The goal of precision medicine is to provide a more precise approach to the prevention, diagnosis and treatment of disease¹. To achieve this goal, precision medicine must be equitable; that is, it must be fairly applicable, available and accessible to all². Unbiased and accessible precision medicine can contribute to reducing avoidable and systemic inequalities in health status, that is, it can contribute to health equity, the state in which everyone has a fair and just opportunity to attain their highest level of health³.

Disparities in prevalence, incidence, clinical and serological manifestations, and outcomes of rheumatic conditions and rheumatological manifestations are well established, with some examples highlighted in Box 1. These disparities are pronounced in under-represented populations, namely racial minority and ethnic groups, medically underserved urban and rural communities, the uninsured and underinsured, older adults, and those with lower education and income⁴. Despite the disproportionate impact of many rheumatic diseases on these communities, the causes of these disparities remain elusive. Identifying the factors that promote disparities is a critical step towards equitable care and precision medicine^5–7.

Developing successful precision medicine initiatives necessitates addressing health disparities. Although precision medicine is meant to focus on individual variability (such as genes, environment and lifestyle), these initiatives are predominantly informed by population-level research⁸. The consequence is that if health disparities are not properly identified and addressed in research, precision medicine is poised to worsen health disparities for groups already disadvantaged by current practices and interventions^8–10. Frameworks for studying health disparities emphasize the need for ecological models that consider biology and environment, including social factors. In theory, this sounds akin to the needs of precision medicine research but in reality there is substantial isolation of biological research from epidemiological studies on environmental determinants of health.

Despite the recognized influence that social and physical environmental experiences have on rheumatic disease, it is not known how these experiences and exposures interact with biology to influence disease outcomes. To date, rheumatic disease research has largely focused on biological mechanisms without inclusion of effects by environmental exposures. Furthermore, genetic studies frequently fail to acknowledge the fact that non-genetic factors are substantial contributors to the disparities. In addition, the persistent use of vague and arbitrary categories of population differences continues to hamper rigorous and reproducible research¹¹. These issues have resulted in a knowledge gap regarding the interactions between genetic and environmental factors that contribute to disparities in disease outcomes. We explored the extent of this knowledge gap in rheumatic diseases by conducting a literature review of studies that intersected both molecular biology and social determinants of health (SDOH), including racism and discrimination. We identified a striking shortage of literature spanning both molecular and epidemiological disciplines. Further, we identified a strong bias in the amount of research focusing on molecular biology in comparison with epidemiological SDOH. This finding highlights the ‘siloing’ of research entities and the general challenges (such as data availability and research infrastructure) in implementing a true precision medicine paradigm that considers the full biological and environmental experiences of a patient; this phenomenon has been identified in other disease categories and has been presented as a cause of hampered progress in alleviating health disparities^12,13.

In this Review, we start by discussing clinical obstacles to health care equity, including access to health care and the use of inaccurate language when labelling population groups. We describe the biopsychosocial model and necessary data towards studying health care disparities, followed by a description of existing barriers in research, focusing on limitations in available data, research silos and the subsequent knowledge gap created in rheumatic disease studies. We then propose solutions and considerations, working towards a better understanding of the biology of rheumatic disease in the context of health disparities. Addressing this knowledge gap is expected to improve precision medicine diagnostics and therapeutics for underserved patients, thus working towards a more equitable paradigm in research and clinical practice.

In addition to disparities in disease susceptibility, prevalence and incidence, another source of health disparities occurs in the health care setting. Health care disparities can manifest in different capacities, leading to delayed or deficient patient care and education and poor patient outcomes. Travel distance to receive specialized care is often a deterrent and a major source of delayed patient care. When examining patients in the Medicaid beneficiary cohort (a government health insurance program that provides health care to some of the most vulnerable populations in the USA), poor access to specialized care for patients with systemic lupus erythematosus (SLE) occurred more often in those enrolled in Medicaid than in those with other health insurance¹⁴. Patients only enrolled in Medicaid travelled 11.5 more miles than those with additional or other health insurance and on average more than 19.8 miles to see a rheumatologist¹⁵. Consequently, patients need to use the emergency department not only for acute care but also as their primary source of medication prescriptions¹⁴. In patients with rheumatoid arthritis (RA) in the Canadian universal public health care system, those who lived >100 km from a rheumatologist were 50% less likely to be seen by a rheumatologist within 3 months of a suspected RA diagnosis by a primary care physician and had decreased continuity of care within the first year of diagnosis¹⁶, both of which are important for early initiation of therapy to prevent disability and damage. As a result of these delays, patients have an accumulation of joint damage requiring surgeries¹⁷. Not only is access to specialized clinical care and medications a source of unequal outcomes but it also contributes to a deficiency in patient education, specifically in addressing disease comorbidities, which further promotes health disparities.

These structural barriers to health care access are also deterrents to the achievement of the goals of precision medicine. Access to health care is not guaranteed in many countries. A comparison of health care models in the USA, Austria and Denmark showed that the effects of precision medicine depend on the structure of the local health care system; in contrast to the Austrian and Danish health care systems, the USA for-profit and insurance-based structure increases the risk that precision medicine will exacerbate existing health disparities and structural injustice^18–20. In addition to health care coverage, the current high costs of precision medicine therapies also preclude equity by excluding those who cannot access them. Changes are needed to prevent health care structures and pharmaceutical business models from supporting existing structural injustices and destabilizing egalitarian health policies¹⁸. Such changes include revisiting, monitoring, and regulating drug pricing models, to encompass the values of social equity and solidarity¹⁸.

Low representation in clinical trials and genomic studies is known to be perpetuated by multiple factors, including geographical distance, mistrust of the health care system, language barriers and fear of encountering implicit bias and stereotyping during care. Under-represented communities are often at a greater risk of poor health outcomes²¹. The confusion about population descriptors and lack of rigorous language when describing population categories continue to stigmatize many communities. Addressing individual biases and using clear and respectful population labels are paramount to realizing the promise of precision medicine²².

Many studies, especially in the USA, describe the epidemiology of rheumatic disease in terms of socially defined, discrete racial categories. Similarly, health disparities research traditionally focuses on self-reported race. However, this methodology poses challenges and limitations. First, race and ethnicity are imperfect proxies for important epidemiological information, including SDOH, such as racism and discrimination, economic stability, health care access and quality, education access, and physical and/or built environmental exposures²³. These social and physical environmental determinants are differentially experienced across racial and ethnic groups owing to historical and contemporary discriminatory policies and practices, resulting in health disparities across groups and geography. Hence, it is not race and ethnicity, but the lifetime of exposures, that contributes to health disparities. Consequently, discrete racial categories are unable to appropriately capture the variability of risks one person might be exposed to in their lifetime. As such, limiting health disparities research to racial categories can impose harm through ascribing risk to arbitrary groupings of heterogenous individuals while also obscuring the causal exposures facilitating risk²⁴. Another weakness of limiting research to racial categories is that many identities intersect, and thus, risk factors might impose differential or compounded effects. For instance, although increased socioeconomic status (SES) is protective, increased exposure to discrimination has the potential to offset these benefits^25–27. In addition, the concept of race is dynamic and changes over time in the wake of administrative decisions and cultural shifts; for example, the list of racial groups on the US Census has changed nearly every decade since the first enumeration in 1790 (ref. 28). Race is also a complex population descriptor because it can be interpreted and measured in many ways (for example, racial identity, racial self-classification, observed race, reflected race)²⁹. Although it may be a useful population descriptor for researchers who wish to measure a consequential form of social status and affiliation, the concept of discrete populations that are static in place and time does not apply to humans.

Another issue is the frequent conflation between race and genetic similarity groups²², often referred to as genetic ancestry groups. Genetic ancestry is a statistical estimate inferred from measures of genetic similarity across individuals using available data, thus without any relationship to the concept of race. The epidemiological literature reveals rheumatic disease disparities based on ascribed race, and as discussed below, the genetics literature has found rheumatic disease disparities based on genetically inferred groups. However, equating these population descriptors is a fallacy, as race is an imperfect proxy for SDOH, and the other captures genetic risk factors. Altogether, this issue highlights the necessity of a multi-faceted approach to studying health disparities, one that includes both genetic and non-genetic, individual-level and population-level stressors and exposures³⁰. Moving forward, it is critical for researchers to utilize accurate and appropriate population descriptions, such as those described by the National Academies of Science, Engineering, and Medicine²⁸ and by Feero et al.³¹

The biopsychosocial model was proposed in response to the reductionist paradigm of viewing diseases through a strict ‘biomedicine’ lens. Instead of viewing the full scope of phenotype and symptoms as explainable by molecular processes (biomedicine model), the biopsychosocial model asserts that health and disease result from an interplay between biological, psychological and social factors³² (Fig. 1a). The biopsychosocial model emphasizes the importance of integrating different domains (such as biological, behavioural, physical and/or built, sociocultural, health care system) and levels (for example, individual, family, community, society). In the context of health disparities, this conceptual framework posits that individuals who perceive certain circumstances as stressful experience a physiological stress response that can be modulated by adverse or protective sociodemographic factors (such as SES), psychological characteristics (for example, depression), behavioural factors (smoking, alcohol use), and coping responses (for example, ability to mobilize social support) to such experiences^30,33. Expansions of this model also include physical environmental exposures (for example, air and water quality)³⁴. Such environmental exposures can enact effects independent of individual perception and/or awareness and can be disproportionately distributed among neighbourhoods and by geographic region^35,36.

Successful implementation of the biopsychosocial model requires availability of appropriate data, measurements, and instruments for the multiple domains contributing towards health disparities (Fig. 1b). Physical environmental exposures are increasingly available at varying geographic granularity. For example, studies of air quality, ranging from local or regional measures³⁷ to national measures³⁸, highlight the increased prevalence of detrimental environmental exposures among minority and low-income populations. Broadly, environmental factors or determinants to health are usually classified into biological, chemical, physical and social environments. As defined by the WHO, other SDOH include economic security and equality, health behaviours and health care resources^39,40. There are many indicators for capturing environmental determinants of health; methodologically, these can be measured via self-reporting (for example, surveys) and/or aggregated data from population statistics with varying granularity^41,42. Given the broad domains of environmental determinants of health, it is critical to recognize that any single indicator will be unable to capture the full complexity of an individual’s environmental determinants of health, including SDOH⁴⁰. For instance, subjective social status instruments in which participants rank their circumstances (such as, income) relative to others in their community (for example, social ladders), have been found to capture perceived well-being⁴³, and thus might capture aspects such as perceived discrimination and social stress, whereas multiple deprivation measures might be better suited to capturing material and/or resource-based socioeconomic factors. Given the intersectionality of identities, and thus experiences and exposures, SDOH measures that are multi-faceted (including more than one indicator) and multi-level (that is, individual, community) are more likely to appropriately capture the complexity of these effects^40,44,45. In rheumatic diseases, data from each of these broad categories (biological, SDOH) are abundantly available. However, these fields are rarely intertwined to investigate disease outcomes under an ecological, combined molecular and SDOH model, severely hindering progress in health disparities research.

Decades of epidemiological research of health disparities in rheumatic diseases have highlighted the existence of systemic issues that drive poor outcomes for patients from under-represented and underserved communities (Table 1). For example, multiple social stressors, such as low household income, poverty, unemployment and racial discrimination, are associated with increased risk and severity of SLE^46,47 and RA^48,49. A marked socioeconomic gradient was reported for these diseases, with individuals in the most deprived group being up to 70% more likely to develop the disease than the least deprived group⁴⁹. Low income was associated with greater odds of osteoarthritis⁵⁰ and gout⁵¹, and worse outcomes for systemic sclerosis (SSc)⁵² and sarcoidosis^53,54. Social deprivation and social vulnerability were associated with osteoarthritis, RA, ankylosing spondylitis, sarcoidosis, and SSc^50,55,56. In countries with lower GDP, patients with psoriatic arthritis had higher levels of disease activity and more patient-reported impact of disease⁵⁷. Also, women, in particular women of understudied and under-represented populations, are disproportionately affected by rheumatic disease; and women, especially African American women, disproportionally experience psychosocial stressors⁵⁵.

Despite stressors co-existing in areas of concentrated poverty and with other environmental factors, protective factors might buffer the negative impacts of stressors⁵⁸. For example, social support decreased depression and anxiety in African American women with SLE^59–62, perceived social support improved depression, anxiety and quality of life in patients with RA^63–65, and social support and spiritual well-being were associated with life satisfaction in patients with SSc⁶⁶. Additionally, alleviation of stressors can influence disease. For instance, for patients with SLE, exiting poverty can mitigate the strong effect of living in concentrated poverty on disease damage⁶⁷. Most epidemiological research has focused on the identification of risk factors, but understanding the mechanisms and extent to which protective factors can offset risk factors is essential to address health inequities (Table 1).

Although critical, the historical focus on epidemiological issues in rheumatic disparities research has sustained an epidemiological research silo (Fig. 2 and Supplementary Table 1, Supplementary Fig. 1) through its failure to intertwine molecular and biological effects. That is, the molecular and physiological mechanisms through which social and physical and/or built environment exposures affect disparities are unknown. For example, air quality and particulate exposures have been associated with disease activity and drug response in rheumatic diseases^68–70, but the biological mechanisms require more exploration. Similarly, epigenetic marks affect gene expression and can govern cell function and physiological response to environmental factors, including SDOH such as SES^71,72 and general perceived stress⁷³. Yet, despite the growing awareness of the effect of social stressors on the epigenome, research aimed at identifying and quantifying the regulatory mechanisms through which social exposures influence rheumatic outcomes is only just emerging⁷⁴. Understanding the interplay and underlying mechanisms by which various positive and negative determinants of health influence epigenomic variation, and how the resulting biological changes might contribute to rheumatic health disparities, will enable biomedical research to directly address outcome disparities.

Large-scale studies into the biological aetiology of rheumatic diseases most commonly focus on genetic variation associated with disease. This line of research has been productive and important. Genome-wide association studies (GWAS) have identified hundreds of genomic loci associated with rheumatic diseases, providing important insights into the biological pathways contributing towards disease. Given the potential to facilitate disease diagnosis, support risk prediction, improve cost and efficiency of trials, and accelerate personalized care, results from these GWAS have been used to construct genetic risk scores (GRS)^75,76. Given that many of these risk factors have been reported to vary among populations (for example, along geographic gradients), genetic variation has been proposed as a partial contributor towards disparities^77,78. In fact, the 2006–2009 strategic plan of the US National Institute of Arthritis and Musculoskeletal and Skin Diseases prioritized the identification of genetic contributions to health disparities, citing as a rationale that rheumatological conditions, including SSc, SLE and osteoarthritis, cluster in populations targeted by health disparities initiatives⁷⁹. However, to date, few genetic risk factors for rheumatic diseases have been linked to health disparities, which is due, in part, to limitations of genetic data, current data availability and the general siloing of genetic research from pertinent environmental and exposure data.

First, it is crucial to understand the general limitations of genetic data, especially in the context of health disparities research. In genetic studies, individuals are categorized into groups on the basis of genetic similarity to genomic reference groups, such as the 1000 Genomes Project populations⁸⁰. Importantly, however, these genetically assigned categories, often labelled continental ‘ancestry’ groups, are not equivalent to self-identified and socially defined categories of race and ethnicity^81,82. Thus, risk allele differences among these genetically defined groups cannot be equated to explaining observed racial health disparities. Furthermore, mapping differences in genetic risk variants among populations requires careful interpretation. Genetic studies disproportionately comprise European-ancestry populations. Even studies including under-represented populations are often more limited in total sample size. Hence, comparison and portability of identified risk variants should always be considered in the context of potential differences in statistical power^83,84. Another complicating factor is linkage disequilibrium, which refers to the tendency of linked alleles to be co-inherited. Most GWAS-reported associations are not causal variants, but instead, ‘tagging’ variants that are highly correlated with the functional variant. Linkage disequilibrium varies by geographically or ancestrally distinct populations. Thus, a genetic association (tagging the causal variant) in one population might not reproduce in another. In studies comprising samples from multiple distinct study populations, linkage disequilibrium can be a powerful fine-mapping tool, as illustrated by SLE studies^85–87. Unfortunately, the availability of large, multi-population cohorts remains limited for most rheumatic diseases. The consequence is that, to date, GRS are built upon associated, tagging variants, which frequently show poor predictability in diverse populations⁸⁸. Thus, as they are currently built, these genetic prediction tools do not equally benefit all patients (particularly patients from populations not used to build these tools), precluding health equity in precision medicine⁸⁸. A potential solution is to move towards GRS that are built upon causal versus tagging variants, which requires extensive functional studies to disentangle linkage disequilibrium in associated regions. As reviewed elsewhere⁸⁹, ongoing research aimed at unveiling the aetiology of rheumatic diseases includes individual molecular profiling, incorporating genetic risk, cell-specific transcriptional signatures, epigenomics, mass spectrometry, tissue and in vitro molecular signatures, and environmental factors. However, the majority of current molecular omics research still focuses on individuals with two copies of chromosome X and of genetic European ancestry⁸⁹, and so the results from this progress are not necessarily translatable or applicable to diverse groups.

The second challenge of applying genetics data to health disparities research is that it must be studied within the context of complex disease aetiology, and not in isolation from influencing environmental and gene–environment factors. Importantly, inherited genetic variants, which are fixed within the genome, do not directly capture alterations imposed by environmental and gene–environment factors. Inclusion of additional omics such as transcriptomics (RNA level), proteomics (protein level) and/or epigenomics (for example, cytosine methylation) offers the potential to capture biological changes caused by stress attributable to systemic racism, diet or other lifestyle factors or exposures that more often affect underprivileged patients. Hence, applying genetics and/or omics knowledge also requires comprehensive and high-quality environmental data to control for environmental confounding, to identify gene–environment associations, and to identify the molecular mechanisms through which environmental exposures affect disease outcomes^90,91. Genes and environmental exposures are inherited together; individuals with more genetic similarity, such as families or groups from the same region, are also exposed to more environmental similarity^90,91. Yet, despite this shared environment, molecular research often negates these factors from studies, contributing to the siloing of biological research. For example, a review of sarcoidosis prevalence emphasized that most family studies focus solely on genetics, without leveraging the opportunity to explore shared environment effects⁹².

Epigenetic studies are well poised to elucidate the relationships between genetics, biology and environmental exposures. Epigenome-wide association studies (EWAS) have identified cell-type-specific DNA methylation changes associated with multiple rheumatic diseases⁷⁸. As reviewed, although DNA methylation varies among population groups, and this variation is partially explained by their distinct genetic ancestry, environmental factors not captured by genetic ancestry are important contributors to variation in DNA methylation. This fact supports the notion that an interaction between environmental, genetic and epigenetic factors underlies health disparities in rheumatic conditions. Mounting evidence suggests that environmental exposures influence epigenetic modifications and trigger alterations in multiple regulatory mechanisms that interact to create a systemic response^93,94. For example, air pollution exposure downregulates hsa-miR-137, which leads to a pro-inflammatory state that contributes to the pathogenesis of RA⁹⁵. In rheumatic diseases, EWAS have largely focused on epigenetic associations with disease, without incorporation of environmental factors linked to determinants of health.

Our review of the literature shows a considerable bias in the number of studies on biological outcomes compared with epidemiological, SDOH research (Supplementary Table 1). On average, the amount of published biological research was 106:1 compared with SDOH research across ten rheumatic diseases (range: 35–271 times). As explored below (in The knowledge gap from research silos), overlap between biological and SDOH research is minimal (<1% for each of the rheumatic diseases explored) (Fig. 2). Despite the accepted influence of environmental factors on cell processes, including the role of stress in inflammation^96,97 and diseases such as SLE, RA and Sjögren’s syndrome⁹⁸, exploration of the biological mechanisms that these exposures enact within the context of health disparities research is limited.

In our review of ten major rheumatic diseases, we identified 71 unique manuscripts that spanned both molecular biology and SDOH topics in the manuscript (Supplementary Table 2). The majority (60.5%) of the 71 manuscripts were classified as reviews, perspectives or protocol papers (Supplementary Fig. 2). Of manuscripts classified as original research, most were genetic studies stratified by race or that adjusted for SDOH in their analytic models. Notably, none reported on the mechanistic links between SDOH and molecular biology outcomes. We recognize that some biologically centred studies that included factors such as income or poverty, but not specific terms such as SDOH or SES, might have been excluded from our search. However, we note that including such factors in the absence of common SDOH or SES identifiers fails to acknowledge their context within a biopsychosocial model and their potential relevance to health disparities. Hence, we limited our intersection search to those that explicitly denoted health disparities, racism, SES and/or SDOH in the manuscript. Altogether, the scarcity of articles identified emphasizes the current knowledge gap of biological–SDOH mechanisms within rheumatic diseases.

After reviews and commentaries, the majority of identified rheumatological research that included biological and epidemiological factures could be classified as genetic risk studies that adjusted for SDOH, primarily using measures of SES. Of note, Cui et al. studied risk prediction for incident SLE among women in the Nurse’s health study cohorts and found that prediction models that included genetics, family history and lifestyle (including income and US geographic region) yielded better performance than models built solely on genetics⁹⁹. This finding highlights the value of incorporating variables beyond biology in prediction models. Other studies focused on assessing genetic risk while adjusting for an SES factor, including single variables (for example, income and education level), social ladders and indices of multiple deprivation. Although multiple deprivation indices and social ladder surveys are more multi-faceted than single variables, we note that these are not without limitations^45,100 and cannot be considered to fully capture the complexity of SDOH. That is, the absence of SDOH effects in a biological model should not be concluded when only a limited measure (such as income) was considered. In our search, we identified several examples in which potential influences of SDOH were downplayed because genetic associations remained after adjustment for single-faceted SDOH variables, further illustrating the siloing of SDOH and molecular research.

The results of this literature search highlight the need for growth in the field. Here, we outline some key considerations and potential solutions towards enabling biopsychosocial research in rheumatic diseases and offering paths towards identifying and alleviating health disparities (Box 2).

Although precision medicine has been promoted for over a decade, a lack of understanding of the effects of determinants of health on health disparities remains a barrier to achieving equitable access to precision medicine²¹. To date, crossover between molecular and SDOH research groups remains limited; this challenge is pervasive across disciplines, but awareness of the problems it poses is growing^12,13. We have illustrated the disciplinary siloing between molecular and epidemiological research by quantifying the current publication gap of integrative research across ten rheumatic diseases. Current work has already shown how inclusion of additional exposures and SDOH can improve biological risk models, but more data are needed to further this area of research. Until we diversify rheumatic disease research, it will remain impossible to implement precision medicine that is informed by both environmental and genetic factors for all patient demographics. Our presented solutions include breaking the epidemiological and biological silos, as well as commitments to equity and structural reform. These changes are necessary to elucidate how exposures affect the heterogenous molecular and cellular pathways leading to pathology, and might highlight novel approaches to avoiding or neutralizing interactions that contribute to rheumatic disease. We note that these challenges are not unique to rheumatic diseases and thus there is benefit in uniting efforts across disease-focused fields (for example, towards standardized SDOH and environmental exposure data, and society-level changes towards better inclusion). The continued development of EHR data, high-throughput omics and molecular profiling offers great opportunities but only if the field works to ensure these data are truly representative of the complex biological and environmental relationships within patient populations.