Abstract
Objective To evaluate the relative prevalence of 8 rheumatic and musculoskeletal diseases (RMDs) across racial and ethnic groups within the National Patient-Centered Clinical Research Network (PCORnet).
Methods Electronic health records from participating PCORnet institutions and systems from January 1, 2013, to December 31, 2018, were used to identify adult patients with ≥ 2 diagnosis codes for rheumatoid arthritis (RA), systemic lupus erythematosus (SLE), osteoporosis (OP), granulomatosis with polyangiitis (GPA), microscopic polyangiitis (MPA), eosinophilic granulomatosis with polyangiitis (EGPA), giant cell arteritis (GCA), and Takayasu arteritis (TAK). Among those with race and ethnicity data available, we compared prevalence of RMDs by race and ethnicity.
Results Data from 28,059,546 patients were available for analysis. RA was more common in patients who were American Indian or Alaska Native vs White, with a prevalence of 11.57 vs 10.11/1000 (odds ratio [OR] 1.15, 95% CI 1.09-1.22). SLE was more common in patients who were Black or African American (6.73/1000), American Indian or Alaska Native (3.82/1000), and Asian (3.39/1000) vs White (2.80/1000; OR 2.43, 95% CI 2.39-2.46; OR 1.39, 95% CI 1.25-1.53; OR 1.26, 95% CI 1.21-1.31, respectively). SLE was more common in patients who were Hispanic vs non-Hispanic (prevalence 3.93 vs 3.45/1000, OR 1.14, 95% CI 1.12-1.16). TAK was more common in patients who were Asian vs White (prevalence 0.05 vs 0.04/1000, OR 1.43, 95% CI 1.00-2.03). OP, RA, and the vasculitides were all more common in patients who were White vs Black or African American.
Conclusion These data provide important information on the prevalence of RMDs by race and ethnicity in the United States. PCORnet can be used as a reliable data source to study RMDs within a large representative population.
Although racial and ethnic demographics within the United States are shifting, the availability of epidemiologic data on rheumatic and musculoskeletal diseases (RMDs) is limited.1,2 Understanding differences in the epidemiology of, and health burden caused by, rheumatoid arthritis (RA), systemic lupus erythematosus (SLE), osteoporosis (OP), granulomatosis with polyangiitis (GPA), microscopic polyangiitis (MPA), eosinophilic granulomatosis with polyangiitis (EGPA), giant cell arteritis (GCA), and Takayasu arteritis (TAK) is important to inform trial recruitment targets, evaluate the generalizability of published research, and provide context for studies of health disparities and resource utilization.
Examination of disease prevalence by race and ethnicity requires defining terms. Race is a socially constructed category determined by the perception of how someone looks, an interpretation that changes by time period and geography. Notably, it is the “embodiment of lived experiences of structural racism.”3 Ethnicity refers to shared culture within a population, manifested in common language, values, norms, or traditions.3 Because race categories are socially constructed, they do not correspond reliably with genetic categories, that is, the genetic origin of a population.3 Nevertheless, race is a critical factor in clinical research because it can be associated with socioeconomic status; access to health services and resources; and the opportunity, inclination, or capacity to participate in human subjects research.
The most reliable prevalence data on RMDs have come from population-based studies.4 Such studies seek to answer research questions for defined populations using administrative or medical records, claims data, surveys, or disease registries as sources,5,6 and results of population-based studies are generalizable to the population in question as a whole.5 Sources used in prior rheumatic disease prevalence data have come from the National Health and Nutrition Examination Survey (NHANES), Medicare data, claims data, and electronic health records (EHRs).7-9 The National Patient-Centered Clinical Research Network (PCORnet) is intended to be a fully integrated research network where vast highly representative health data, research expertise, and patient insights are built-in and accessible to health stakeholders. It was built with the premise that having patients as equal partners in all aspects of the network would revolutionize the way we seek research answers.10 PCORnet represents close to 10% of the population, mainly those seen in academic medical centers.10 Although PCORnet does not encompass the entire US population, it provides health-related data on a large subsection of the population and has the potential to provide reliable prevalence data on RMDs.
In general, existing data on RMDs have focused on White populations, with knowledge gaps existing regarding the effect and prevalence of these conditions on racial and ethnic minority populations within the US.11,12 Ensuring adequate representation in research of racial and ethnic groups requires examining the epidemiology and prevalence of RMDs in these populations. This study sought to (1) evaluate the relative prevalence of each disease across racial and ethnic groups within PCORnet, and (2) compare the racial and ethnic distribution of each of the 8 RMDs under study in PCORnet.
METHODS
Study population/overview. PCORnet currently includes data from 337 hospitals, 338 emergency departments, 1024 community clinics, and over 169,000 physicians.10 PCORnet primarily draws data from EHRs. In this analysis, EHR data from participating PCORnet institutions and systems (Supplementary Table S1, available from the authors upon request) were queried, identifying all adults aged ≥ 18 years with a diagnosis code of any kind between January 1, 2013, to December 31, 2018. Both International Classification of Diseases, 9th revision (ICD-9) and 10th revision (ICD-10) codes were used in the query (Supplementary Table S3); ICD-9 to -10 crosswalks were created using forward-backward mapping procedures using a General Equivalence Mapping approach.13,14 Within this population, we selected patients with diagnosis codes from healthcare providers for RA, SLE, OP, GPA, MPA, EGPA, GCA, or TAK from ≥ 2 encounters separated by at least 90 days, an approach that has been validated in prior rheumatic disease research.15-20 We chose conditions that represented the range of both common and rare RMDs, excluding those, such as spondyloarthritis, that may be more challenging to identify based solely on diagnosis codes. Details on medications, laboratory results, and other testing (eg, t-scores from dual energy x-ray absorptiometry) were not available for this analysis but may be used in future studies with PCORnet to implement alternative algorithms for identifying RMDs, specifically for RA16,17,21,22 and for SLE.23-25
Race and ethnicity. In PCORnet, race and ethnicity data are extracted from EHRs of each institution and mapped to a Common Data Model based on the US Office of Management and Budget standard, which is compatible with data fields in 2015 US census data tables.26 Race for this analysis included American Indian or Alaska Native, Asian, Black or African American, Native Hawaiian or other Pacific Islander, White, or other/missing. Hispanic ethnicity indicated an individual of Cuban, Mexican, Puerto Rican, South or Central American, or other Spanish culture or origin, regardless of reported race.
Analysis. Adult patients with race and ethnicity data were included in this analysis; patients for whom we could not identify race and ethnicity (ie, missing data, > 1 race, other) were excluded. We tested the null hypothesis that there was no association between race and ethnicity and the prevalence of different RMDs within PCORnet. To further assess representativeness, racial and ethnic distributions in PCORnet data were compared to 2015 US census data tables.26 Prevalence of RA, SLE, OP, GPA, MPA, EGPA, GCA, and TAK were then compared across racial and ethnic groups to the overall PCORnet population. Univariable logistic regression was used to assess associations between race and ethnicity with different diagnoses within PCORnet. Two software systems were used for analysis: SAS (version 9.4; SAS Institute) and MedCalc (version 20.110; MedCalc Software). Both systems were needed because raw data were configured as cumulative data by race and ethnicity and diagnosis, not 1 row/person, so MedCalc was needed to convert the data for odds ratios (ORs). The study was not considered human subjects research and was not subject to institutional review board review because only aggregate-level data were analyzed.
RESULTS
A total of 28,059,546 adult patients in PCORnet were seen between January 1, 2013, and December 31, 2018 (Supplementary Table S2, available from the authors upon request). A total of 23,457,477 adult patients with race data and 24,963,746 patients with ethnicity data were included, after excluding 4,602,069 (16%) with other, missing, or > 1 race, or after excluding 3,095,800 (11%) with other or missing ethnicity. Frequency of other, missing, or > 1 race were similar across the rheumatic diseases studied (ranging 11-17%), as was frequency of missing or other ethnicity (ranging 6-11%). The distribution of patients of known race and ethnicity in PCORnet was similar to that in the 2015 US census data, with Black or African American patients somewhat overrepresented in the PCORnet population (18% PCORnet vs 13% US census) and Asian patients somewhat underrepresented (3% vs 6%; Supplementary Table S2). In PCORnet, there were 271,752 patients with ≥ 2 visits at least 90 days apart with a diagnosis of RA (prevalence 9.68/1000), 98,915 with diagnoses for SLE (3.53/1000), 557,269 for OP (19.86/1000), 4379 for GPA (0.16/1000), 1699 for MPA (0.06/1000), 3224 for EGPA (0.11/1000), 8505 for GCA (0.30/1000), and 1029 for TAK (0.04/1000; Table 1).
Prevalence (/1000) of selected RMDsa by race and ethnicity in PCORnet, 2013-2018.
The prevalence of each RMD within each racial and ethnic group is shown in Table 1, and differences in disease prevalence by race and ethnicity from univariable logistic regression models are shown in Table 2. The prevalence of RA was 10.11/1000 in patients who were White and was higher in those who were American Indian or Alaska Native (11.57/1000, OR 1.15, 95% CI 1.09-1.22 vs White) and lower in patients who were Asian (6.29/1000, OR 0.62, 95% CI 0.60-0.64), Black or African American (8.45/1000, OR 0.82, 95% CI 0.81-0.83), or Native Hawaiian or Pacific Islander (4.81/1000, OR 0.49, 95% CI 0.44-0.55). Prevalence of RA was lower in Hispanic vs non-Hispanic patients (8.04 vs 9.95/1000, OR 0.81, 95% CI 0.80-0.82). In contrast, compared to an SLE prevalence of 2.80/1000 in patients who were White, prevalence was higher in patients who were American Indian or Alaska Native (3.82/1000 patients, OR 1.39, 95% CI 1.25-1.53), Asian (3.39/1000, OR 1.26, 95% CI 1.21-1.31), or Black or African American (6.73/1000, OR 2.43, 95% CI 2.39-2.46). SLE prevalence was higher in Hispanic vs non-Hispanic patients (3.93 vs 3.45/1000, OR 1.14, 95% CI 1.12-1.16).
Odds of diagnosis of RMDs among patients in PCORnet, 2013-2018, stratified by race and ethnicity.
OP prevalence was highest in patients who were White (22.48/1000 patients; Table 1 and Table 2). Compared to patients who were White, prevalence was modestly lower in those who were Asian (20.70/1000, OR 0.92, 95% CI 0.91-0.94) and substantially lower in patients who were American Indian or Alaska Native (11.19/1000, OR 0.53, 95% CI 0.50-0.57), Black or African American (8.63/1000, OR 0.38, 95% CI 0.38-0.39), or Native Hawaiian or Pacific Islander (8.77/1000, OR 0.39, 95% CI 0.36-0.43). Prevalence was lower in Hispanic vs non-Hispanic patients (12.24 vs 20.64/1000, OR 0.59, 95% CI 0.59-0.60).
Among the vasculitides, disease prevalence was highest among patients who were White or non-Hispanic for GPA, MPA, EGPA, and GCA (Table 1 and Table 2). In TAK, disease prevalence was higher in patients who were Asian vs White (0.05 vs 0.04/1000, OR 1.43, 95% CI 1.00-2.03). TAK prevalence was lower in patients who were Black or African American and in those who were Native Hawaiian or Pacific Islander vs White and was lower in Hispanic vs non-Hispanic patients.
DISCUSSION
This evaluation of over 28 million adult patients from participating PCORnet institutions, with demographics closely mirroring the overall US population, provides data on the prevalence of key RMDs in different racial and ethnic groups. Despite the potential for underdiagnosis in racial and ethnic minorities, these populations had a significantly higher prevalence of certain conditions vs patients who were White or non-Hispanic. A diagnosis of RA was more common in patients who were American Indian or Alaska Native, and TAK was more common in patients who were Asian. SLE, in particular, disproportionately affected racial and ethnic minorities, with higher prevalence in American Indian/Alaska Native, Asian, Hispanic, and especially Black/African American patients compared to White and non-Hispanic patients. Although further work is needed to advance the understanding of the drivers of racial and ethnic differences in the epidemiology of rheumatic diseases, this study provides information that may help researchers in planning to evaluate the generalizability of published studies, develop recruitment targets to promote representative research, and conduct future evaluations of disparities among patients with rheumatic conditions.
These findings demonstrate the potential of PCORnet as a data source to study RMDs. We found that the population in PCORnet is consistent with the general US population, making this a highly representative data source for research. With PCORnet including approximately 10% of the US population, we were able to identify a large number of patients with both common and rare RMDs. The availability of diagnosis, laboratory, and medication data makes PCORnet a rich data source for future research.
There are many potential drivers of divergent disease prevalence in different racial and ethnic groups. Misdiagnosis or underdiagnosis may result from health system policies, clinic/hospital practices, lack of access to care, physician-patient relationship dynamics, or misconceptions about disease prevalence in different populations.27,28 Differences in socioeconomic status and environmental factors may affect risk of disease onset.29-31 In addition, although race is a social construct and is a poor surrogate for ancestry, especially at the individual level, differences in genetic risk factors across populations may also contribute in some cases.32,33
As of 2015, approximately 1.3 million adults reported having RA in the US.2 In our study, more than 15% of patients with RA were non-White. Rates of RA were modestly lower in patients who were Black or African American, Asian, Native Hawaiian or Pacific Islander, and Hispanic compared to patients who were White and non-Hispanic, although these differences could be due in part to underdiagnosis. Even with the potential for underdiagnosis, rates of RA were highest in patients who were American Indian or Alaska Native, consistent with previous studies, highlighting the importance of RA in this population.34,35 Despite the burden of disease in non-White patients, research in RA continues to focus primarily on White populations, with 1 study in 2013 evaluating 240 randomized controlled trials of RA finding that White patients are often overrepresented in the trials, making up approximately 97% of patients.11,27
Our findings are consistent with previous studies showing that the prevalence of SLE in the US is higher among Asian, African American, African-Caribbeans, and Hispanic Americans in comparison to White Americans.36 An evaluation of 5417 cases of SLE across 4 registries from the Centers for Disease Control National Lupus Registry found that the prevalence of SLE was highest among Black or African American patients.37 Several studies attempted to examine why rates of SLE are particularly high among the Black or African American population. Studies such as Lupus in Minorities: Nature vs Nurture (LUMINA) have identified environmental, socioeconomic, psychosocial, genetic, and clinical risk factors that differ among racial and ethnic groups.28,33
An evaluation of the NHANES from 2005 to 2010 estimated that 10.2 million US adults aged 50 years and older had OP, accounting for approximately 10.3% of the population at the time of the 2010 US census.7 Of these 10.2 million adults, 7.7% were non-Hispanic White, 0.5% were non-Hispanic Black or African American, and 0.6% were Mexican American.7 Prior research finds Black or African American adults are at lower fracture risk than White adults.38,39 Similarly, we found that OP was more common in White and non-Hispanic patients compared to racial and ethnic minorities. Genetic differences in bone size, density and structure, and calcium regulation have been cited as possible explanations for observed differences in risk across racial groups.40-42 Nevertheless, disparities in OP screening and access to information may result in underdiagnosis.43,44
Previous research on the racial and ethnic breakdown of various vasculitides is limited; GPA and MPA can affect those of any race or ethnic background but appear to more commonly affect patients who are White,45,46 as we found in our study. In prior research, GCA appears to be no more prevalent among patients who were White than Black or African American.47 We found a higher prevalence of GCA in White and non-Hispanic patients, but there is the potential for bias leading to underdiagnosis in racial and ethnic minorities. As other studies have shown, we found the highest prevalence of TAK in Asian patients.48,49 Because all these forms of vasculitis are rare, however, these diseases may be unrecognized and thus underdiagnosed.
Despite the large sample size and geographic diversity of the centers and institutions represented in this study, including approximately 10% of the US population, there are limitations to note. We identified patients with diseases of interest based on the presence of 2 or more diagnosis codes separated by at least 90 days. Although this approach may lead to some misclassification, it is similar to methods used in other evaluations of administrative data and has been validated for several of the diseases under study. Further, this approach has been shown to lead to some misclassification, it has been examined in several of the diseases under study and shown to have reasonable accuracy with positive predictive value at around 60% in RA,16 SLE,18 and antineutrophil cytoplasmic antibody–associated vasculitis.19 This misclassification might be expected to make prevalence appear more similar across racial and ethnic groups. As described in the Methods, data on medications, laboratory results, and other testing were not available for this analysis, but some of this information is available from the PCORnet Common Data Model and may be used to implement more specific algorithms for future studies. In addition, limited access to care and other social inequities may result in lower disease prevalence measures in underrepresented groups,50 especially as a majority of data in PCORnet come from academic centers. Our measures of the relative prevalence of RMDs across racial and ethnic groups should be considered conservative. Areas of interest for future research include comparing disease prevalence and outcomes at community centers vs academic centers in PCORnet and examining drivers of racial and ethnic disparities.
It was not possible to assess to what degree differences in disease prevalence by race and ethnicity were driven by socioeconomic and other factors. We excluded patients with missing, other, or more than 1 race, and therefore patients who identify as multiracial were not included.
In conclusion, these findings demonstrate that certain rheumatic diseases, such as SLE, RA, and TAK, disproportionately affect racial and ethnic minority groups in the US, whereas others are modestly less prevalent among patients of racial minorities and Hispanic ethnicity. These findings may help to inform recruitment targets, motivate evaluations of the generalizability of published research, and provide some preliminary context for studies of health disparities and resource utilization including hospitalization, readmission, and surgery rates. Autoimmune disease-specific patient-powered research networks have used prevalence data from PCORnet as benchmarks for the representativeness of their patient registries in inflammatory arthritis and vasculitis.51 Future research should focus on placing such data into appropriate contexts to evaluate and/or address health disparities in adult patients with rheumatic disease.
ACKNOWLEDGMENT
The authors would like to thank Nupur Parikh for her help reviewing the literature to inform the Introduction and Discussion sections of this article.
Footnotes
This work was partially supported by a Patient-Centered Outcomes Research Institute (PCORI) award (CRG-1807-0001) to administer the Autoimmune and Systemic Inflammatory Syndromes Collaborative Research Group (ASIS CRG) of PCORnet. PCORnet has been developed with funding from PCORI. The statements in this article are solely those of the authors and do not necessarily represent the views of organizations participating in, collaborating with, or funding PCORnet or PCORI.
WBN is the Principal Investigator on grants/contracts from AbbVie, Amgen, Janssen, PCORI, and Scipher Medicine, and is an employee of the Global Healthy Living Foundation (GHLF), which receives grants, sponsorships, and contracts from pharmaceutical manufacturers and private foundations. A full list of GHLF funders is publicly available (https://www.ghlf.org/our-partners/). All financial interests of WBN and coauthors that could create a potential conflict of interest (COI) or the appearance of a COI with regard to our work on this manuscript have been mitigated. The remaining authors declare no conflicts of interest relevant to this article.
- Accepted for publication August 2, 2023.
- Copyright © 2023 by the Journal of Rheumatology
