990). A recent Commonwealth Fund report documents the over-whelming barriers to care faced by non-English speaking Hispanics (Collins, Tenney, and Hughes, 2002). Language barriers often vary considerably by country of origin within racial and ethnic groups. For instance, whereas less than 2 percent of Hawaiians and 15 percent of Japanese live in households where English is not spoken well, 26–42 percent of Thais, Chinese, Koreans, and Vietnamese, and more than half of Laotians, Cambodians, and Hmong live in such households (U.S. Bureau of the Census, 1990; IOM, 2003).
A number of federal regulations encourage the use of interpreters in the health care setting (U.S. DHHS, 2000a; Office for Civil Rights, 2000; Perot and Youdelman, 2001). However, language barriers continue to pose significant problems for both patients and providers. In one recent survey, 43 percent of the Hispanics living in households where Spanish was the primary language reported having difficulty communicating with and understanding their doctor (Doty and Ives, 2002). Only half of the patients who said they needed an interpreter when visiting a doctor said they always or usually got one. In many instances (43 percent), when an interpreter was available, he or she was a member of the patient’s family; rarely (1 percent) was the interpreter professionally trained (Collins, Tenney, Hughes, 2002).
Many providers are acutely aware of how language barriers and other cultural differences constrain their ability to provide effective care (IOM, 2002). In a survey of Los Angeles providers who participated in care programs sponsored by the county health authority, more than 70 percent felt that language and culture are important to the care of their patients and more than half believed that their patients did not adhere to medical treatments as a result of linguistic or cultural barriers (Cho and Solis, 2001).
In sum, routine collection of information on patients’ primary language or language preference is also an essential step in identifying patient subgroups where language barriers may be present and in developing culturally competent care (Betancourt, Green, and Carillo, 2002; IOM, 2002).
Collecting information on a patient’s primary language is legal and authorized under Title VI of the Civil Rights Act of 1964, though few federal statutes require it (Perot and Youdelman, 2001). Some health plans already routinely collect such data. For example, Kaiser Permanente in northern California has made cultural competency a priority for several years. As part of this effort, those using Kaiser data systems for patient care
(such as making an appointment) cannot do so unless the language preference field has been filled in: the system thus prompts the user to indicate the patient’s language preference and the need for translator services. Information sheets in the patient’s native language describing his/her condition or treatment can often also be provided.
SES is a multidimensional concept that reflects an individual’s access to material and social resources and assets including income, wealth, and educational credentials, as well as an individual’s prestige or status in society as reflected in access to consumption of goods, services, and knowledge (Krieger, Williams, and Moss, 1997). SES is measured in a variety of ways. Income and/or education measures are the most common in the United States; measures such as occupational class may be used in Europe. Regardless of how SES is measured, numerous studies have demonstrated a consistent socioeconomic gradient in which health status, and often the quality of care received, decreases with declining SES. The gradient generally persists even when individual risk factors, including being in a minority racial or ethnic group, are taken into account (Robert and House, 2000; Kaplan, Everson, and Lynch, 2000; Gornick, 2000).
Although race or ethnicity and SES may exert some independent effects, the two are often interrelated; hence, information on SES can help highlight the underlying sources of disparities in health status and care between and among different racial and ethnic groups (IOM, 2002; Wong et al., 2002). For example, education and income levels vary substantially among minority racial and ethnic groups. Among adults, Asian Americans/ Pacific Islanders were most likely to have at least a high school education (83 percent), followed by African Americans (72 percent), and Hispanics (53 percent) (Bennet and Martin, 1995). Asian Americans/Pacific Islanders also had the highest median income ($55,500); that of African Americans and Hispanics was substantially lower ($33,400 and $30,400, respectively) (U.S. Bureau of the Census, 2001). However, although Asian Americans/ Pacific Islanders tend to have the highest education and income levels overall, SES varies considerably between Asian subgroups.
Although there is wide consensus that SES plays a major role in health disparities and should be taken into account whenever possible, exploring the links between SES and health disparities is a relatively new field of research. Many key issues remain unresolved—in particular the best ways to measure SES (Gornick, 2000). Consider education, one of the most widely used measures. It is popular because it is easy to measure and collect, it applies to persons who are not active in the labor force (e.g., unemployed, retired), and it is relatively stable over an adult’s life span, regardless of
changes in health. However, the economic and health correlates of educational level may vary by age, birth cohort, gender, class position, and perhaps most importantly, by race or ethnicity (Krieger, Williams, and Moss, 1997). For example, economic returns for a given educational level are larger for whites than for African Americans, Hispanics, and American Indians (U.S. Bureau of the Census, 1991). College-educated blacks are four times more likely than their white counterparts to become unemployed and experience consequent drops in income (Wilhelm, 1987). The limitations of individual measures have led experts to encourage simultaneous use of multiple measures to assess SES (Krieger, Williams, and Moss, 1997; Robert and House, 2000).
PUBLIC AND PRIVATE-SECTOR USERS OF RACIAL AND ETHNIC DATA
As suggested above, a variety of stakeholders have considerable interest in racial and ethnic data on health and health care. In this section, we focus on public- and private-sector entities that are likely to use such data.
The following federal and state agencies, which are tasked with the primary responsibility to deliver or purchase health care need racial and ethnic data to ensure that care of similar quality is being provided to all users.
Center for Medicare and Medicaid Services (CMS)
The most obvious federal user of racial and ethnic data is CMS. Monitoring and reporting quality of care has been a long-standing function of the Medicare program. Research reports and other studies of the Medicare program have examined differences in utilization of Medicare services according to race. Until recently, data were available and sufficiently reliable to report about care for blacks and whites only. Improvements to the system through which Medicare receives data about race and ethnicity have made it possible to examine differences in utilization for major racial and ethnic groups, although these data still have many limitations (Lauderdale and Goldberg, 1996; Arday et al., 2000).
Such analyses have revealed important differences in the extent and nature of disparities for different minority subgroups and the need to tailor interventions to specific subgroups. Nevertheless, CMS does not yet report quality of care measures for all of the main racial and ethnic subgroups. As we discuss below, CMS’s ability to accurately gauge such disparities and
effectively target interventions to meet the needs of its beneficiaries is undermined by problems with existing data sources.
CMS shares responsibility for Medicaid with state governments. Because nearly 50 percent of Medicaid beneficiaries come from racial and ethnic minority groups, improving quality of care in the Medicaid program is likely to have significant advantages for beneficiaries. Geppert and colleagues in this volume (see Appendix E) discuss the state government as a user of racial and ethnic data in more detail.
Health Resources and Services Administration (HRSA)
HRSA finances delivery of care through community health centers. Unlike Medicare, HRSA collects data on the race and ethnicity of community health center users and has reported quality of care measures by race and ethnicity.
Substance Abuse and Mental Health Services Administration (SAMHSA)
SAMHSA does not directly provide care for individuals with substance abuse and mental health problems. However, it does have responsibilities for ensuring that substance abuse services are available to those in need. Fulfilling this function requires knowing the racial and ethnic characteristics of the population in need so that appropriate programs can be made available.
Department of Veterans Affairs (VA) and Department of Defense (DOD)
The VA and DOD also directly operate health care programs and have responsibilities for providing high-quality care to enrollees. The DOD has ready access to data on race and ethnicity and has conducted a number of focused studies to ascertain whether there are racial or ethnic disparities in care. The VA has had to rely largely on Social Security Administration data to obtain racial and ethnic information and until recently was able to examine quality for whites and blacks only. A recent effort to collect data about satisfaction and health status from all VA users has generated self-reported racial and ethnic data that can be used to monitor and report on quality of care. Health care providers in VA and DOD facilities are increasingly held accountable for performance on standard quality indicators but are not routinely assessed for how their performance varies between patients in different racial and ethnic groups.
Other DHHS Agencies
In addition to those federal agencies involved in the direct provision of care, other agencies of DHHS have a responsibility to ensure the equitable distribution of health care assets across the population. This includes the Agency for Healthcare Research and Quality (AHRQ), which has responsibility for monitoring health insurance and medical expenditures, and the Centers for Disease Control and Prevention (CDC), which often needs such information to fulfill public health functions. Key public health functions that require the collection of racial and ethnic data include surveillance, prevention, including immunization and health education, and communication.
Federal Employee Health Benefit Program (FEHB)
The FEHB purchases health insurance on behalf of federal employees and their dependents. As a prudent purchaser, the program has a responsibility to create conditions that ensure high-quality care for its enrollees. It also has an interest in ensuring that, as a steward of federal resources, it is getting value for the health care dollar, and that groups for which it pays the same premium actually receive equivalent services. It is highly likely, however, that FEHB cannot completely meet either of those conditions because the ability to measure quality of care for different population groups does not currently exist.
Demands for accountability may be even more pronounced in the private sector than in the public sector. The groups most affected by this trend and most in need of data on race and ethnicity are insurers and providers. As discussed in an earlier section, employers and individual consumers also are increasingly interested in this information.
Private Health Insurers
Insurers in the managed-care and fee-for-service sectors are increasingly held accountable for providing high-quality care. Many large employers have begun requiring measures of quality, such as through HEDIS. However, lack of racial and ethnic data in the private insurance market has largely prevented assessment of quality of care for different racial and ethnic populations, let alone holding insurers accountable for providing care of similar quality. Private purchasers have interests similar to those of public purchasers in this regard. Yet they are unable to ensure that they are
getting good value for all of their employees. Their inability to do so has broad implications for firms. Because good health is essential for a productive workforce, employers with large minority workforces may inadvertently undermine their organization’s productivity by contracting with insurers or health plans that have large disparities in the quality of the care they provide.
Health Care Providers
Increasingly, health care providers such as hospitals, clinics, and physician practices are also being held accountable for providing high-quality care. But as previously discussed, achieving accountability is impossible without adequate data. One additional point is worth noting in this area. The IOM report Unequal Treatment (IOM, 2003) suggests that subconscious biases and other institutional processes may contribute to disparities in care. Without data on the existence and extent of these disparities at an institutional or practice level, it is impossible for providers—be they individuals or organizations—to know whether they have such problems and whether they are making progress in addressing them.
APPROACHES TO OBTAINING DATA ON RACE AND ETHNICITY
A variety of existing data sources contain information about individuals’ race and ethnicity that is linked, or can be linked, to information about their health and health care. Such data sources have been crucial to documenting disparities in health and care for minority and low-income individuals in a variety of settings and circumstances.
Unfortunately, existing data sources, which are often targeted to specific populations, tell us little about large segments of the population and are undermined by inaccurate coding of race and ethnicity or lack of comparable measures of race and ethnicity. Below we describe some existing data sources and note limitations to their use. We focus primarily on Medicare data; descriptions of other data sources are available elsewhere (IOM, 2003; Bierman et al., 2002; U.S. Office for Civil Rights, 2001).
With approximately 40 million beneficiaries, 15 percent of whom are minority, Medicare files should, in theory, be a source of racial and ethnic information on a significant portion of the population. In practice, however, the usefulness of these data has been limited by problems with their accuracy and completeness. Information on the race and ethnicity of Medicare beneficiaries is contained in the CMS enrollment database (EDB).
CMS does not actually collect racial and ethnic information directly; instead it uses entitlement information collected by the Social Security Administration on a voluntary basis from individuals applying for a new or replacement social security card. This information is then transferred to a file known as the master beneficiary record file (MBR), which, in turn, is used to populate the racial and ethnic categories in the EDB (Arday et al., 2000).
The MBR includes only four racial and ethnic categories: white, black, other, and unknown. Most experts consider Medicare data on race and ethnicity to be useful for comparing blacks vs. whites (or nonblacks); however, its usefulness for other racial and ethnic categories is more limited. The accuracy of racial and ethnic codes in the EDB may be further undermined by the fact that the EDB assigns the race of the primary wage earner to the approximately 20 percent of beneficiaries entitled to Medicare because of their association with the wage earner.
Beginning in 1994, CMS undertook a number of steps to improve the accuracy and completeness of racial and ethnic coding in the EDB. The main change was that CMS began to periodically update the EDB using another SSA file, known as NUMIDENT, that added three categories of new race and ethnicity: Hispanic, Asian American or Pacific Islander, and North American Indian or Alaska Native. Unfortunately, the SSA opted to ask applicants whether they were Hispanic in a single question about race and ethnicity rather than separately from the question about racial categories. Consequently, Hispanics were effectively coded as race rather than ethnicity (Lauderdale and Goldberg, 1996). In addition to NUMIDENT updates, CMS was able to further correct the coding in the EDB by conducting a direct-mail survey to over 2 million beneficiaries whose race was listed as “unknown or other” or who had a Hispanic surname or country of birth. CMS has also begun to work with the Indian Health Service to correct misclassification of Native American beneficiaries.
Although these improvements have substantially increased the accuracy and completeness of racial and ethnic codes in the CMS EDB, several recent analyses indicate that the accuracy of the codes is still less than 60 percent for Hispanics, Asians, and Native Americans (Perot and Youdelman, 2001). Moreover, starting in 1988 the SSA began assigning social security numbers at birth without collecting racial and ethnic data, consequently, NUMIDENT files may not include racial or ethnic data on most individuals born in the United States after 1988.
CMS also monitors the performance of the Medicare program through both Consumer Assessment of Health Plans Survey (CAHPS) and the Medicare Current Beneficiary Survey (MCBS). Fortunately, both of these surveys collect self-reported data on race and ethnicity, so accuracy and completeness are not major problems. However, the ability to analyze important
subgroup responses in both of these surveys is limited by sample size considerations. If data on race and ethnicity were universal in Medicare data systems, it would be possible to oversample populations of interest when CAHPS and MCBS are administered.
Imputing Data on Race, Ethnicity, and SES
Even if the standardized collection data of on race, ethnicity, and SES were required, widespread implementation and the collection of complete data on all patients would likely take several years. In the meantime, it may be possible to impute measures of race, ethnicity, and SES from existing data sources when other sources are unavailable. The two most widely used methods of imputation for this are geocoding and surname analysis. Both of these approaches are relatively easy to perform; however, both may have limited usefulness for identifying members of some racial or ethnic subgroups.
This is a well-validated technique in which an individual’s address is linked to census data about the area in which he or she lives in order to estimate the individual’s racial, ethnic, or socioeconomic characteristics (Krieger, Williams, and Moss, 1997). Geocoding can be performed easily using commercially available software or by contracting with a vendor. In general, the smaller the census area considered, the better the estimate of an individual’s characteristics. For example, Zip Codes span relatively large geographic areas containing upwards of 30,000 people, and they are not typically homogeneous in sociodemographic make-up. In contrast, census block groups averaging 1,000 residents or fewer correspond to the size of a small neighborhood and are generally quite homogeneous.
Geocoding is considered to be more useful for imputing socioeconomic characteristics of individuals based on the SES characteristics of their neighborhoods than for identifying an individual’s race (Krieger, Williams, Moss, 1997). Nevertheless, geocoding can provide reliable estimates of racial characteristics among certain minority groups that tend to live in segregated neighborhoods. For example, in one recent study, geocoded measures of black race showed high concordance with individual-level measures, and analyses of disparities in quality of care in managed-care plans had essentially identical outcomes whether the geocoded or individual measure of race was used (Fremont et al., 2002). Nevertheless, geocoding may produce relatively high rates of misclassification of race or ethnicity when applied to geographic areas where racial or ethnic segregation is low for the group(s) of interest.
This technique uses an individual’s last name to estimate the likelihood that they belong to a particular racial or ethnic group. The technique has been applied in a variety of ways such as to identify local concentrations of ethnic groups or to estimate the completeness of racial or ethnic identification when information is incompletely recorded (Abrahamse, Morrison, and Bolton, 1994; Lauderdale and Kestenbaum, 2000). The use of surname analysis has been limited to identifying Hispanics and Asian Americans, for which there are well-developed surname dictionaries. For example, in a recent sponsored project by the Commonwealth Fund, selected health plans used only surname analysis to identify Hispanic plan members; and found significant disparities in care between Hispanics and non-Hispanics (Nerenz et al., 2002).
Use of surname analysis for other racial or ethnic groups is much more problematic than for Hispanics and Asians and is generally not done. For instance, many Native Americans have names given to them by settlers and would be frequently misclassified as non-Hispanic whites using this technique. Depending on the name and context, misclassification can also be a significant problem even among Hispanics and Asians (Abrahamse, Morrison, and Bolton, 1994). The utility of surname analysis may increase over the next several years as new dictionaries and techniques (e.g., incorporating information from geocoding) are made available that make it possible to more accurately identify racial and ethnic subgroups among Asians (Lauderdale and Kestenbaum, 2000) and Hispanics (Peter Morrison, oral communication, August 2002).
CHALLENGES TO COLLECTING RACIAL, ETHNIC, AND SES DATA
The potential benefits of collecting racial, ethnic, and SES data are clear to many. However, data collection raises a number of concerns that must be addressed. These include methodological considerations, patient privacy and confidentiality, civil rights laws, and burdens that collecting additional data will create for various entities.
The most appropriate approach to obtaining data depends on the specific circumstances in which the data are collected; multiple methods will probably be needed until self-reported racial and ethnic data are collected in a standardized fashion (e.g., using OMB racial and ethnic categories) and are more widely available. For example, because most private insurers do
not now collect data on race or ethnicity, a mix of methods may be needed to bridge to the goal of achieving complete and accurate information over the next several years. Collecting data by self-report at or shortly after enrollment as part of an intake process is likely to provide the most accurate data for those who respond. Such an approach could be merged with imputed data, such as estimate based on geocoding, until such time as primarily collected data are available for all enrollees.
How long it might take to achieve totally self-reported data probably depends on the nature of the enrolled population and on market dynamics, including disenrollment patterns and the number of plans in an area. Collecting additional data during patient encounters could conceivably fill some important gaps, but this approach is likely to yield data of inconsistent quality and to be time consuming and impractical, especially considering the challenge of collecting data on enrollees who do not obtain care regularly. Surveys sent to all enrollees for whom racial or ethnic data are missing can help to fill gaps, as has been the case for Medicare. However, plans will need a clear reason to undertake such an effort. Their willingness to do so may depend on external pressures for reporting about quality and on consumer/purchaser demand for information.
Concerns about Privacy and Confidentiality
Many patients and consumer advocates view collecting information about an individual’s race, ethnicity, and socioeconomic characteristics as intrusive and a potential invasion of privacy. Thus, ensuring that collected information remains confidential is crucial if routine collection is to occur. The passage of the Health Insurance Portability and Accountability Act (HIPPA), and the related privacy rule should go a long way to ensure that individually identifiable information about a patient’s race, ethnicity, and SES is not shared inappropriately.
Nevertheless, the HIPAA Privacy Rule’s primary focus is to protect personal health information rather than information about race, ethnicity, and SES. In addition, covered entities have considerable latitude in how information in medical records is used within their organization (i.e., for treatment, payment, and routine operations) as well as to which business associates to disclose information. Thus, concerns among patients, consumer advocates, and others about the privacy and confidentiality of racial and ethnic data are likely to persist.
Exposure to Civil Rights Litigation
Increasing attention to patient privacy and the financial and criminal penalties associated with violating the Privacy Rule have also likely height-
ened providers’ concerns that routine collection of racial and ethnic data may violate civil rights laws. Indeed, a widespread misconception held by many providers and health plans is that collecting racial and ethnic data violates federal and state civil rights law (Nerenz et al., 2002; IOM, 2003; Bierman et al., 2002). A recent review of federal laws revealed no laws that prohibit the collection of racial and ethnic data (including language preference). Rather, there are numerous laws and program-specific statutes already in effect or scheduled to go into effect within the next several years that encourage or require the collection of racial and ethnic data (Perot and Youdelman, 2001). Similarly, reviews of state laws have shown that only four states have any sort of restrictions on collecting such racial or ethnic data, and these varied as to whether the restriction applied before, during, or after enrollment. At least one state required health plans to maintain information on enrollee’s race and ethnicity (Perez and Satcher, 2001; Bierman et al., 2002).
Previous reviews have not explicitly addressed the legality of collecting data other than race and ethnicity that relate to an individual’s socioeconomic status. However, many believe that the collection of such data for the purposes of monitoring and improving quality is consistent with civil rights laws designed to protect patients from differential treatment or discrimination based on personal characteristics.
Some plans and providers report concerns that the routine collection of racial and ethnic data and the reporting of various performance measures by race and ethnicity place them at substantially increased risk of class action lawsuits were disparities in their plans to become apparent. While this is a significant concern, it is not clear that the threat of litigation is greater when such data are collected than when they are not.
Two examples illustrate the ambivalence of health plans about the risks of collecting data on race and ethnicity. One health plan executive, in explaining why his plan was “a little gun shy,” described a case in which an individual had voluntarily provided racial data at the time of application for an individually underwritten policy. Although the insurer is adamant that the individual policy was denied on the basis of multiple preexisting conditions, the individual sued the plan on the basis of racial discrimination. By contrast, an executive of a different health plan expressed his firm conviction that the act of examining quality for different racial and ethnic groups was evidence of his company’s taking action to prevent discrimination, and would “probably immunize” the company against a successful civil rights lawsuit. Although some potential legal exposure could be avoided by limiting data collection until after a coverage decision has been made, there is probably no way to fully guard against legal action.
Burden of Collecting Data
In addition to concerns about patient privacy and civil rights, the entities that collect and analyze information about racial, ethnic, and SES characteristics of persons they serve may face significant costs (Nerenz et al., 2002). Adding data elements to large administrative data sets maintained by health plans can be expensive in terms of time and resources spent reconfiguring files and forms. In addition, depending on how data are obtained (e.g., at time of enrollment or visit, surveys of existing members), plans may need to devote substantial resources to actually filling in the data fields.
Once data are collected, there can be additional costs associated with analyzing and reporting measures of care and performance stratified by race, ethnicity, and SES. For example, NCQA HEDIS measures that require chart abstraction can be expensive for plans. To keep costs to a minimum the NCQA requires plans to obtain data only for a sample of approximately 400 enrollees. However, these sample sizes are not sufficient to conduct meaningful analyses for minority subgroups. Consequently, plans would need to oversample these groups (or sample larger overall populations) in order to have sufficient power to detect racial, ethnic, or SES differences. Small sample size is less of a problem for HEDIS measures that can be calculated solely from administrative data and that focus on care for common conditions such as hemoglobin A1c checks in diabetics (Fremont et al., 2002).
Health plans have already invested substantial time and resources to reconfigure their data systems to meet HIPAA requirements. Unfortunately, although race and ethnicity are included as optional “situational” data fields for some types of patients and standardized forms, they are listed as “not used” or are not allowed on others. Consequently, routine collection of standardized racial and ethnic data for many types of enrollees is hindered rather than facilitated under current HIPAA rules (see the Bocchino paper in Appendix G for a detailed discussion).
Finally, in addition to the risk of litigation because of perceived misuse of racial or ethnic data (discussed above), some plans fear suffering business losses if minority populations they serve view the collection of such data as an effort to ration care. Such populations may also be reluctant to enroll in plans if publicly released “report cards” suggest poorer health and outcomes among minority or low-income groups, although the differences may reflect case mix and the effects of poverty rather than lower-quality care. Without either statistical adjustments for case mix differences, which are extremely difficult to present in report card format, or education of consumers, these report cards could unfairly hurt plans’ efforts to increase market share in minority and low-income populations (IOM, 2003; Bierman et al., 2002).
Although considerable challenges to collecting racial and ethnic data remain, we believe there are reasons for optimism. One such reason is the emergence of innovative initiatives in the private sector. For example, Aetna, a large national insurer, has undertaken an extensive minority health initiative. There appear to be two related reasons for the company’s decision. First, after examining U.S. demographic trends, Aetna has concluded that an increasing proportion of its members will be minorities and, therefore, both wants to and has to measure and ensure high-quality care for these members. Second, Aetna’s leadership has articulated a moral imperative to act on the IOM report (2003), and to move toward ensuring that care for its members is not different on the basis of race or ethnicity.
Aetna has adopted a multipronged approach. First, it has begun collecting data on the race and ethnicity of its members (at or after the time of enrollment, in all states in which it is legal to do so) with the goal of using quality of care algorithms to measure quality for different racial and ethnic groups. Second, it is strengthening efforts to ensure that minority providers are part of its provider networks. Finally, the company is altering its marketing strategies, using a more diverse workforce, and stressing its commitment to cultural competence.
Although much of this effort is relatively recent, the insurer’s staff report little, if any, consumer resistance to voluntarily providing racial and ethnic information. In fact, they have received significant positive feedback and expressions of appreciation from consumers. When asked about potential concerns about respondent burden in providing racial and ethnic data, staff stressed that such information was obtained during routine assessments of health status and needs, and that the major burden was “having to change all of our forms” and the software programs to read them.
Several other plans have reported collaboration with their state Medicaid agencies to examine racial and ethnic data for their Medicaid enrollees. The federal Medicaid Managed Care regulations encourage data sharing, and these plans report that use of these data has already led to quality improvements, particularly in the areas of diabetes and depression care.
Improving and broadening the collection of racial and ethnic data are critical to fulfilling core goals of the U.S. health system. While such efforts will pose significant challenges and entail real costs, ignoring this important need may ultimately be higher than definitively solving the problem.
There is growing consensus that racial and ethnic disparities in health care reflect serious problems in the overall quality of care that may affect
any patient. In this respect, failure to collect racial and ethnic data will hamper much-needed efforts to improve the quality of care, and exacerbate the erosion of trust in the consistency and quality of U.S. health care (IOM, 2003).
Without accurate information on the racial and ethnic characteristics of populations served, public health efforts and associated resources are likely to be poorly targeted and may miss large segments of the populations most in need. Poorly managed chronic conditions or undiagnosed disease can result in more severe disease, worse outcomes, and higher health costs.
Since minority populations make up an increasing proportion of the workforce, their unmet health needs can substantially reduce workers’ quality of life and productivity, which in turn can affect the economy at all levels.
Minority populations constitute an increasingly large segment of the health care market; thus lack of data about their care will undermine efforts to support consumer choice and stimulate market forces.
The inability to routinely monitor the health needs and quality of care received by minority populations violates existing federal statutes and contradicts fundamental values of equity and fairness in this country. DHHS has already been sued for not requiring providers to collect and report uniform data on race and ethnicity. Without further initiatives to collect these data, additional lawsuits directed at government agencies and private health plans are likely.
Collection of reliable data on race and ethnicity is feasible, and there are good examples of its current collection and use in both the public and private sectors. The federal government can take a leadership role by ensuring the universal, ongoing collection of data in the Medicare program, and can stimulate its use in Medicaid programs and the private sector. Although private-sector data collection will likely remain voluntary, both insurers and employers have already demonstrated leadership in making the case for such data, and will hopefully set new industry standards. We anticipate that their efforts will be further encouraged by federal action in the Medicare program and increased consumer demand for such information.
The ongoing debate about the uses of and appropriate measures and methods for collecting SES and language preference data is encouraging. As has been the case in the area of racial and ethnic data collection, voluntary pilot efforts will likely inform the feasibility and usefulness of such data collection.
In the long run, the completeness and accuracy of data on race, eth-
nicity, language preference, and SES will improve only if this information is collected for meaningful and actionable purposes. To that end, we offer recommendations in the following areas.
Uniform Standards and Training
Standards must be developed to ensure uniform collection of data at federal, state, and local levels. In addition, there is a critical need for guidance about how these data should be collected in different settings (e.g., when, how, by whom) and training for frontline personnel in how best to do this.
Recommendation A1: Create a centralized body that can provide guidance and oversight regarding standards for and collection of data. This body should propose a set of incentives for the collection of such data, as well as penalties for failing to collect data in ways that meet a minimal standard.
Recommendation A2: Provide an adequate budget to ensure the availability of such training and data collection at state and local levels.
The collection and use of data are critical to the functioning of the public health infrastructure, and provide a basis for ensuring that public health systems are accountable to different populations and communities.
Recommendation B3: Ensure sampling of all major racial and ethnic groups for all major epidemiologic and health status data collection efforts funded by the federal government, including those that provide important subnational data. These include, but are not limited to the Behavioral Risk Factor Surveillance System (BRFSS), National Health Interview Survey (NHIS), Medicare Expenditure Panel Survey (MEPS), Healthcare Cost and Utilization Project (HCUP), National Health and Nutrition Examination Survey (NHANES), Consumer Assessment of Health Plans Survey (CAHPS), Medicare, Medicaid, and federal employees. This will require a commitment to oversampling smaller populations in each of these surveys, and an appropriate budget for doing so.
Recommendation B4: Ensure the collection of data on race and ethnicity for all vital records.
Health Care Delivery System
Data collection in the health care system itself provides the basis for assessing disparities in care and for benchmarking progress. While these recommendations apply to Medicare, Medicaid, and the private sector, they may also be relevant to the VA and Department of Defense health care systems.
Update and ensure the accuracy of data on race and ethnicity for current and future Medicare beneficiaries. This can be accomplished in a variety of ways, including geocoding and merging other self-reported racial and ethnic data, such as from CAHPS, into the EBD.
Continue progress on identification of Native Americans. Extend assignment of Native American status beyond those living on reservations to urban dwellers.
Eliminate dependency on the Social Security Administration for racial and ethnic data, or create a permanent fix to the problem of incomplete and inaccurate information. Current estimates of how much CMS spends on this range from a few million to several hundred million dollars annually. It is thus likely that redeployment of these funds would be sufficient to implement a permanent solution. In addition, steps need to be taken now to address the fact that data on race and ethnicity have not been collected in the social security card application process since 1988.
Tie elimination of disparities to Medicare performance. The Medicare program should monitor quality of care for different racial and ethnic groups by plan, by hospital, and by state. Quality improvement organizations may be one vehicle to do this. In addition, CMS should use its purchasing power in working with Medicare+Choice plans to develop and implement data collection plans and initiatives to eliminate disparities. CMS should also consider whether the collection and reporting of accurate information about race and ethnicity should be a condition of hospital and nursing home participation in the Medicare program.
Extend the requirements for reporting performance for racial and ethnic groups in SCHIP to the Medicaid program.
Make racial and ethnic data collection based on self-report (when feasible) mandatory for all new Medicaid beneficiaries, and update data on
current beneficiaries as eligibility redeterminations are made. Require states to report these data to CMS and enforce this requirement.
Provide Medicaid programs with sufficient financial support to accomplish this, and consider using financial incentives to meet this goal (such as tying a portion of the federal contribution to data completeness and accuracy).
The capacity of health plans and insurers, health systems, employers, providers, and consumers to use data on race and ethnicity to improve health and health care depends on availability of such data. The IOM report Unequal Treatment (2003) has provided further evidence for a business case for such data.
Recommendation D7: The Department of Health and Human Services, standard-setting organizations, providers and employers should take steps to modify HIPAA standards to facilitate the reporting of racial and ethnic data. This could be accomplished if DHHS and others clearly defined the business case for reporting these data. Change the accompanying guide that describes the racial and ethnic data elements for ANSI X12N 837 from a listing of “not used” to “situational” in order to facilitate reporting of racial and ethnic data.
Recommendation D8: Develop educational programs to stimulate consumerism through information provided to employers and individuals about how they can use such data.
Recommendation D9: The Department of Health and Human Services, the American Association of Health Plans, employer groups, and others should conduct education and outreach to health plans, purchasers, and employers regarding the legality of collecting data on race and ethnicity.
In addition to research on disparities, expand research on data, such as types of data, relationships between variables such as race or ethnicity, SES, and education, and methods for oversampling smaller populations.
Recommendation E10: The Department of Health and Human Services should provide funding to expand research on ways to collect accurate data, including geocoding.
Recommendation E11: The Department of Health and Human Services should provide funding to expand on best ways to make self-report acceptable, training to collect data, and techniques for oversampling.
Recommendation E12: The Department of Health and Human Services and the Department of Justice should fund research to identify mechanisms that balance personal protection with protections for plans and others that use data for appropriate purposes as well as mechanisms to protect against misuse of data or detect possible redlining.
Abrahamse, A.F., P.A. Morrison, and N.M. Bolton 1994 Surname analysis for estimating local concentration of Hispanic and Asians. Population Research and Policy Review 13:383-398.
Arday, S.L., D.R. Arday, S. Monroe, and J. Zhang 2000 HCFA’s racial and ethnic data: Current accuracy and recent improvements. Health Care Financial Review 21:107-116.
Ayanian, J.Z., P.D. Cleary, J.S. Weissman, and A.M. Epstein 1999 The effect of racial differences on patients’ preferences in access to renal transplantation. New England Journal of Medicine 341:1661-1669.
Bennet, C.E., and B. Martin 1995 The Asian and Pacific Islander population. In Population Profile of the United States: 1995. Washington, DC: U.S. Government Printing Office.
Berwick, D.M. 1998 Crossing the boundary: Changing mental models in the service of improvement. International Journal of Quality Health Care 10:435-441.
Betancourt, J.R., A.R. Green, and J.E. Carrillo 2002 Cultural Competence in Health Care: Emerging Frameworks and Practical Approaches. New York: The Commonwealth Fund.
Bierman, A.S., N. Lurie, K.S. Collins, and J.M. Eisenberg 2002 Addressing racial and ethnic barriers to effective health care: The need for better data. Health Affairs (Millwood) 21:91-102.
Burstin, H.R. 1993 Do the poor sue more? A case-control study of malpractice claims and socioeconomic status. Journal of the American Medical Association. 270:1697-1701.
Cho, J., and M. Solis 2001 Healthy Families Culture and Linguistics Resources Survey: A Physician Perspective on Their Diverse Member Population. Los Angeles: LA Care Health Plan.
Collins, K.S., A. Hall, and C. Neuhaus 1999 U.S. Minority Health: A Chartbook. New York: The Commonwealth Fund.
Collins, K.S., K. Tenney, and D.L. Hughes 2002 Quality of Health Care for African Americans: Findings from The Commonwealth Fund 2001 Health Care Quality Survey. New York: The Commonwealth Fund.
Doty, M.M., and B.L. Ives 2002 Quality of Health Care for Hispanic Populations: Findings from The Commonwealth Fund 2001 Health Care Quality Survey. New York: The Commonwealth Fund.
Fiscella, K., P. Franks, M.R. Gold, and C.M. Clancy 2000 Inequality in quality: Addressing socioeconomic, racial, and ethnic disparities in health care. Journal of the American Medical Association 283:2579-2584.
Fremont, A.M., and C.E. Bird 2000 Social and psychological factors, physiological processes, and physical health. In Handbook of Medical Sociology, C.E. Bird, P.C. Conrad, and A.M. Fremont, eds. Englewood Cliffs, NJ: Prentice Hall.
Fremont, A.M., S. Wickstrom, C.E. Bird et al. 2002 Socioeconomic, Racial/Ethnic, and Gender Differences in Quality and Outcomes of Care as it Relates to Cardiovascular Disease. Rockville, MD: Agency for Healthcare Research and Quality.
Gornick, M.E. 2002 Measuring the effects of socioeconomic status on health care. Pp. 45-74 in Guidance for the National Healthcare Disparities Report, E.K. Swift, ed. Washington, DC: National Academy Press.
2000 Disparities in Medicare services: Potential causes, plausible explanations, and recommendations. Health Care Financial Review 21:23-43.
Hoffman, C., and M. Pohl 2000 Health Insurance Coverage in America: 1999 Data Update. Washington, DC: The Kaiser Commission on Medicaid and the Uninsured.
Institute of Medicine 2003 Unequal Treatment: Confronting Racial and Ethnic Disparities in Health Care. B.D. Smedley, A.Y. Stith, and A.R. Nelson, editors. Committee on Understanding and Eliminating Racial and Ethnic Disparities in Health Care. Board on Health Sciences Policy. Washington, DC: The National Academies Press.
2002 Speaking of Health: Assessing Health Communication Strategies for Diverse Populations. Committee on Communication for Behavior Change in the 21st Century: Improving the Health of Diverse Populations, Board on Neuroscience and Behavioral Health. Washington, DC: The National Academies Press.
2001 Crossing the Quality Chasm: A New Health System for the 21st Century. Committee on Quality of Health Care in America. Washington, DC: National Academy Press.
2000 To Err Is Human: Building a Safer Health System. Linda T. Kohn, Janet M. Corrigan, and Molla S. Donaldson, Editors. Committee on Quality of Health Care in America. Washington, DC: National Academy Press.
1999 The Unequal Burden of Cancer: An Assessment of NIH Programs and Research forMinorities and the Medically Underserved. M.A. Hayes and B.D. Smedley, eds. Washington, DC: National Academy Press.
Kaplan, G.A., S.A. Everson, and J.W Lynch 2000 The contribution of social and behavioral research to an understanding of the distribution of disease: A multilevel approach . In Promoting Health: Intervention Strategies from Social and Behavioral Research, B.D. Smedley and S.L. Syme, eds. Washington, DC: National Academy Press.
Krieger, N., D.R. Williams, and N.E. Moss 1997 Measuring social class in U.S. public health research: Concepts, methodologies, and guidelines. Annual Review of Public Health 18:341-378.
Laouri M., R.L. Kravitz, W.J. French et al. 1997 Underuse of coronary revascularization procedures: Application of a clinical method. Journal of American College of Cardiologists 29:891-897.
Lauderdale, D.S., and J. Goldberg 1996 The expanded racial and ethnic codes in the Medicare data files: Their completeness of coverage and accuracy. American Journal of Public Health 86:712-716.
Lauderdale, D.S., and B. Kestenbaum 2000 Asian American ethnic identification by surname. Population Research and Policy Review 19:283-300.
Lurie, N. 2002 Measuring disparities in access to care. Pp. 99-147 in Institute of Medicine, Guidance for the National Healthcare Disparities Report, E.K. Swift, ed. Washington, DC: The National Academies Press.
Malloy, M.H. 1998 Effectively delivering the message on infant sleep position. Journal of the American Medical Association 280:373-374.
National Cancer Institute 2003 Plans and Priorities for Cancer Research. Bethesda, MD: National Cancer Institute.
National Center for Health Statistics 2001 Health, United States, 2000, with Socioeconomic Status and Health Chartbook. Hyattsville, MD: National Center for Health Statistics.
2000 Health, United States, 1999, with Socioeconomic Status and Health Chartbook. Hyattsville, MD: National Center for Health Statistics.
National Committee for Quality Assurance 1999 HEDIS 2000: Narrative—What’s in it and Why it Matters. Washington, DC: National Committee for Quality Assurance.
Nerenz, D.R., V.L. Bonham, R. Green-Weir, C. Joseph, and M. Gunter 2002 Eliminating racial/ethnic disparities in health care: Can health plans generate reports? Health Affairs (Millwood) 21:259-263.
Office of Management and Budget 1977 Racial and Ethnic Standards for Federal Statistics and Administrative Reporting. (Statistical Directive 15). Washington, DC: Office of Management and Budget.
Perez, T.E., and D. Satcher 2001 Letter to Dr. Bruce Zimmerman, President, American Diabetes Association, from T.E. Perez, Director, Office for Civil Rights, U.S. DHHS, and D. Satcher, Assistant Secretary for Health and Surgeon General, U.S. DHHS, January 19, 2001.
Perot, R.T., and M. Youdelman 2001 Racial, Ethnic, and Primary Language Data Collection in the Health Care System: An Assessment of Federal Policies and Practices. New York: The Commonwealth Fund.
Robert, S.A., and R. House 2000 Socioeconomic inequalities in health: An enduring social problem. In Handbook of Medical Sociology, C.E. Bird, P.C. Conrad, and A.M. Fremont, eds. Englewood Cliffs, NJ: Prentice Hall.
Schneider, E.C., A.M. Zaslavsky, and A.M. Epstein 2002 Racial disparities in the quality of care for enrollees in medicare managed care. Journal of the American Medical Association 287:1288-1294.
Shapiro, M.F., S.C. Morton, D.F. McCaffrey et al. 1999 Variations in the care of HIV-infected adults in the United States: Results from the HIV Cost and Services Utilization Study. Journal of the American Medical Association 281:2305-2315.
Smith, D.B. 1999 Health Care Divided: Race and Healing a Nation. Ann Arbor: University of Michigan Press.
U.S. Bureau of the Census 1991 Current Population. Washington DC: U.S. Bureau of the Census.
1990 1990 Census of Population: Asians and Pacific Islanders in the United States. Washington, DC: U.S. Bureau of the Census.
U.S. Department of Health and Human Services 2000a Healthy People 2010: Understanding and Improving Health. Washington, DC: U.S. Government Printing Office.
2000b Healthy People 2010. Washington, DC: U.S. Department of Health and Human Services.
U.S. Office for Civil Rights 2001 HHS Directory of Health and Human Services Data Resources. Washington, DC: U.S. Office for Civil Rights.
2000 Title VI of the Civil Rights Act of 1964: Policy Guidance on the Prohibition Against National Origin Discrimination as It Affects Persons with Limited English Proficiency. Washington, DC: U.S. Office for Civil Rights.
U.S. House of Representatives 2002 H.R. 5187: Patient Navigator, Outreach, and Chronic Disease Prevention Act of 2002. 107th Congress, 2d session.
Virnig, B.A., N. Lurie, Z. Huang, D. Musgrave, A.M. McBean, and B. Dowd 2002 Racial variation in quality of care among Medicare+Choice enrollees. Health Affairs (Millwood) 21(6):224-230.
Weinick, R.M., S.H. Zuvekas, and J.W. Cohen 2000 Racial and ethnic differences in access to and use of health care services, 1977 to 1996. Medical Care Research Review 57:36-54.
Wilhelm, S.M. 1987 Economic demise of blacks in America: A prelude to genocide? Journal of Black Studies 17:201-254.
Williams, D.R. 1999 Race, socioeconomic status, and health. The added effects of racism and discrimination. Annals of the New York Academy of Sciences 896:173-188.
Wong, M.D., M.F. Shapiro, W.J. Boscardin, and S.L. Ettner 2002 Contribution of major diseases to disparities in mortality. New England Journal of Medicine 347:1585-1592.