Socially-assigned race and health: a scoping review with global implications for population health equity

The following research questions were formulated to guide the scoping review:

Search strategies were developed by the public health librarian (NT) using controlled vocabulary and free-text terms combining two main concepts: (1) socially assigned race and (2) health outcomes (Additional file 1). The published literature was systematically searched in PubMed, Web of Science, EBSCO, and Proquest (Additional file 2). Due to the low number of results in our preliminary searches, we expanded our search to all subscribed databases within the EBSCO and Proquest platforms using the “Choose databases” and “Change databases” features respectively. The search strategy was inclusive of all countries, years of publication, and study designs (qualitative, quantitative, or mixed-methods). Searches were limited to English-language studies published in the peer-review literature through May 2019 and updated on September 27, 2019. The WHO Global Health Library, Bielefeld Academic Search Engine (BASE), Directory of Open Access Journals (DOAJ), MedNar search engine, OpenGrey, EThOS (British Library), and Google Scholar were searched for grey literature and unpublished reports. Reference lists of relevant grey literature, as well of the included studies were hand-searched to identify additional studies. Colleagues and researchers in relevant fields were contacted for assistance in identifying additional studies. Peer-reviewed journal articles were included if (1) socially-assigned race was the primary ‘exposure’ of interest in the analysis and (2) socially-assigned race and the association with either a health, healthcare, or health-related outcome was examined. Studies that included socially-assigned race as a covariate or control variable were excluded.

The final search results were exported into EndNote and duplicates were removed. Each title and abstract was screened for inclusion by two independent reviewers (JAL, KW). The full-text of the remaining articles were screened independently by two reviewers (JAL, KW) to determine if the inclusion criteria were met. If there were any discrepancies that arose between the two reviewers, a third reviewer (JC) was consulted to resolve the disagreement by consensus.

The study team developed a data-charting form to determine the appropriate variables to extract and document the relevant study characteristics. Two reviewers (JAL, KW) independently charted the data and discussed the results. The articles included in the sample were characterized by article identifiers (e.g., authors, year of publication), study identifiers (e.g., country, sample size, study design), study sample characteristics (e.g., N, age, sex, race/ethnicity), purpose of study, socially-assigned race/ethnicity measure, outcome, and result by narrative format. Tables were created to reflect the studies included, study designs, publication years and key characteristics of the study populations. An assessment of the quality of the studies was not performed, which is in alignment with scoping review methodology.

Studies operationalized socially-assigned race either as perceived (the race one believes others assume you to be) or observed (by an interviewer based on observable characteristics). The majority of studies ascertained perceived socially-assigned race by respondents’ answers to the question: “How do other people usually classify you in this country?” [6, 11, 20, 23‐25, 30, 33, 34]. In the United States, this question is most frequently collected via the Behavioral Risk Factor Surveillance System (BRFSS). BRFSS is an annual state-level, telephone survey of non-institutionalized adults living in the United States that monitors health behaviors, chronic health conditions, and utilization of preventive health services. The socially-assigned race question is collected via an optional module or included as a state-added question among all racial/ethnic groups the survey is administered to. Other studies conducted in the United States examining socially-assigned race that focused specifically on the experiences of Latinx populations, utilized the 2011 Latino Decisions/ImpreMedia survey. This survey is nationally representative of Latinxs and is designed to ascertain information about attitudes and experiences of health, health care, policy, discrimination, and detailed information on national origin, nativity, acculturation, and citizenship. In New Zealand, socially assigned race was collected during the administration of the New Zealand Health Survey (NZHS; 2006–2007). NZHS collects self-reported data from individuals aged 15+ on physical and mental health and health service use. In Veenstra [29] the question about socially-assigned race was queried among Canadians living in Vancouver and Toronto as: “… And what about other people who you meet, what racial background do other people tend to think you are? Do they think that you’re white, Asian, South Asian, Black, Southeast Asian, or Aboriginal, or perhaps some combination of these, or maybe something else I haven’t mentioned?“ [29].

In several studies, socially-assigned race was observed, where the interviewer classified participants’ race/ethnicity or skin tone (as a proxy for race) [7, 12, 15, 31, 36]. Several studies utilized data from the National Longitudinal Study of Adolescent Health to Adults (Add Health). In Add Health, at the end of the interview, interviewers were asked questions about the participant including assigning the participant to a racial category based on the interviewer’s observation [7]. Perceived appearance, one’s own subjective evaluation of their appearance, was used to capture socially-assigned race in two studies [32, 37]. In another study, the interviewer’s rating of skin tone was used as a marker of socially-assigned race for its “generalized perception of other” [13]. Lopez et al. uses a slightly different approach to measure socially-assigned (i.e., ascribed) race by querying respondents about self-perceived race, socially-ascribed race and street race [33]. The socially-ascribed race was asked similarly to the question on the BRFSS. The street race question asked: “If you were walking down the street, what race do you think other Americans who do not know you personally would assume you were based on what you look like?” There were five street race categories including: white, Latinx, black, Arab, and Mexican.

In the United States, the collection of socially-assigned race using data from the BRFSS is inclusive of multiple racial/ethnic groups and multiracial individuals. Research has shown that congruence between self-identified and socially-assigned (concordance) appears to be greatest among individuals that self-identify as white or black (98.4 and 96.3% respectively.) [20] Disagreement between self-identification and socially-assigned race (discordance), occurs most frequently among Latinx, AI/AN, NHOPI, and multiracial groups [14, 20‐22]. For example, Jones et al. demonstrated that among those who self-identify as Latinx, 63.0% were socially-assigned as Latinx, 26.8% white and 3.5% black [20]. Among American Indian, 47.6% were classified by others as white and for NHOPIs, 35.1% were socially assigned as white [20].

Socially-assigned race studies conducted in New Zealand, included Māori, Pacific peoples, Asian and European ethnic groups [25]. In this context, concordance between self-identified and socially assigned race was greatest among individuals who self-identified as European (97.6%, which is considered the dominant ethnic group) and Asians (92.7%). Discordance was highest among individuals who self-identified as Māori, Pacific, and those who identified as multi-ethnic [25]. In a study conducted across several Latin American countries, interviewers reported their assessment of respondent’s race/ethnicity according to the following 5 categories: white (blanca/branca); mixed-white (mestizo/parda), mixed-black (mulato), indigenous (indigena), and black (negra/preta) or other [31]. It was reported that interviewers’ classifications of respondents matched respondents’ self-classifications in 61% of cases [31]. In a study with respondents from Vancouver and Toronto Canada, a lower percentage of Black (73.9%) and South Asian (65.7%) respondents reported a greater mismatch between self-identified and socially-assigned race in comparison to Whites (95.9%) and Asians (91.3%) [29].

The multidimensionality of race is rooted in theory about reflected appraisals which specifies that an individual’s idea of self is in part derived from social interactions with others [33, 38, 39]. The literature appraised emphasizes a relational dimension of socially-assigned race that identifies “a group’s location within a social hierarchy (e.g., minority versus majority status)” (page 251 [9]) and underscores how this hierarchy differentially affects group well-being and health [40]. Social assignment of race/ethnicity is experienced according to others’ perceptions and in part reflects a racial hierarchy rooted in accrued privilege [14, 25]. The hierarchy confers an advantage to individuals racially classified as white and penalizes those perceived by others as a member of a historically oppressed group (i.e., lower on the racial hierarchy).

The question that arises is whose perceptions are relevant and does this affect the construct validity of a measure of socially-assigned race? When using a measure of perceived socially-assigned race, we rely on the respondent to indicate their perception of how others are likely to ascribe them and surveys seldom inquire about the race/ethnicity of the perceiver. An underlying assumption of socially-assigned race is the significance of the classification of members of a higher-status or dominant group. These members typically have greater access to power and resources and have a tendency to reify historical, institutional and systemic inequalities that foster and maintain the power dynamic [16]. However, the racial/ethnic background of the “perceiver” is unknown. Moreover, how one is perceived may vary by the race/ethnicity of the perceiver [14]. To our knowledge, datasets that collect socially-assigned race, do not additionally inquire about the race/ethnicity of the “perceiver” or the situation in which the external classification occurs. Further, it is not clear whether there are certain factors that may influence how one perceives external racial attribution. Vargas and Stainback [14] sought to contextualize factors that influence incongruence between self-identified race and socially-assigned race using data from the Portraits of American Life Study [14]. Their findings suggest that individuals who reported a mismatch (incongruence) between self-identity and social-assignment were more likely to report lower levels of ethnoracial unity (i.e., feel less close to other members of self-identified racial group) and racial identity salience (i.e., lack of connection with other members of self-identified racial group) in comparison to individuals who were congruent on self-identification and social-assignment [14]. Qualitative research designs may prove to be particularly informative in systematically evaluating these issues which may help improve the construct validity of measures of socially-assigned race.

The health impact of the generalized perception of others may differ by the racial/ethnic groups targeted for racialization [18]. An important conceptual consideration is related to assumptions surrounding classification as a lower-status or “minority” group member. Racial classification reflects physical, socioeconomic, and cultural perceptions of an individual [21]. There may be observed differential impacts on health based on the type of perceived racial classification. In the United States, some have argued that there is a hierarchical system of racial classification that presupposes racial discrimination. For example, Latinx populations may be perceived as white, Latinx, black/African-American. The extent to which patterns in health risk are associated with perceptions as Latinx, versus black/African-American, versus multi-racial group is unclear. It is possible that the health risk may mirror the intraethnic heterogeneity of health outcomes such as diabetes which align with the racial stratification of Latinx groups [41]. More specifically, racial differences in diabetes prevalence are highest among Latinxs who self-identify as black (i.e., Puerto Ricans, Dominicans) in comparison to those who self-identify as white or Latinx [41‐43]. Comparisons of these health differences have been under-investigated in large part due to insufficient sample sizes and are worthy of further exploration.

Another conceptual consideration is the choice for reference group. This is particularly applicable in studies exploring level of agreement between self-identified and socially-assigned race. The majority of studies in this review, that were conducted in the United States, used self-identified non-Hispanic whites who were socially-assigned as non-Hispanic whites as a referent group. This choice of reference is theoretically relevant for studies probing the health advantage of being perceived as white. However, alternative choices for referent groups, for example, being self-identified and socially-assigned as a non-white racial/ethnic group have also been employed in studies to facilitate interpretation of the outcome [23].

There are several methodological issues related to study design, data availability and analytic strategies that deserve further attention. The majority of studies assessed were quantitative. However, employing qualitative or mixed-methods research designs would be an important contribution to further elucidate the mechanisms underpinning socially-assigned race and health. Utilizing these designs has the potential to gain in-depth understanding of one’s lived experience that may help to generate robust theories and elucidate pathways through which social-assignment is related to health. Further, the detail information obtained from qualitative techniques could be useful for informing the interpretation and corroboration of quantitative data. Additionally, qualitative techniques that combine innovative approaches such as the use of multimedia vignettes or simulated and virtual reality platforms may be used to assess the scope of bias due to socially-assigned race among health care providers.

The availability of data sets collecting and documenting socially-assigned race poses a challenge to generating future research investigating socially-assigned race and health. Overall, the studies were cross-sectional, with longitudinal investigations remaining unexplored. Large population-based surveys such as the BRFSS collects data on socially-assigned race. However, the use of this survey has been limited because states opt-in for the collection of this data. Beginning in 2002, selected states included the Reactions to Race module, which also asked questions about race consciousness, emotional and physical reactions to race-based treatment, and perceived differential treatment in employment and healthcare settings. The Reactions to Race module is not included as part of the core component of the BRFSS questionnaire that comprises a set of standard questions asked by each state each year. Instead, it has been considered an optional module, where states make the choice to adopt the modules to be administered for a given year. Stepanikova et al. [24] used socially-assigned race collected in BRFSS by pooling data across years (2004–2013) and 17 states and the District of Columbia to yield a large sample size [24]. In recent years, some of the questions from the BRFSS Reactions to Race module were included as ‘state-added’ questions - when an individual state elects to include questions of their choosing that may include a subset or a single question from an optional module or validated scale. Inclusion of state-added questions are not reported on the main BRFSS website and can be only be determined by reviewing each individual state’s BRFSS data documentation. This presents a challenge, because it is difficult to determine the extent to which states are administering the socially-assigned race question or other Reactions to Race module questions.

The utility of socially-assigned race in BRFSS has been critiqued for its lack of representativeness, particularly for national Latinx populations [11]. Because of the limited number of states that administered the Reactions to Race module, it largely reflects Latinx populations that are predominantly Mexican, whereas states that have higher concentrations of Puerto Ricans, Dominicans, Central Americans, and Cubans have not well represented. Surveys such as the 2006 Portraits of American Life, the 2011 Latino Decisions/ImpreMedia, and the Latino National Health and Immigrant Survey include samples that are intended to be more representative of the Latinx population in comparison to the BRFSS data. These surveys have the capacity to explore differences in socially-assigned race by finer delineations of national origin, acculturation, and citizenship. While these data sets are ideal to answer questions and understand socially-assigned among Latinx populations, the data sets tend to have smaller sample sizes, are not conducted in consecutive years, and collect limited health outcome data in comparison to the BRFSS.

The socially-assigned race literature can benefit from the extension and focus on other racial/ethnic groups or historically oppressed populations. However, the issue of sufficient sample size is a major challenge for examining socially-assigned race in other racial/ethnic groups such as Native Americans, Native Hawaiians and Other Pacific Islanders, multi-racial, or indigenous populations. A study that was conducted using data from Vancouver and Toronto, Canada could not analyze data for mismatches between self-identified and socially-assigned race for Aboriginal and Southeast Asian populations [29]. Further, we identified a dearth of data sets that contain socially-assigned race measures data on adolescents. One study used data from Add Health to examine the link between socially-assigned race and health among Native American and white adolescents [36]. However, in our review of the literature this was the only study conducted among adolescents.

Future socially-assigned race research would benefit from theoretically driven analytical considerations related to model specification. Many of the included studies are minimally adjusted for potential confounders. Included studies have assessed the association between socially-assigned race and self-reported health without adjusting for behavioral factors and health characteristics (i.e. physical activity, smoking, BMI, fruit/vegetable intake) which have been documented to be associated with self-rated health. Additionally, using measures of citizenship status, nativity, and time in the U.S. as potential effect modifiers may help clarify some of the observed health patterns among Latinx populations. In addition to including other health-related covariates in analyses and conducting more detailed assessments, we posit that coping styles may also differ by ascribed race and should be examined.

The majority of studies examined the main effects of socially-assigned race on health outcomes. Further explanation of mechanisms and potential effect moderators by the relationship between socially-assigned race and health may afford a theoretical foundation to disentangle processes that influence racialization and subsequent inequities in health and healthcare. Most studies included measures of socioeconomic status, such as household income, educational attainment and occupational status as mediators of the association between socially-assigned race and health. Few studies have yet to test the suggested mechanisms through which socially-assigned race is posited to operate such as exposure to individual-level discrimination. Moreover, potential variables identified as effect modifiers, such as neighborhood racial-ethnic residential segregation or stress buffers (e.g., vigilance and anticipatory stress) that may diminish or amplify health effects, have rarely been explored.

Although, our knowledge of racial health inequalities is predominantly ascertained from studies that measure self-identified race/ethnicity, we see great utility in incorporating measures of socially-assigned race in population health studies. It is imperative that we advocate for and include questions about socially-assigned race in addition to other multidimensional measures of race in representative population-based datasets. The incongruence between self-identified and socially-assigned race can help in answering questions related to the persistence and maintenance of racial health inequalities that warrant further empirical investigation. Towards this end, there is a real need for inclusive race/ethnicity data collection efforts in our public health monitoring and surveillance systems and surveys to move closer to achieving health equity.

Use of a single, unidimensional measure of race does not provide sufficient detail about intraracial processes of racialization and health. Studies determining the explanatory power that socially-assigned race has in differentiating intraracial experiences of race and racism and subsequent variations in health are needed. We have a minimal understanding of the extent to which socially-assigned race captures variations in population health. There are increasing efforts to disaggregate the health status of Latinx populations according to foreign-born status and country of origin to capture additional variation in health profiles. However, fewer studies capture racial heterogeneity among Latinx population by using a measure of socially-assigned race to broaden our knowledge regarding Latinx health inequalities. For example, the patterns of health risk and advantage have not been fully explored among Latinxs who are socially-assigned either as Latinx, Black, white, or some other racial/ethnic group. Socially-assigned race-specific reporting of health may uncover variations that are obscured by the use of self-identified race. There are some findings which suggest that educational and economic profiles also vary by socially-assigned race and we know from prior research on self-identified race and health, that these factors are part of the pathway through which “race” influences health [44].

Additional research that considers how measuring socially-assigned race affects population health disparities estimates are warranted and have the potential to provide greater insight into the health consequences of the social construction of race and potential targets for social and policy approaches to address inequality. The findings also call attention to the magnitude of racial/ethnic health and healthcare inequalities and how these estimates may be affected by the way race/ethnicity is collected and measured [8, 35]. Saperstein [35] compared the association between interviewer-classified race and self-identified race to receipt of various health screenings (e.g., pap smear, blood pressure, and breast exam) among women [35]. The results suggested that interviewer-classified race, as compared to self-identified race, was a stronger predictor of racial differences in health screenings. A study that used data from BRFSS found that self-identified and socially-assigned race were both independently associated with perceived discrimination in health care [24]. The findings from this study revealed that socially-assigned race was a better predictor of perceived discrimination in health care in comparison to self-identified race. Another study evaluated self-reported race/ethnicity and interviewer-ascribed race/ethnicity and income inequality in Brazil [45]. The magnitude of association between interviewer-ascribed race/ethnicity and income inequality was larger when compared to a measure of self-reported race/ethnicity. These studies illustrate the importance of considering multiple dimensions of race. In the aforementioned examples, the mechanisms of inequality were best represented by a measure of socially-assigned race. The operationalization of race such that it is ascribed by someone else may resemble racial discrimination and implicit bias and could lead to a more appropriate estimation of the magnitude of disparities. These examples also show how the reliance upon a single measure of race, namely self-identified race/ethnicity, can underestimate the level of health inequities. However, it is not clear the extent to which one measure of race may be more (or less) strongly associated with health or factors that influence health. It is possible that the relative strength of self-identified versus socially-assigned race varies by health outcomes and may be a function of theoretically distinct mechanisms that are responsible for the health disparity. Saperstein [35] demonstrated that socially-assigned race was a stronger predictor of health outcomes that were encounters in health care or clinic settings versus self-identified race which was a stronger predictor of group differences in outcomes. An explanation of this difference is related to the inherent value given to others classification of one’s race and the implicit biases and prejudices that accompany it [35]. Scholars contend that socially-assigned race may be more closely associated with institutionalized racism and experiences of discrimination [29]. Assumed cultural differences and stereotypes about one’s race may be more salient to the quality of interactions with health care providers, receipt of a health screening during a medical visit, or receipt of pain medication in an emergency room, more so than self-identified race. Future lines of inquiry would substantially benefit from the thoughtful, theory-driven selection of a specific dimension of race and health outcomes since different measures of race may offer different explanations or have different implications for addressing the inequality [35].

The social construction of race/ethnicity and racial hierarchies around the world varies depending on the social, historical, and political context of an area. Though substantial work has contributed to our understanding of the racialization process, vis-à-vis socially-assigned race and health, much of this work comes from a United States perspective and has been largely conducted among Latinx populations. Though there are few studies on Native Americans, it is not clear to what extent the health status of members of other racial/ethnic groups are differentially affected by socially-assigned race. Exploring socially-assigned race in other areas (e.g., Europe) with increasing ethnic diversity would be informative. While the implications of socially-assigned race extend beyond the United States to other regions (e.g., Latin America) and countries (e.g., New Zealand) around the world, we need to be cautious about extrapolating findings from one area to another. Research could benefit from a deeper understanding of the process of racialization among other non-dominant racial/ethnic groups or indigenous peoples outside of the United States context. Relatedly, new areas of research interest may entail exploring an expanded set of health care and physiological health outcomes for a more comprehensive picture of health. Moreover, new technologies (e.g., machine learning, automation, algorithms, and phenotype recognition) may influence social assignment of race. While there is evidence regarding the impact of these technologies on racial bias and profiling, the potential impact on health and healthcare related outcomes is unknown and warrants future study. Given the substantial gaps in qualitative and quantitative data collection on socially-assigned race, creative ways to link existing datasets with big data sources and use of innovative qualitative techniques may provide opportunities to generate new insights and a comprehensive understanding of the relationship between socially-assigned race and population health across various contexts. Future scholarship that includes socially-assigned race as a variable to measure and monitor population-level health status and track the racialized experiences of historically oppressed, marginalized, and indigenous groups around the world is a crucial next step for population health inequalities research.