In addition to the data-driven model of causal fairness, we propose a normative framework to adjudicate morally permissible uses of race in precision health—that is, uses of race in precision health that conform to the moral standards (i.e., ethical considerations) necessary for the practice of good medicine30. Here, we understand morally permissible to mean morally optional, in that to perform the action (in this case, using race in precision health) is neither morally required nor morally prohibited. We have chosen to focus on morally permissible uses of race in precision health because we are not (yet) convinced that using a racial classification in precision health is ever morally required. However, we are convinced that wholesale elimination of race in precision health is not required, so there may be contexts in which it would not be morally prohibited30.

The moral standards most relevant to the current discourse include the virtues of justice, benevolence, and trustworthiness, where justice is understood as respect for another’s moral status and the consequent rights that accompany such a moral status; benevolence as goodwill toward others, or the proper care and concern of another; and trustworthiness as the state of being technically and morally competent in whatever is being entrusted. Here technical competence is a variety of know-how about the technical aspects of precision health, and moral competence is how individuals or institutions are disposed toward the aforementioned virtues30. Note, these virtues cohere with The NASEM Report’s guiding principles for “scientifically valid and trustworthy research”17. Additionally, like The NASEM Report’s guiding principles, we believe these virtues are “mutually reinforcing”17. Justice, benevolence, and trustworthiness are necessary conditions to promote the aims of precision health in a way that provides total health and well-being and does not compromise the integrity of its recipients31. Justice preserves the dignity and respect of those receiving care, and consequently, promotes more holistic well-being. Benevolence ensures proper disposal to the moral considerations relevant to the person(s)/community of interest. Trustworthiness among health care professionals and their institutions is particularly important, given that the asymmetric dynamic between health providers and their patients exposes patients to a certain level of vulnerability.

Unlike The NASEM Report, we emphasize virtue, that is “excellences” of character. Where principle-based frameworks (like The NASEM Report) often focus primarily on action guidance, virtue-based frameworks (like ours) prioritize the character of the relevant persons and institutions, while also providing action guidance. Thus, in what follows, while we provide requirements for adjudicating morally permissible uses of race in precision health, we will be less concerned with providing a mechanical decision procedure and more concerned with highlighting the proper conditions for cultivating the sort of dispositions that promote good medicine when working with diverse populations. We are concerned both with what medical practitioners and stakeholders do and with how they are disposed toward their practice/investment (i.e., how they are). This is, in part, because rules can be misapplied and abused if those beholden to them do not know or are not inclined to value the virtues necessary for good medical practice. Still, despite the differing approaches, we take our convergence on moral standards with The NASEM Report to add further support for those standards; it demonstrates a sort of robustness of evidence.

Finally, note that these standards distinguish themselves from legal norms in that legal norms are standards of evaluation that are primarily concerned with the scope, nature, and legitimacy of political structures, whereas justice, benevolence, and trustworthiness—while relevant to legal norms—are also concerned more generally with human and societal flourishing and well-being. Given these considerations, we will understand morally permissible actions to be actions that, whether performed or not, are conducive to the promotion of justice, benevolence, and trustworthiness, and thus, human and societal flourishing and well-being.

Morally permissible uses of race should be able to prevent or mitigate the social harms associated with race-based medicine, including (amongst other things) unjust treatment due to essentialism8, while also preserving the benefits of using race in medicine, including the ability to track racism’s impact on health outcomes and medical care13,32 and the etiology of medically relevant genetic differences between human subpopulations for both monogenic and polygenic traits and diseases33. Given the above, we propose that researchers and clinicians use the race-in-medicine (RIM) normative framework to determine morally permissible uses of racial classifications in precision health30. According to RIM, it is morally permissible to use race in medicine if and only if: (a) when applicable, social determinants of health are sufficiently engaged prior to or in tandem with the use of race (social determinants requirement), (b) the medical end(s) sought cohere with the aim of medicine and are best acquired using race (harm minimization requirement), and (c) the use of race does not violate the relevant just legal norms constraining medical practice more generally (legal norms requirement). These requirements are necessary to fulfill, because without fulfilling them, practitioners of precision health risk violating the virtues of justice, benevolence, and trustworthiness30. What’s more, these requirements are sufficient, because taken together, they properly address the relevant objections and concerns raised when considering the use of race in medicine30.

Note, RIM is agnostic toward race theories in that it does not privilege a particular theory of race and is applicable whatever one’s theory of race may be30. Additionally, RIM is comprehensive in that it is meant to cover all branches of medicine—medical research, diagnosing, treatment, and education—and consequently, all these branches as they relate to precision health. Finally, RIM is comprehensive in its application at the individual (patient-physician) level, at the institutional level (e.g., medical schools and hospitals), and at the governmental level.

Given the above, how might precision health practitioners and stakeholders apply RIM’s requirements? Note, fulfillment of each condition will vary depending on context and competency. So, in what follows, we offer a case study of glomerular kidney disease, providing additional examples when useful. Glomerular kidney disease is associated with APOL1 genotype, which is common in populations of sub-Saharan African ancestry33, and is characterized by racial and ethnic disparities in incidence, prevalence, treatment, and outcome, with Black American adults and Hispanic adults disproportionately negatively impacted34,35,36,37. Despite glomerular kidney disease’s association with APOL1 genotypes, risk for the disease is modulated by social determinants including, but not limited to, socioeconomic status36,37,38,39,40,41,42, neighborhoods and housing43,44, diet (including access to certain foods)45,46, insurance47, and exposure to air pollutants48,49. Given these known socioenvironmental factors, according to the social determinants requirement, research programs seeking to investigate glomerular kidney disease along racial lines would need to prioritize engaging these exogenous factors. For research teams, this may mean partnering with others or creating a subgroup of researchers who focus on the correlations between these exogenous factors and glomerular kidney disease before dividing genomes into groups that reflect some sort of racial classification. This shift may, amongst other things, require hiring a technically diverse group of researchers, including those trained to engage social determinants of health. Increased demographic diversity generates novel solutions50,51 so increasing the demographic diversity of precision health’s research workforce could help illuminate novel pathways previously overlooked. At the institutional level, this priority shift might be implemented through incentives like targeted grants, which could dramatically alter the research landscape. For example, the National Human Genome Institute currently earmarks just 5% of its congressional budget for studying ethical and social implications of genomics52, despite billions of dollars spent each year to treat illnesses linked to genetic as well as social and environmental factors53.

Engaging social determinants of health requires technical competence to identify confounding variables in medical research, as well as moral competence to intervene in environments where health burdens are largely modulated by unjust structures (e.g., housing discrimination and access to healthcare)54. One might worry that our recommendation to ‘engage’ social determinants of health is insufficiently prescriptive to offer genuine guidance to practitioners. However, like The NASEM Report, we acknowledge the context-sensitivity of clinical research and practice [17, p. 99]. We aim here to reflect that sensitivity by offering examples that illustrate how addressing individual psychological biases like essentialism may require modifying the surrounding social structures in which these biases are embedded, rather than providing a mechanical decision procedure that is ill-equipped to capture all the relevant medical and social contexts as well as “evolving best practices over time” [17, p. 99]. At the same time, the causal fairness framework described above serves as a more explicitly prescriptive framework for engaging social determinants of health that compliments RIM and can be embedded into RIM’s social determinants requirement (along with other frameworks for engaging social determinants of health).

We move now to the harm minimization requirement. Although the APOL1 genotype is common in populations of sub-Saharan African ancestry33, not every person of sub-Saharan African ancestry possesses the APOL1 genotype55. What’s more, not every person with the APOL1 genotype will develop kidney disease56. This suggests that a racial classification is ill-suited to accomplish the ends sought in studying the (epi)genetic and environmental structures implicated in glomerular kidney disease. Instead, a more fine-grained classification system that distinguishes between at-risk populations is better suited for investigation. This coheres with and encompasses The NASEM Report’s principles of equity and justice, validity and reproducibility, and transparency and replicability, as it takes seriously the ethical and social implications of model choice in precision health (equity and justice), demands that the level of classification reflect the population(s) observed (validity and reproducibility), and requires researchers to consider why and whether the classification scheme being used is best suited given the aims of the research program. It is also here, at the harm minimization requirement, that we believe our causal fairness model for precision health plays an important role. Again, under our causal fairness model, if the use of a racial classification for (say) diagnosis or recommendation does not lead to the same prediction for an otherwise identical patient who has the same genetic feature as a racial group but is not categorized in that race, then, according to the harm minimization requirement, an alternative classification scheme better suited to causal fairness should be used.

It should be noted, however, that in the case of medical research regarding glomerular kidney disease, racial self-reports may still be of value. This is especially true when considering the relationship between race and social determinants of health, given that the distribution of these determinants is modulated, in part, by racism13,31. Additionally, where genomic information is unavailable and it is too inefficient to test everyone, the use of self-reported racial membership may aid in evaluating how risk for glomerular kidney disease is distributed within and between populations. Finally, although the harm minimization requirement appears to preclude the use of racial classifications in precision health overall (as there appears to always be a better alternative), wholesale elimination of race is also not the answer because alternative classifications are subject to the same essentialist biases as race, as described above.

Finally, the legal norms requirement will require endorsement and adherence to current legal norms that demonstrate respect for the moral status of research participants and the consequent rights that accompany such a moral status. It is not enough that the practice be legal, since history is replete with examples of laws that violated the rights of individuals based on their racial membership. The legal norms informing the use of race in precision health must, then, be just30. For example, the NIH Revitalization Act of 1993 was meant to establish guidelines for the inclusion of minority populations in clinical research funded by the US federal government. However, with the advent of genomic research and population genetics, blind adherence to this act risked further entrenching the social imagination in essentialist assumptions. Thus, to prevent unjust use of the NIH Revitalization Act of 1993, application of these legal norms must be modified to address such risks.

Several legal norms constrain research on glomerular kidney disease. For example, the right to privacy torts can be understood as just legal norms, given they were created to protect the dignity of persons by preserving their ability to exercise autonomy over the most fundamental and intimate parts of their lives57. Any use of race in glomerular kidney disease research that unwarrantedly jeopardizes the anonymity of research participants should be emended or else eschewed. Additionally, in keeping with equal protection under federal law, glomerular kidney disease researchers reliant on genomic tests should make sure their use of participants’ results cohere with the Genetic Information Nondiscrimination Act of 2008, which is meant to protect individuals against group-based, genetic discrimination in health coverage and employment.

Though not an exhaustive examination of how just legal norms factor into precision health, these observations demonstrate the inescapability of legal systems when practicing precision health, and thus the need for a normative framework that explicitly considers the legal norms constraining precision health and medical practice more generally30. Of course, writing and emending legislation should not fall on the shoulders of precision health researchers and practitioners, given their limited resources and lack of legal expertize. But the legal norms requirement does signal that precision health should be a multidisciplinary collaboration, where medical researchers, IRB committees, and health institutes regularly partner with legal scholars, lawyers, social scientists, and (legal) philosophers with an eye toward the relationship between historically marginalized communities and the law.

Though not a mechanical decision procedure, RIM reinforces the virtues necessary for ethical precision health by addressing social determinants of health (some of which are modulated by historical injustices and racism), highlighting competing population descriptors, and acknowledging the need for just legal norms to constrain medical practice. An upshot of RIM is its broad scope allows it to be comprehensive and capture all relevant contexts in which precision health takes place.