From Race and Ethnicity as Proxy Variables to Measuring What It Means: The Next Evolution of Health Equity Research

The aggregation of racial/ethnic data in public health and biomedical research has long obscured the health needs of Asian and Native Hawaiian and Pacific Islander (NHPI) communities. Although the 1997 revisions to Office of Management and Budget (OMB) Directive 15 called for separating these groups, many government agencies and researchers continued to combine them, even up to the present, referring to them as “Asian or Pacific Islander,” “Asian,” or even “Other.” This practice falsely implied uniformity across diverse populations with distinct biological, cultural, and social experiences. As a result, critical health disparities were often masked, sometimes to meet statistical assumptions rather than being empirically informed, which stalled progress toward health equity.

In this issue of AJPH, Li et al. (p. 1671) demonstrate the importance of moving from broad, higher-order racial/ethnic categories to more specific, lower-order groupings to uncover health inequities. By disaggregating Asian and NHPI data into distinct Asian and NHPI subgroups, they revealed that only certain subgroups drove the disparities in late-stage HIV diagnosis seen when subsumed under their larger group “Asian and NHPI” compared with Whites. For 2010 through 2014, increased odds were seen among Filipinos, Native Hawaiians, and other Pacific Islanders. From 2015 through 2022, elevated risk remained for Filipinos, Asians with no ethnicity data, and Pacific Islanders other than Native Hawaiians.

IMPROVING THE SPECIFICITY OF RACIAL AND ETHNIC DATA

Li et al.’s study contributes to a growing body of research that goes beyond OMB recommendations by disaggregating Asian and NHPI data into specific subgroups. Similar efforts have been made in studies of COVID-19 risk and mortality,¹ cancer incidence,² and cardiovascular disease prevalence.³ The 2024 revisions to the OMB standards now encourage collecting more granular racial/ethnic data that better reflect individuals’ identities. Other scholars have also emphasized the need for culturally respectful and meaningful disaggregation for Asian and NHPI populations.^4,5

It is important to note that Li et al. separated Native Hawaiians from other Pacific Islander groups but did not further disaggregate smaller subgroups (e.g., Sāmoan or Chuukese), despite collecting race/ethnicity information on 10 distinct NHPI groups, because of small sample sizes (< 10 individuals) and confidentiality concerns. Similarly, for Asians, they collected data for 12 distinct Asian subgroups, but aggregated some of them. Although aggregation is often justified to avoid unreliable risk estimates, broad racial categories can produce misleading and highly inferential results. Nevertheless, their decision to disaggregate at an intermediate level enhanced data granularity beyond their broader, higher-order categories.

Disaggregating racial/ethnic data to the most detailed level can help create effective policies, targeted interventions, and better resource allocation. Li et al.’s work identified Asian and NHPI subgroups at the greatest risk for late-stage HIV diagnosis. Delayed HIV diagnosis can cause severe damage to the immune system, life-threatening infections, premature death, treatment difficulties, psychological distress, and increased transmission risk. Li et al. did suggest that the improvement in late-stage HIV diagnosis in Native Hawaiians over time may be due to the increased attention and resources (e.g., education and access to testing) directed their way. Their findings highlight the need for precise racial/ethnic data to promote health equity.

HANDLING MIXED-RACE/ETHNICITY DATA

What is not clear in Li et al.’s study is how they treated mixed-race Asians and NHPIs. Following OMB definitions (e.g., “A person having origins in any of the original peoples of Hawaii, Guam, Samoa, or other Pacific Islands”), it is likely that they included both single-race and multirace Asians and NHPIs in their higher-order analysis of racial data. However, how they assigned persons who were Asian + NHPI is not clear in their report. Nevertheless, it highlights an important issue in health equity research: how to fairly classify mixed-race individuals.

According to the 2020 US Census, about 17% of Asians and over half of NHPIs identify with more than one race, yet many may choose their respective Asian or NHPI identity as primary. Unpublished data from the Native Hawaiian Health Research Project show that over 90% of people with Native Hawaiian ancestry prefer this identity despite multiple racial backgrounds. However, a common practice is to separate single-race Asians and NHPIs from their mixed-race counterparts without additional context to guide classification. For example, research that distinguishes between single-race and mixed-race Asians and NHPIs, including combinations like Asian + NHPI or NHPI + White, has found higher rates of cardiovascular disease among mixed-race individuals than their single-race counterparts.⁶ Without more detailed information, the causes of these disparities remain unclear, underscoring the need for more nuanced approaches to racial/ethnic categorization.

Recently, researchers have explored improved methods for analyzing multiracial/ethnic data. One study compared two approaches: the additive model, where every race/ethnicity a person selects is counted, and the mutually exclusive model, where individuals are assigned to a single racial/ethnic category.⁷ Using 2021 Medicare Consumer Assessment of Healthcare Providers and Systems (CAHPS) survey data and treating all possible race/ethnicity combinations as the gold standard, researchers found that although both models performed similarly in predicting patient experience outcomes, the additive model approach was better at capturing differences related to social determinants of health, such as age, education, and mental health. The additive model avoids assuming that all multiracial individuals share similar experiences or risks.

If we accept that race/ethnicity is a social rather than a biological construct,⁸ we must recognize the risk of misclassification when individuals’ preferred identities are not captured—for example, by asking, “Which racial/ethnic background most influences who you are?” in the case of multiracial individuals. Without such information, the findings may lack validity because of misclassifications. Even more sophisticated analytical techniques—such as the additive model—which attempt to account for multiracial individuals more inclusively still rely on assumptions other than a person’s preferred identity or identities. These limitations underscore that race/ethnicity is a proxy variable—an imperfect stand-in for the complex social, cultural, historical, and structural forces that shape health.

MOVING BEYOND PROXY VARIABLES

There is growing recognition of the need for more nuanced approaches to understanding racial/ethnic disparities. As a social construct, race/ethnicity serves as a proxy for factors that shape health behaviors, risks, and outcomes, such as sociocultural norms and socioeconomic conditions; unequal access to education, quality health care, and other resources; historical and contemporary experiences; exposure to systemic and interpersonal racism; and power dynamics. For example, studies have shown that educational attainment,⁹ acculturation-related factors,¹⁰ socioenvironmental conditions,¹¹ and experiences of racism¹² can help to explain racial/ethnic differences in health risks and outcomes.

To advance health equity, we must move beyond using race/ethnicity as proxies and strive to measure the underlying social, structural, and historical factors they represent. Disaggregating data to more granular subgroups, understanding mixed-race identities, and applying innovative analytical approaches are important and necessary. However, these efforts must be coupled with a greater commitment to understanding and addressing the root causes of disparities, including systemic racism, socioeconomic inequities, and cultural context, and their nuanced effects across distinct racial/ethnic groups. This is the next step in the evolution of health equity research.