Racism Measurement Framework: A Tool for Public Health Action and Accountability

Racism has long been recognized as a principal determinant of health inequity,1,2 but recent events have galvanized local and state leaders across the United States to declare racism a public health emergency,3,4 with at least 23 resolutions across the state of Ohio, including Franklin, Cuyahoga, Hamilton, Lorain, Summit, and Montgomery counties, along with the cities of Akron, Canton, Columbus, Cleveland, Dayton, and Youngstown, among others.5–17 Such declarations coincide with reports highlighting the disproportionate impact of COVID-19 on communities of color and the Black Lives Matter protests that were catalyzed by recent deaths at the hands of law enforcement.18–20 The dual crises of COVID-19 and long-standing racism have put health inequities into the center of our collective awareness, further supporting the Columbus City Council’s demand for a “data-driven focus on poverty, economic mobility, and other factors that impact the social determinants of health (SDoH).”6 Several proposed conceptual models of SDoH address racism,21–23 but in the absence of a measurement framework there remains a gap in applying a data-driven approach that will enable health systems, government agencies, academic centers, community organizations, and public health professionals to combat racism.

COMMENTARY search do not challenge the underlying white framing ("the overarching worldview that encompasses important racial ideas, terms, images, emotion and interpretation, and lens by which white supremacy is perpetuated") of racial disparities in health outcomes, and call upon researchers to develop research questions and methodologies that "take apart and critically analyze health inequities within the context of the white racial frame." 27 Our aim in this paper is to propose a measurement framework with a series of metrics that captures the multidimensionality of racism and its effects on health. The framework is intended to provide a comprehensive list of indicators to inform monitoring and evaluation of anti-racist strategies that tackle social and health deficits attributed to racism.

METHODOLOGY
Racism is "a system of structuring opportunity and assigning value based on the social interpretation of how one looks, that unfairly disadvantages some individuals and communities, unfairly ad-vantages other individuals and communities, and saps the strength of the whole society through the waste of human resources." 28 It is important to note that multiple definitions of racism exist; however, most consider racism to be multidimensional and functioning on multiple levels. Figure 1 illustrates our conceptual model of racism that draws heavily from Jones' theoretic framework and reflects dimensions of systemic racism, interpersonal racism, and internalized racism-and their points of intersection. 29 We began our narrative review by reflecting on seminal articles of "Race and Health" as well as articles that emerged using the simple search term of "racism measurement." [28][29][30][31] From there, we used citations (citations of the paper and citations within the paper) to expand our literature search to identify more recent reflections and measurement indicators. Literature was gathered as it aligned with the dimensions in the conceptual model: systemic, interpersonal, and internalized. Metrics were included in the measurement framework if they had been used in at least 2 studies related to the measurement and/or description of disparities caused by racism.
This conceptual model is informed by a review of the literature and reflects the established multidimensionality of racism, often organized into 3 distinct, yet overlapping, domains: systemic, interpersonal, and internalized. The measurement framework is developed to characterize the levels of racism, the impacts of racism, and the hypothesized pathways between racism and health outcomes. The framework can be used to develop a call for action to prioritize strategic planning goals and interventions that tackle racism at multiple levels and to monitor and evaluate whether those efforts are creating sustainable changes towards health equity.

COMMENTARY
Gaps in measurement were identified from the literature as either a dimension where no metric was found or a metric with substantial quality concerns. Lastly, we present recommendations for how this framework can be applied in the public health context in Ohio to understand health disparities and assess the effectiveness of anti-racism activities across multiple dimensions of racism.

Systemic Racism
Systemic racism refers to the control of and access to labor, material, and symbolic resources within a society along racial lines. It is comprised of (1) structural racism, which involves the cumulative and compounding effects of an array of societal factors, including the history, culture, ideology, and interactions of institutions and policies that systematically privilege White people and disadvantage people of color; and (2) institutional racism, which refers to the unfair policies and discriminatory practices of particular institutions, such as schools and workplaces, that routinely produce racially inequitable outcomes for people of color and advantages for White people. 30 It represents "the ways in which societies foster discrimination through mutually reinforcing inequitable systems…creating interconnected systems that embed inequities" in practices, laws, and policies. 31 As such, systemic racism can be measured as the effects of policies on health (access and outcomes). In this section, we provide a brief overview of examples of racist policies and measures that track their impact on health outcomes. There are a number of well-established metrics for systemic racism in the policy sectors of housing, voting, criminal legal, economic opportunity, and health care, and Table 1 demonstrates how these metrics are linked to the SDoH and health outcomes.

Housing
Residential housing patterns are the most well-studied metric of systemic racism. Residential segregation affects socioeconomic status (SES) and social mobility, as residence determines access to education and employment opportunities. 32,33 America's history of systematic and institutionalized racism led to "redlining," whereby neighborhoods were color-coded on maps, with communities of color designated as undesirable and color-coded red, making it more difficult and expensive for Black Americans to obtain loans and purchase homes. 34 In 1937, the Home Owners' Loan Corporation redlining maps codified racial residential segregation into federal policy and practice from the 1940s through the 1960s, 25 where redlining remains at the heart of modern residential segregation metrics. For instance, contemporary Census tracts have been classified by the redlining maps to measure the Black-White disparity in odds of mortgage loan denial. 35 Metrics. Two established measures include (1) index of dissimilarity, which is the proportion of Blacks or Whites who would have to move out of their neighborhood to achieve an even distribution, or complete integration; and (2) isolation, which is the degree to which Blacks have potential contact with non-Blacks. 32

Voting
Measures of systemic racism in voting policy are outlined below. It is necessary to consider the effects of voter suppression laws, which are known to disproportionately affect people of color. As an example, 70% of voters in Georgia that were purged from the voter registration rolls were Black, 36 and across the country, 1 in 13 Black Americans cannot vote because of restrictive measures that have disproportionate effects. 36,37 Metrics. Established measures in voting policy include proportions of Blacks to Whites who (1) were registered to vote, (2) actually voted, and (3) were elected to the state legislature. 38

Criminal Legal
Racial disparities in arrests and incarceration contribute to disparities in health outcomes caused by racism, with a recent study indicating that a greater number of criminal legal contacts among Black individuals contributes to more depressive symptoms, compared to White individuals. 39 We also must understand how the outcomes of the criminal legal system vary based on the color of the victim. Studies of execution outcomes point to inequity of Black and White lives, as defendants convicted of killing White victims were executed at a rate 17 times greater than those convicted of killing Black victims. 40,41 Metrics. Measures of racism in the criminal legal system include proportions of Blacks to Whites who were (1) incarcerated, (2) disenfranchised felons, and (3) on death row. 35,38

Economic Opportunity
There is a strong correlation between the racial opportunity gap and the racial mortality gap, reflecting how differences in health outcomes are tied to differences in income, poverty, unemployment, and education. 42 When considering the effects of racism on economic opportunity, it is necessary to consider the generational accumulation of wealth and privilege. 35,43 Wealth disparities stem from policies that perpetuated racial divides in the labor market, namely the New Deal's Fair Labor Standards Act of 1938 that largely excluded workers of color from its labor protections. 44 While this policy has been expanded, agricultural and domestic workers are still exempt from many protections, where racial and ethnic minorities comprise 58-70% of the workforce. 44 Metrics. Established measures of economic opportunity include (1) earnings rank gap, which is the sex-specific difference between a Black individual's percentile in the Black earnings distribution and the position in the White earnings distribution 45 ; (2) racial opportunity gap, which is the sex-specific difference in income percentile ranking in adulthood between Black and White children born to families at the same income level in the same county 42

Health care
To assess systemic racism in health care policy, we consider racial biases in coverage and services received by both public and private insurers. In terms of coverage, the proportion of physicians participating in Medicaid is declining, influencing what health services exist in a given locale. 46,47 Restricted health access for Medicaid patients disproportionately affects racial and ethnic minorities, who are 2.5 times more likely to have Medicaid coverage than non-Hispanic Whites. 46,47 To add to the disparity, we see that White patients receive better quality of care than Black patients. 48,49 Furthermore, health systems that rely on commercial prediction algorithms can exhibit significant racial bias, with Black patients having significantly greater illness burden than Whites at the same level of algorithm-predicted risk. 50 Similarly, the Coronavirus Aid, Relief, and Economic Security (CARES) Act allocated funds to hospitals as a function of their past revenue, which disproportionately affected Black counties, who received the same level of relief funding as counties with lower COVID-19 burden and financial need. 51 Such racial biases arise because existing algorithms predict health care costs rather than illness, but health care costs are a poor proxy for health care needs because health-related expenditures are impacted by barriers to health care access. Notably, barriers to health care access may be further worsened in the transition to telehealth with the additional barriers of the digital divide and telehealth literacy, resulting in reduced use of health services among Black patients. 52 Metrics. Measures include (1) proportions of physicians participating in Medicaid by county racial demographics or neighborhood segregation indices, (2) illness burden at same level of algorithm-predicted risk 47,50 and (3) digital divide associated with telehealth. 52

Interpersonal Racism
As previously noted, there is a considerable distinction between needing and receiving health care, which provides insight into interpersonal racism in health care. Interpersonal racism, whether intentional or unintentional, refers to discriminatory practices and beliefs directed at persons in the targeted social group by persons in the privileged, or dominant, social group, taking the form of suspicion, devaluation, scapegoating, and dehumanization. 29,53 Interpersonal racism and systemic racism often overlap because these interpersonal practices and beliefs become embedded in structural relations. 29 This interplay between interpersonal and systemic racism arises from power dynamics and extends across multiple sectors, with studies highlighting the negative effects of subtle indicators of racial identity on employers' perceptions of qualifications, competence, and "professionalism." [54][55][56] Health care provider bias has particularly salient effects on health outcomes, as providers' perceptions regarding patients' value, self-reliance, competence, and deservingness lead to inequitable medical testing, referrals, treatment, procedures, and pain management. [57][58][59] Providers' implicit biases are particularly pernicious towards Black women, often reinforcing stereotypes of Black women being uneducated, licentious, problematic drug users, with an overdependence on public assistance. [60][61][62] Biases are present in more subtle ways, including exhibiting fewer positive, rapport-building nonverbal cues with Black patients. 63 Interpersonal racism, in conjunction with systemic racism, contributes to the already disparate health outcomes in Ohio. It is important to note that the purpose of measurement is not to merely document disparities but to incite anti-racist action by identifying opportunities for intervention. 64,65 Recognizing that these discrepancies in health outcomes are the result of social barriers, with provider bias being among the most frequently cited barrier, 66 medical education and health systems have adopted cultural competency training as a standard of practice, evidenced by the national standards for Culturally and Linguistically Appropriate Services (CLAS) in health care. 67 Many studies, however, have deemed cultural competency training inadequate, and others have noted its ability to ignore oppressive structures of power and to compound cultural, ethnic, and racial stereotypes. [68][69][70][71] As an alternative to cultural competency training, structural competency has been proposed as a framework for responding to social, political, and economic structures that influence the health of patients. 70,72 The Structural Vulnerability Assessment Tool was developed for clinicians to take inventory of patient vulnerability and, in this way, can minimize the influence of provider biases on patient care plans. 68 Metrics. Health services metrics provide a reliable means of capturing interpersonal racism between patients and providers as well as the interplay between interpersonal and systemic racism. Such metrics include patient satisfaction with health services and perceived quality of care; trust in health system and professionals 73 ; communication and relationship with health service professionals; frequency of delaying, or simply not getting, health care; and visits and admissions to hospitals and emergency departments. 74

Internalized Racism
The demarcation between interpersonal and internalized racism is often nebulous, which demonstrates that even though racism is multidimensional, these different levels interact and build off of one another. Internalized racism is the incorporation of attitudes, beliefs, or ideologies into one's worldview that maintain or exacerbate the unequal distribution of opportunity across ethnic and/or racial groups. 75,76 Such racism takes on one of two forms: internalized oppression or internalized dominance, both of which hinge on the concept of embodiment, in which engagement with the material and social world transforms the body. 76,77 As such, internalized racism can contribute to allostatic stress, causing accelerated aging and earlier onset of disease due to biological "weathering," 78 as well as patient disempowerment and fatalism, which negatively affect health care utilization and preventive care. 79

COMMENTARY
A number of survey instruments exist, each tailored by target audience and survey length, including, but not limited to, the Everyday Discrimination Scale, Major Experiences of Discrimination, Chronic Work Discrimination and Harassment, Heightened Vigilance Scale, Index of Race-Related Stress (IRRS), Perceived Ethnic Discrimination Questionnaire (PEDQ), Workplace Prejudice/ Discrimination Inventory (WPDI), Experiences of Discrimination (EOD), Racism and Life Experience Scales (RaLES), Multidimensional Inventory of Black Identity (MIBI), Schedule of Racist Events (SRE), and Workplace Racial Bias Measure (WRB). 30,80,81 Metrics. Surveys are a common measurement technique to assess internalized racism, such as the Perceived Racism Scale, which assesses behavioral and emotional responses to the experience of racism as well as frequency of exposure. 82 Surveys typically ask about distinct contexts (workplace, health care, school, etc) to capture the interplay between internalized, interpersonal, and systemic racism. 83 Despite their ubiquity, surveys' reliance on selfreport data presents a unique set of problems, namely responseshift bias and social desirability bias. 84,85 The Implicit Association Test (IAT) can be adapted to evaluate whether those most affected by discrimination are willing to recognize and report it as such, thereby minimizing cognitive problems affecting self-report data. 86

OPPORTUNITIES FOR USE OF MEASUREMENT
Following the review of the metrics associated with systemic, interpersonal, and internalized racism, we have identified several gaps that can be addressed with a more robust measurement framework. Understanding the interplay between racism and health is not enough, however; as Nancy Krieger notes, "the reason to study how injustice harms health is not to prove that injustice is wrong, since it is, by definition…the reasons are to deepen understanding of how injustice shapes population health…and to generate evidence for accountability." 53 To "generate evidence for accountability," as Krieger puts it, we have identified gaps in the literature, outlining the following list of opportunities for use of measurement, which is not about new measures, but, instead, using accurate measures consistently and mapping such measures to investments to become more accountable.

Systemic Racism
(1) Develop a baseline understanding of racial disparities across sectors in Ohio. One starting point would be the development of a "Disparity Report" for the State of Ohio, mirroring the initiative of the New York City Disparity Report, which includes indicators in education, economic security and mobility, health and wellbeing, and personal and community safety, with each metric segmented by race. 87 The Ohio Department of Health has put forth reports on disparities in infant mortality and pregnancy-related mortality caused by racism, 88,89 but to share disparities in an actionable way, organizations and institutions should embrace the framework described herein to consider the many dimensions that contribute to such outcomes. Table 2 in the Appendix provides an example of a more comprehensive analysis that is possible with existing measures, presenting disparities in health outcomes alongside disparities in the built environment, economy, education, and criminal legal system, as it presents intermediaries between racism and metrics shown in Table 1. More work, however, is needed to incorporate aspects of structural racism in preparing health equity and anti-racism data reports. 70,90,91 (2) Understand how existing algorithms contribute to racial bias, which sectors are affected, and other labels on which the algorithm can be trained to mitigate the effects of racial bias. Using the health system's algorithm for high-risk care management programs as a continued example, alternative label choices include avoidable costs (eg, emergency visits and hospitalizations), number of chronic conditions, and a combined health prediction with cost prediction, which can reduce racial bias by as much as 84%. 50 Highlighting the importance of how an algorithm is trained is relevant beyond the health care sector. For instance, credit-scoring algorithms predict outcomes by income, a measure that reflects structural inequalities related to employment and salary. 92 Similarly, hiring algorithms exhibit bias with respect to race and sex. 93,94 Further research is needed on the data sources, such as claims data, that contribute to the racial bias in algorithms when cost prediction is used.
(3) Explore dimensions of ethnicity. To assess the extent of variation within racial and ethnic groups and identify which subgroups are most affected, researchers should collect more detailed ethnic data (eg, breaking down Hispanic into Cuban, Dominican, Mexican, and Puerto Rican). Further sampling questions regarding citizenship, voting rights, and place of birth of participants, their parents, and their grandparents will not only enable better understanding of ethnic diversity but also provide an opportunity to more precisely study how contemporary immigration policy adversely affects health. 28,95,96 (4) Use new tools in policy analysis to quantify impact. Within policy analysis, there are ample opportunities to contribute because there is a lack of empirical evidence to identify which mechanisms of segregation-educational opportunity, labor market, or housing quality-should be tackled first, which would have the largest impact, and which is most likely to trigger ripple effects to other pathways on an individual's health. 95 The development and deployment of a "Disparity Report" could help in this regard. To identify the proximal mechanisms that link segregation to health, data scientists and statisticians must leverage (a) new models that capture dynamic historical processes that influence each other over time at multiple levels of analysis to provide a new way of seeing and understanding persistent problems, and (b) new statistical techniques, such as instrumental variable (IV) analyses, to estimate the causal effects of segregation by reducing or eliminating bias related to selection and unmeasured confounding variables. 97,98 Ohio Journal of Public Health, December 2020, Vol. 3, Issue 3 ISSN: 2578-6180

Interpersonal Racism
There exist a few opportunities for health systems to assess how individual and systemic changes can promote the adoption of antiracism in care delivery.
(1) Capture a baseline measure of potential effects of interpersonal racism in health care. Quantitative and qualitative indicators should be considered, such as patient satisfaction with health services and perceived quality of care; trust in health system and professionals; communication and relationship with health service professionals; frequency of delaying, or simply not getting, health care; and visits and admissions to hospitals and emergency departments. 74 (2) Evaluate how clinical organizations adhere to anti-racism efforts such as cultural competency standards. Although cultural competency has for many years been proposed to mitigate the effects of interpersonal racism in health care, more research has shown that structural competency in terms of changes at the health system level are needed to affect behavioral change. 29 Baseline measures could provide better insight into the benefits of training in structural competency rather than cultural competency (or other existing standards of practice) as well as facilitate routine monitoring of progress in its institutional anti-racism agenda, much like Ohio State University Wexner Medical College has outlined, with its commitment to evaluate efforts and deliver annual anti-racism reports. 90 The CLAS Standards require health care organizations to establish culturally and linguistically appropriate goals and policies that are incorporated into the organization's clinical operations, providing an early roadmap, but a more elaborate measurement framework is needed to evaluate performance against the standards and inform quick course correction if efforts do not result in systemic change. Organizations such as Multiethnic Advocates for Cultural Competence, 99 which is a community expert in cultural competency training, would serve as an excellent partner in developing organizational and practice-based measures that align and/or expand on the CLAS standards. Ohio's COVID-19 Minority Health Strike Force Blueprint has also highlighted several strategies for dismantling racism through cultural and linguistic competency. 100 (3) Develop better methods of quantifying discrimination. Additional research is needed to determine optimal approaches for comprehensively measuring discrimination in all its dimensions, including chronicity, recurrence, severity, and duration. 95

Systemic Racism
(1) Participate in community-based and community-engaged research and evaluation. Among academic centers, there is an even greater need for community partnerships and collaborations, evaluating interventions for (a) the extent to which they may be differentially effective across diverse subsets of the populations and (b) cost-effectiveness, but it is critically important that researchers engage with communities and build on their existing knowledge and insights by incorporating community members in all phases of research including study design, data collection, analysis, and dissemination. 34,95,101 In Ohio, there are several organizations invested in addressing the SDoH and racism through effective interventions, as highlighted in Table 2 in the Appendix. These models need to be evaluated, disseminated, and adopted by others attempting to address complex problems. When pursuing community-engaged research, research budgets should take into account the need for translation resources and services as well as compensation for survey respondents. 34

Interpersonal Racism
(1) Consider adoption of the Structural Vulnerability Assessment Tool. The Structural Vulnerability Assessment Tool operationalizes the concept of structural competency and provides a way to assess patients' nonmedical social needs, while minimizing the influences of provider bias. 68 Importantly, the Structural Vulnerability Assessment Tool includes questions about discrimination, providing an opportunity for medical centers and health systems to assess discrimination without having to administer another survey or questionnaire to a subset of the patient population.

Internalized Racism
(1) Reconsider race as a variable in research. As a starting point, researchers must reconsider race as a mere "dummy variable," 102 taking into account their a priori assumptions of how race plays into the model as a way of recognizing potential biases, which is a critical component in the Public Health Critical Race Praxis (PHCRP) framework. 1,27,34 The PHCRP contributes an ant-racism lexicon based on critical race theory (CRT) that avoids the repeated use of "vulnerable," a common euphemism embedded in public health literature that is used to soften, if not avoid, the discussion of racism. 103,104

CONCLUSION
The COVID-19 pandemic, the deaths of George Floyd and Breonna Taylor among countless others, coupled with the resurgence of the Black Lives Matter movement, has advanced investments in anti-racism efforts, but it remains to be determined if the implementation of these efforts will result in meaningful change. 105 A measurement framework is needed to understand the issue of racism more deeply, challenge the white racial framing by considering racism's many dimensions, 27 and elucidate whether antiracism efforts are changing the pathways toward more equitable health access and outcomes. In this way, organizations and agencies can apply this framework to their own anti-racism agendas to differentiate what is working from what is not and to identify what can be generalized to other settings. This level of evaluation is needed to translate declarations of "racism as a public health emergency" into actions, advancing the pursuit of health equity in the state of Ohio.