The validity of race as an indicator of distinct, genetically different population groups has been widely questioned (Chaturedi and McKeigue 1994; McKenney and Bennett 1994; Senior and Bhopal 1994; Williams, Lavizzo-Mourey, and Warren 1994; Beutler et al. 1996). Greater genetic variation exists within racial groups than between them (Senior and Bhopal 1994; Williams, Lavizzo-Mourey, and Warren 1994). In fact, the variation in gene frequency is approximately 85 percent within racial groups, and only 15 percent between racial groups (Freeman 1998).

For these reasons, many researchers have shifted paradigms, defining race as a social construct based on phenotypic genetic expression rather than as a biological construct (Sheldon and Parker 1992; LaVeist 1994; Senior and Bhopal 1994; Williams, Lavizzo-Mourey, and Warren 1994; Freeman 1998; Jones 2001). As Beutler et al. (1996) argue, while race as a biological construct is illusory, its function as a social–psychological and social–political construct is very real. In fact, the U.S. Bureau of the Census uses race as a social–political construct rather than a biological one.

Within the framework of race as a social–political construct, race is used to understand the health consequences of variations in factors such as health care quality and utilization, adequate housing, education, and nutrition. Race, in this sense, is a multidimensional construct and a predictor of exposure to external health risks posed by environmental, social, and behavioral factors (Fullilove 1994; Hahn and Stroup 1994; LaVeist 1994; Freeman 1998). Freeman (1998) argues that biologic expressions of race result in social interactions, which in turn produce racial and ethnic disparities in morbidity and mortality (i.e., discrimination). Indeed, Kent et al. (2001) contend that the best way to understand race is to view it as a social construct that is influenced by social and political factors. From this perspective, much of the racial differentiation in cardiovascular disease can be explained by social–environmental factors influencing susceptibility and access to medical care (Manson and Ridker 1990). For example, members of racial minority groups may have dietary and lifestyle habits and socioeconomic status that might lead to predispositions for cardiovascular disease and less access to preventive and emergency medical care and advice.

This evidence suggests a new taxonomy for racial identification is needed, focusing on socio-environmental effects of phenotypic characteristics of patients, rather than on the characteristics themselves. Proponents of this view might suggest that, instead of classifying a patient's “race,” we should classify “racism” inherent in the patient's immediate and broader environments. As Jones (2000) argues, while race as a variable only roughly approximates socioeconomic status, culture, and genetic history, it is precisely related to social classification, which significantly impacts daily life experiences. This argument suggests that race is a social construct that includes the effects of racism on an individual. Jones (2000) further asserts that differences in health outcomes associated with race are a direct result of the effects of racism. According to Jones (2001), racism encompasses elements of institutionalized racism and internalized racism. Each of these elements is discussed in the following sections.

Institutionalized racism is defined as having differential levels of access, based on race, to societal goods, services, and opportunities (Jones 2000). Williams (1999) argues that institutionalized racism negatively affects access to educational attainment, employment opportunities, and attainment of higher levels of socioeconomic status. We propose two types of institutionalized racism as most important in the context of health services research: educational racism and access racism. A person living in an environment characterized by institutionalized educational racism would have little exposure to health education materials, instruction, or preventive medical advice. This shortcoming is attributable to institutional factors: racism resulting in illiteracy, and racism causing a dearth of racial minority members being able to acquire medical education, get licensed to practice medicine, and return to their communities to practice. In the Veterans Affairs (VA) health services setting, educational racism manifests itself as, for example, veterans who are fully qualified to receive services but primarily because of lack of health education and information, fail to utilize the services available to them.

Someone living in an environment characterized by institutionalized access racism would have difficulty obtaining urgent medical assistance (e.g., medical visits without an appointment, emergency care) in a timely manner. In this case, racism causes delays by way of overload: too many patients in a low-socioeconomic status (SES) geographic area for the meager urgent care/emergency care equipment available from the institution(s) and government(s) for that area.

Residential segregation is an example of access racism. In fact, Williams (1999) argues that residential segregation is the single most important type of racism that negatively affects health, primarily because it affects access to adequate educational systems and medical care. Ernst (1999) has documented access racism, showing that biologic expressions of race result in discriminatory social interactions that produce racial disparities in morbidity and mortality. Experiencing institutionalized access racism has been linked with experiencing poorer health status, and this association appears to be strongest with mental health outcomes (Williams, Neighbors, and Jackson 2003). Another example of access racism in health services emerges from stories of “economic profiling” of minority patients, such as the following: several years ago, the African American wife of a preeminent medical scientist in Houston (also African American) visited a particular health clinic for the first time and, before being given the opportunity to identify her health insurance, was “profiled” and brusquely routed to a financial office and forced to fill out forms—to see whether she could qualify for Medicaid!

In contrast to these two types of institutionalized racism, Jones 2000 Jones 2001 defines internalized racism as feelings of resignation, helplessness, and hopelessness. These feelings may manifest themselves in risky health behaviors. One example of internalized racism is cultural racism. A person living in an environment characterized by cultural racism could, for example, possess a lifestyle steeped in unhealthy habits, most of which are the result of longstanding institutional racism (e.g., the higher proportion of fast-food restaurants located in minority neighborhoods, the lack of easy accessibility to fitness centers and playgrounds, and the lack of safe neighborhoods in which to walk or jog for exercise) and impoverishment. Individuals exposed to cultural racism may decide that a healthier lifestyle, even if desirable, is simply too difficult to achieve in their (unhealthy) home environment. This lack of health self-efficacy may often be accompanied by broader doubt about achieving a brighter future for themselves in general. In a VA health services context, the effects of cultural racism on patients may, for example, be seen in veterans who visit their VA facility for an illness, then fail to adhere to their provider's treatment recommendations, or simply give up further treatment altogether, not because of lack of access or lack of knowledge of the illness and what treatment can do to cure it, but rather because they cannot envision themselves being able to succeed in changing their health habits in their home environment.

We contend, therefore, that it would be a useful pursuit for health services researchers to endeavor to develop measures—even crude ones—of the prevalence of the above-described forms of racism in patients' lives, with a goal for the near future of having this information collected reliably from patients along with the conventional categorizations of race. The real and potential impacts of these forms of racism could then be estimated for each patient and, optimally, interventions to minimize these impacts could be taken by health care providers. This type of strategy is already being used with some degree of success in the context of mitigating health illiteracy in racial minority patient populations (Rudd, Goldberg, and Dietz 1999).

Many researchers have recently argued that the terms “race” and “ethnicity” should not be used synonymously (Sheldon and Parker 1992; Hahn and Stroup 1994; McKenney and Bennett 1994; Senior and Bhopal 1994; Warren et al. 1994). Writings on this issue have suggested that one's race and/or ethnicity should be treated as an affiliation rather than a genetic predisposition, and individuals should be extended the respect of being allowed to specify the affiliation(s) of their choosing, in a way that suits them (Bhopal, Rankin, and Bennett 2000).

Race, as it is used in health-related research, consists of personal identity and group identity facets as well as the more familiar biological indicators. Ethnicity, in contrast, is most commonly used as an entirely social–political construct, referring to the sharing of a common culture, including shared origin, shared psychological characteristics and attitudes, shared language, religion, and cultural traditions (Sheldon and Parker 1992; Chaturedi and McKeigue 1994; LaVeist 1994; Senior and Bhopal 1994; Beutler et al. 1996; Freeman 1998). Thus, ethnicity refers to cultural identification, which is fluid and may change over time. For example, Sillitoe and White (1992) report that while in the early 1980s individuals in Britain responded with no comment to an item assessing their ethnicity as West Indian, in more recent years, Britons of this background have successfully lobbied for the term “black British” to be used instead. The reason for this change is that many “West Indians” now currently residing in Britain are Britain-born.

The concept of ethnicity has evolved to its current conceptualization as being a construct separate from a person's race, although in many cases the two co-occur. This growing recognition has led to the realization that each of the racial groups of Asian American, African or black American, American Indian or Native American, and Caucasian or white American includes a series of ethnic groups. For example, persons of Hispanic ethnicity include white, black, and Asian races (phenotypes), while persons of sub-Saharan African ethnicity(ies) are almost exclusively of black race (phenotype) and persons of Pacific Island ethnicity are almost exclusively of Asian race (phenotype). This raises a critical point: boundaries of ethnicity are not precise, and may be fluid across geopolitical boundaries. This is true not only at the international level, but at the intranational level as well. As a result, the term “ethnicity” is not always understood by study participants, even across adjacent local communities, and likely requires further interpretation in a relevant local context. Ethnicity may have an indirect effect on health outcomes by influencing health beliefs, the manner in which symptoms are expressed, physical functioning, entry into health service delivery systems, and medical treatment processes (Atkinson, Casas, and Abreu 1992; Marin, Gamba, and Marin 1992; Williams and Jackson 2000). In fact, some researchers suggest that ethnicity (i.e., cultural identification) be assessed as part of the clinical encounter, in order to make sense of patients' responses to treatment.

The strength of relationship between ethnicity and health outcomes appears to be influenced by “acculturation,” that is, the extent to which members of an ethnic group have adopted the beliefs and practices of another ethnic group. Acculturation occurs when individuals from one ethnic group come into contact with individuals from another ethnic group (Redfield, Linton, and Herskovits 1936). Acculturation may be assessed in a variety of ways, such as by examining individuals' language and food preferences, social activities, or political identification (Padilla 1980). In some cases, higher levels of acculturation are correlated with the adaptation of negative health behaviors and subsequent poorer health outcomes (Hubert, Snider, and Winkleby 2005; Rosenberg, Raggio, and Chiasson 2005; Vaeth and Willett 2005), while in other cases lower levels of acculturation are correlated with poorer health outcomes (Rahman et al. 2005; Zsembik and Fennel 2005). These associations appear to be related to the specific ethnic groups examined (i.e., Mexican Americans versus Latinos from Caribbean islands) (Zsembik and Fennell 2005) and to the outcomes being measured (i.e., dietary practices versus health services use and self-perceptions of health) (Lara et al. 2005).

Next, we present a brief overview of the U.S. Government's official policy for categorizing race and ethnicity. The key role of such a policy cannot be overemphasized; the 2003 Institute of Medicine Report, Unequal Treatment: Confronting Racial and Ethnic Disparities in Health Care (Institute of Medicine 2003, pp. 31–35), highlights the importance of accurately measuring race and ethnicity as variables in health services research. Therefore, having a unified, federal policy in place for guiding the collection of race and ethnicity data is critical.

Of particular importance for health services researchers is the fact that, under revised Directive Number 15, respondents are to be permitted to mark more than one racial identification category. This provision enables analyses that compare survey responses between individuals who choose a single race and those who choose multiple races (e.g., African American race only versus African American race plus another race).

Next, we briefly describe three different approaches for measuring race and ethnicity as variables. While we do not advocate any one of these approaches, we think it is appropriate to highlight them. These approaches are particularly important in VA health services research, because we know that issues related to utilization of health services and treatment efficacy are closely tied to racial and ethnic health disparities.

Importantly, on the 2000 Census, individuals self-identified as being Hispanic or non-Hispanic. This item preceded the racial identification item in an effort to reduce confusion on the part of respondents (Sawyer 1998). Table 1 shows the wording and structure of the Hispanic ethnic identification item, which includes Mexicans, Mexican Americans, Chicanos, Puerto Ricans, and Cubans as Hispanics, and allows the respondent to write in another Hispanic identification. The only versions of Census 2000 that did not include this categorization were the forms used in American Samoa, the Commonwealth of the Northern Mariana Island, Guam, individual forms used in the Pacific Island and the U.S. Virgin Islands, and military forms used in the Pacific Islands.

Table 1 Racial and Ethnic Identification Items from Census 2000

Table 2 Racial, Ethnic, and Religious Identification Items Used in the Sillitoe and White (1992) Study

The SMAS (Stephenson 2000) was developed and validated to assess the degree of ethnic identification (extent of acculturation) among individuals from five ethnic groups. Acculturation was defined as the degree of immersion in dominant and ethnic societies (Stephenson 2000). Immersion was defined as language, interaction, food, media, and level of acceptance by the dominant culture. The SMAS was not designed to measure cultural change among acculturating individuals. This instrument was developed using an ethnically diverse research team including community professionals and consultants, and was tested among African Americans (n=6), individuals of African descent (n=18), Asian Americans (n=6), caucasian (European) Americans (n=14), and Hispanic Americans (n=10) (Stephenson 2000).

The psychometric properties of the SMAS were assessed using exploratory factor analysis. The 32-item SMAS was found to include two factors. These are ethnic society immersion (ESI, Factor 1) and dominant society immersion (DSI, Factor 2). The ESI factor includes 17 items and the DSI factor includes 15 items. Coefficient α's were 0.86 for the entire scale, 0.97 for Factor 1, and 0.90 for Factor 2. The range of item-total correlations was 0.51–0.87 on Factor 1, and 0.57–0.83 on Factor 2. In terms of validity testing of Factors 1 and 2, it was discovered that both ESI and DSI were significantly correlated with ethnic group affiliation (r=−0.39, p<.001 for ESI and r=0.46, p<.001 for DSI) (Stephenson 2000). In addition, Factors 1 and 2 of the SMAS have been found to correlate significantly with two other widely used instruments designed to measure ethnic identification/acculturation, the Acculturation Rating Scale for Mexican Americans-II (ARMSMA-II) (Cuellar, Arnold, and Moldonado 1995) and the Bidimensional Acculturation Scale for Hispanics (BAS) (Marin and Gamba 1996).

Our intent in providing the instruments described above is to draw attention to the variety of approaches available for measuring race and ethnicity as variables. These approaches may be particularly applicable for use in health services research conducted with VA patient populations, which are becoming increasingly more racially and ethnically diverse.

We expect that this commentary communicates the need for researchers to rigorously explore reasons for the associations among racial and ethnic group status, socioeconomic status, and health outcomes. Understanding these associations is critically important for health services delivery. As Sheldon and Parker (1992) note, the inconsistencies in the measurement of race and ethnicity across many research studies would not be tolerated in the measurement of other study variables. The importance of this work for health services research is highlighted by the fact that without some standard means of measurement, the validity of comparisons of health-related data based on these constructs is questionable.

Thus, a need exists to develop ethnically competent instruments that function equivalently in different ethnic groups. However, in order to ascertain the function of instruments in different ethnic groups, it is first imperative to develop consistent measures of race and ethnicity that can be used in health services research and medical care interventions. Other aspects of health services research related to the valid assessment of race and ethnicity include assessments of racial- and ethnic-associated use of specific types of health services and procedures, disease incidence and prevalence, differences in health outcomes, and mortality rates.

Our recommendations for health services researchers build upon a call for consistency by Ford et al. (2002), who suggested that all investigators use a core set of basic demographic items while also using expanded, population-specific modules. This would allow data pertaining to members of these population groups to be described in greater detail, while retaining a core set of items that could be used in a consistent manner in many different studies. More than one expanded module could be used per site, in addition to the core set of items. For example, research sites in Southern California could use an expanded module for Asian Americans, which could contain categories such as Korean, Chinese, Japanese, Vietnamese, etc. Research sites in New York could use an expanded module for blacks, which could include categories such as Barbadian, Haitian, Jamaican, Nigerian, Panamanian, Senegalese, Trinidadian, etc. (Ford et al. 2002). Individual researchers using existing data could combine these data into core demographic categories based on the Census 2000 classification scheme. This would make possible a comparison of researchers' collected data with Census-based population estimates. Improved and consistent methods of data collection could result in higher-quality data being collected, which could lead to the identification of race- and ethnicity-specific health services needs. Jones (2001) suggests that investigators explicate the reasons for any racial and ethnic health disparities they find in their data. These investigations have the potential to lead to the development of interventions designed to reduce or eliminate these disparities.

In addition, individual researchers could need the following suggestions related to the use of race and ethnicity data in health services research. First, race- and ethnicity-related differences in quality of care, access to care, treatment efficacy, and other outcomes related to health services research could be vigorously investigated and explicated (Jones 2001; Institute of Medicine 2003; Betancourt and Maina 2004). Second, the effects of socioeconomic status on race- and ethnicity-related outcomes could be examined, with the purpose of disentangling these effects (Williams 1999; Jones 2001). Third, the effects of racism on study outcomes could be examined for members of different racial and ethnic groups. Fourth, investigators could provide a rationale for the collection of race and ethnicity data (i.e., collection of this information is related to a documented disparity) (Jones 2001). Fifth, the manner in which race and ethnicity data were collected could be reported (i.e., self-reported, observer rating, allowance for selection of multiple categories, etc.) (Jones 2001). Sixth, new data demonstrate the influence of geographic differences on quality of health care received (Baicker, Chandra, and Skinner 2005). Therefore, the interactive effects of geographic location, race, and ethnicity also need to be examined in future research.

It is imperative that health services researchers find ways to measure race and ethnicity more accurately. This is of great relevance to VA research, because the VA patient population is rapidly becoming more racially and ethnically diverse. In order to enhance the quality of VA health services research, we must now identify the best ways to measure race and ethnicity in a valid and reliable manner. In this way, the effects of race and ethnicity on assessments of quality of care, access to care, and treatment efficacy can be conducted with greater precision by VA health services researchers.

The research reported here was supported by the Measurement Excellence and Training Resource Information Center (METRIC) of the U.S. Department of Veterans Affairs, Health Services Research and Development Service. Dr. Kelly is supported by a career development award granted by the U.S. Department of Veterans Affairs, Health Services Research and Development Service. The views expressed in this manuscripts are those of the authors and do not necessarily reflect the views of the U.S. Department of Veterans Affairs.

This research was also supported by Department of Defense Grant No. DAMD 17-96-1-6246; National Institutes of Health R24 EXPORT Center Grant No. RFA-MD-04-002; and National Institutes of Health Resource Center for Minority Aging Research (RCMAR) Grant No. 1 P30 AG 21677. The authors thank Mr. Ken Kato, Ms. Ellen Matthiesen, and Ms. Shannon Hancock for their assistance with the manuscript preparation.