Abstract
Objective Radiographic axial spondyloarthritis (r-axSpA) has a 7-year average diagnostic delay. Although the effects of sex or gender on time to diagnosis have been evaluated, the role of social determinants of health remains understudied. We assessed whether time from initial clinical documentation of r-axSpA symptoms to r-axSpA diagnosis (diagnostic delay) varies based on sex, race, ethnicity, and/or the presence of social needs.
Methods We studied patients with r-axSpA from a tertiary center from 2000 to 2022. The cohort was built with the Observational Health Data Sciences and Informatics (OHDSI) network. For the primary analysis, we assessed the time from back pain and/or spinal pain to r-axSpA diagnosis and, secondarily, the time to r-axSpA from any other r-axSpA–related condition. To estimate differences in diagnostic delay, we employed an accelerated failure time parametric survival model.
Results We included 404 patients (mean age 49 years; 38.6% female), with 25.5% identifying as Black, 31.1% as other or unknown race, and 14.1% as Hispanic. Patients with a documented social need had a 21% increase in time from back pain to r-axSpA diagnosis (95% CI 0.93-1.56). In patients with any r-axSpA–related condition, time to diagnosis similarly increased by 21% (95% CI 0.92-1.57). Considering that there is an average time to diagnosis of 34 months, a social need increased time to diagnosis by 7 months.
Conclusion This study reveals a trend toward diagnostic delay in r-axSpA related to social need, sex, race, and ethnicity. Future studies should focus on referral strategies to enable prompt diagnosis and optimize care.
Radiographic axial spondyloarthritis (r-axSpA; formerly known as ankylosing spondylitis) is a chronic, systemic, immune-mediated condition characterized by inflammation involving the axial skeleton, peripheral joints, entheses, and extramusculoskeletal sites.1-4 Individuals with r-axSpA may initially present with nonspecific back pain and may not undergo appropriate rheumatologic evaluation, thereby causing a delay in diagnosis averaging 7 years from symptom onset.5-10 In some cases, people with r-axSpA may initially experience inflammation outside of the spine, such as inflammation of the eye (uveitis), skin (psoriasis [PsO]), bowel (inflammatory bowel disease [IBD]), or peripheral arthritis/tendonitis, making a prompt r-axSpA diagnosis more challenging.
Despite significant advances in management over the last 2 decades,11-13 with the use of therapies including tumor necrosis factor inhibitors (TNFi), interleukin 17 inhibitors (IL-17i), and Janus Kinase inhibitors (JAKi), longer time from symptom onset to treatment portends poor response to therapy.14-17 Further, patients with active r-axSpA experience reduced quality of life and functional limitations if their disease is not controlled.18-20
The reasons for this delay are not well understood, primarily due to the complexities of disease manifestations, systemic barriers such as healthcare access, and healthcare provider knowledge and practice. For example, studies performed on race and ethnicity in r-axSpA21-23 have identified that Black and Hispanic/Latinx patients tend to exhibit more functional impairment, worse radiographic severity, and higher inflammatory markers compared to White patients. Although there could be a biological reason for this delay, these populations have also been underrepresented in clinical research, creating a bias that may lead to milder cases in these populations going undiagnosed.24 Moreover, the presence of a social need on r-axSpA diagnostic delay is underexplored, further complicating our efforts to gain a comprehensive understanding of this delay. By investigating these factors, we can enhance our understanding of diagnostic delays and develop strategies to promote early referral and diagnosis, and equitable care and outcomes for individuals with this condition.21-23,25
In this study, using data from patients receiving care at a tertiary care center, we assessed whether the time from the initial clinical documentation of r-axSpA symptoms to r-axSpA diagnosis varied based on sex, race, ethnicity, and/or social needs, and whether documented presence of a social need was associated with a longer time to diagnosis. To optimize cohort identification and data collection using a standardizing coding system, we used the Observational Health Data Sciences and Informatics (OHDSI) platform.26
METHODS
Data source. We performed an observational study using electronic health record (EHR) data from a large urban safety-net hospital from 2000 to 2022 that serves a diverse population, with > 70% of patients identifying as people of color and living in lower-income communities. Approximately 50% of the patient population is insured by Medicaid through Massachusetts Medicaid (MassHealth) accountable care organizations. One-third of the patients use a primary language other than English.27-31
Data collection. To facilitate data collection, we used the platform by OHDSI, a collaborative global initiative comprising researchers, data scientists, informaticians, healthcare providers, and regulators from across the world with the aim to conduct rigorous observational research through the Observational Medical Outcomes Partnership (OMOP) Common Data Model (CDM) toolset.32-35 The OMOP CDM ensures uniformity and transparency in data collection, providing a homogeneous coding system for an organized and structured healthcare data analysis. This framework is applicable to analysis of a regionally defined population and enables the conduct of large-scale, multisite studies across different institutions and organizations.
ATLAS is an application developed by the OHDSI community for defining cohorts, performing cohort studies, and analyzing real-world patient data within the OMOP CDM framework.36 In this study, practicing rheumatologists (MD and JWL) identified relevant features and criteria (eg, diagnoses, medications, visits, measurements) to define covariates. We built concept sets by associating these features with codes from standard diagnostic ontologies through ATLAS and, when necessary, mapped nonstandard concepts. Using these concept sets, we constructed cohort definitions. To ensure accuracy, we tested the cohort definition by verifying counts and assessing sample cohort members, confirming that the inclusion and exclusion criteria accurately represented the patient population. Figure 1 summarizes the process of code selection to develop concept sets, define cohorts, and extract data using OHDSI tools to perform analyses.
Flow diagram for a study using the OMOP common data model. Healthcare systems (top row) have unique data structures that can be mapped onto the OMOP common data model (dashed arrows). Analysis inputs are defined using concept sets derived from standard ontologies (eg, International Classification of Diseases, Systematized Nomenclature of Medicine) to generate the analytic data set. OMOP-structured data sets from multiple institutions can be pooled and analyzed using standard methods (solid arrows). OMOP: Observational Medical Outcomes Partnership.
Study population. Eligible patients were adults with r-axSpA defined using ≥ 1 International Classification of Diseases, 9th revision (ICD-9) or ICD-10 codes based in an OMOP concept set that included specific codes 720.0 and M45, as well as the nonspecific ICD-9 code 720.x (Supplementary Table S1, available from the authors upon request). For our inclusion criteria, patients were required to have ≥ 1 outpatient visit ≥ 12 months prior to the date of r-axSpA diagnosis, ensuring a period in which r-axSpA–associated conditions could be assessed between 2000 and 2022. These dates represent the availability of data capture within the EHR database. Random samples of patients included in the r-axSpA cohort definition were evaluated by 2 independent reviewers (MD and JWL) for validation purposes. No other formal validation process was done for this study. We excluded those whose initial visit provided evidence that they had already received an r-axSpA diagnosis, including treatment with biologic agents and/or a note of the diagnosis at their initial visit.
Outcome of interest. The main outcome of interest was the time to diagnosis, calculated as the time from back/spinal pain diagnosis or other first documented r-axSpA–associated condition to the date of r-axSpA diagnosis. For the primary analysis, we assessed the time from back pain/spinal pain to r-axSpA diagnosis. For the secondary analysis, we examined the time from either back/spine pain, peripheral arthritis, PsO, IBD, or anterior uveitis to r-axSpA diagnosis (Supplementary Tables S2-S7, available from the authors upon request).
Covariates. We extracted data on age at axSpA diagnosis; sex; patient-reported race; patient-reported ethnicity; use of nonsteroidal antiinflammatory drugs, disease-modifying antirheumatic drugs, TNFi, steroids, and JAKi; inflammatory markers; HLA-B27; axial imaging studies; referrals to physical therapy and pain-management programs; procedures; visits to primary care, rheumatology, and emergent or urgent care; and the documentation of any social need from the THRIVE-validated social determinants of health (SDOH) screening questionnaire implemented by Buitron de la Vega et al.27 The THRIVE questionnaire is administered to all patients at primary care visits and consists of questions addressing housing stability, food security, transportation access, medication affordability, employment status, education status, and caregiving needs. The presence of any reported social need on the THRIVE questionnaire was coded as a binary variable. For example, for housing insecurity, the question was “Do you currently live in a shelter or have no steady place to sleep at night?” with the choices being yes/no. Individuals with no documented THRIVE questionnaire were excluded from the primary analysis.
Analytic methods. Data were first characterized descriptively, and age, sex, race, ethnicity, and the presence of a social need were defined as main exposure variables. Histograms were generated to better visualize diagnostic time between subgroups by age, sex, race, ethnicity, presence of a social need, and year of r-axSpA diagnosis. The primary analysis used the accelerated failure time (AFT) parametric survival model to estimate the average time to r-axSpA diagnosis, with adjustment for age, sex, race, and ethnicity as confounders. Unlike Cox proportional hazards models, an AFT model directly estimates the effect of exposures on the time to diagnosis.37 We fit 4 models using different distributions (generalized gamma, Weibull, exponential, and log normal). We then used the Akaike information criterion (AIC) and −2 log likelihood test to compare model fit. In a sensitivity analysis, we imputed missing social need information as “no documented unmet social needs.”
RESULTS
We initially identified 761 patients with r-axSpA seen between 2000 and 2022 using our cohort definition. After exclusion of individuals without documented back pain or an r-axSpA–associated condition prior to r-axSpA diagnosis, we were able to include 404 patients with r-axSpA in our analyses (mean age 49 years; 38.6% female; Table). Of these individuals, 25.5% identified as Black, 43.4% as White, and 31.1% had another or unknown racial designation; 14.1% identified as Hispanic/Latinx ethnicity. Of the 404 individuals with r-axSpA, 61 (34.1%) documented a presence of a social need (mean age 49.5 years; 21% female). Of these, 36% identified as Black and 32% as other or unknown race.
Demographics and baseline characteristics of patients with r-axSpA who had previous documentation of any clinically defined r-axSpA symptom.
Longer time to diagnosis was seen in those who were > 45 years old compared to those who were younger; in women vs men; in Black and Hispanic/Latinx patients vs White patients; and in those with documented social needs vs those without (Figure 2 and Figure 3).
Histograms of proportion of individuals (Y-axis) by time to r-axSpA diagnosis from back pain (X-axis), stratified by (A) ethnicity, (B) race, (C) sex, (D) age, and (E) presence of a social need. NA: not applicable; r-axSpA: radiographic axial spondyloarthritis.
Histograms of proportion of individuals (Y-axis) by time to r-axSpA diagnosis from other previously documented r-axSpA–associated conditions (X-axis), stratified by (A) ethnicity, (B) race, (C) sex, (D) age, and (E) presence of a social need. NA: not applicable; r-axSpA: radiographic axial spondyloarthritis.
For the AFT models, using the appropriate comparison criteria for nested (−2 log likelihood) and nonnested (AIC) models, we selected the gamma distribution model as having the best fit for the time to diagnosis. Patients with a documented social need had a 21% increase in time from back pain to r-axSpA diagnosis (95% CI 0.93-1.56; Figure 4), adjusting for sex, race, and ethnicity. In the model for any r-axSpA–related condition, time to diagnosis increased similarly by 21% (95% CI 0.92-1.57). Given that mean time to r-axSpA diagnosis was 34 months among those without social need, these findings indicate that a social need increased the time to r-axSpA diagnosis by 7 months, after accounting for age, sex, race, and ethnicity.
Influence of age, sex, race, ethnicity, and presence of social needs on mean time to r-axSpA diagnosis. The dotted line at 1.0 indicates that there is no difference in the average time it takes for the diagnosis to occur, with an increase in the value of the risk factor. r-axSpA: radiographic axial spondyloarthritis.
DISCUSSION
In this study, we examined how sex, race, ethnicity, and presence of social needs influenced diagnostic delay in r-axSpA. We found that women (vs men), Black (vs White) patients, and Hispanic/Latinx (vs non-Hispanic/Latinx) patients tended to experience longer diagnostic delays. Further, we observed a worrisome delay in time to diagnosis among patients with documented social needs. This study highlights the relevance of addressing healthcare disparities in the assessment of patients with r-axSpA. To conduct this study, we innovatively employed the collaborative framework provided by the OHDSI community to identify an r-axSpA cohort and extract EHR data using standardized coding schema. This approach allowed us to use advanced tools to build not only a precise cohort, but also a transparent, reproducible, and scalable study design. With additional infrastructure, the analyses used in our single-center study could also facilitate future multicenter studies addressing disparities in our field.
In our study, women with r-axSpA had longer times to diagnosis compared to men, consistent with findings from other studies.3,38,39 Although women tend to have more severe manifestations of SpA, it is not clear if this is related to biological sex differences or from delayed diagnosis in some individuals. Observational studies of axSpA have demonstrated that women tend to exhibit more widespread pain and extramusculoskeletal manifestations at diagnosis than men.3,38,39 Imaging findings like radiographic sacroiliitis and syndesmophytes occur less frequently in women, leading to cases being missed or misdiagnosed, and magnetic resonance imaging scans of the sacroiliac joints may show bone marrow edema up to 1 year after pregnancy,32 increasing the possibility of a delay in diagnosis in women tested during that time. These clinical differences in r-axSpA between women and men are not only solely biological and also carry important social implications. The delay in diagnosing r-axSpA in women raises questions about healthcare access, treatment within the healthcare system, and missed therapeutic opportunities, all which merit further investigation.40 Our findings highlight the need for more sex-oriented data on r-axSpA to improve equality in outcomes and quality of life for men and women.
Patients from minoritized racial backgrounds and patients of Hispanic/Latinx ethnicity experienced longer delays before r-axSpA diagnosis. Prior epidemiologic studies have suggested a lower prevalence of r-axSpA and HLA-B27 positivity among Black individuals.15,21-25,39,41,42 Notably, such prevalence data should be considered in the context of the rigor of these initial studies and the representativeness of data collected. In other studies of axSpA,21,43 Black patients have exhibited more severe manifestations, both subjectively (as measured by disease activity scores) and objectively (as assessed by laboratory variables and imaging features) compared to Hispanic/Latinx and White patients. Although these studies have hypothesized a difference in severity of axSpA by racial and ethnic group, it is plausible that these differences are related to underdiagnosis of axSpA in non-White and non-Hispanic individuals. These disparities may be linked to the possibility that race has been heterogeneously reported in rheumatologic clinical guidelines,44 creating implicit bias when it comes to diagnosing diseases that have been traditionally perceived to affect certain races, further leading to diagnostic delay and increased morbidity.44,45
This study employed a standardized questionnaire to assess social needs, such as housing stability, food security, transportation access, medication affordability, employment status, education status, and caregiving needs,25 in an r-axSpA population receiving care at our institution. We found that patients with ≥ 1 social need experienced a longer delay in r-axSpA diagnosis compared to those without any documented social need. Previous studies from European populations,8,46-49 as well as studies from China, Korea, and Taiwan,1,18 have analyzed how factors such as lower educational attainment, lower income, unemployment, difficulty remaining in current work, and low perceived social support increase the burden of disease in individuals with r-axSpA. Also, aspects like food security, housing stability, and transportation have been sparsely described in these populations, further emphasizing the growing need for a standardized screening tool for social determinants of health.9,27,50 The presence of a social need in the context of a chronic illness depicts a troubling scenario, as individuals with uncontrolled axSpA may ultimately experience reduced mobility, reduced productivity,51-53 and increased burden on their families, leading to higher costs for the healthcare system. Given the diverse population in the US and the heterogeneity of health insurance coverage,45 there is an urgent call for increased support for these vulnerable patients to ensure equitable access to health care. Education is a crucial component in addressing diagnostic delays. Efforts should first focus on primary care providers and other clinicians who commonly manage chronic back pain in adults by training them to recognize early symptoms and extraspinal manifestations of r-axSpA.54-56 Moreover, medical schools and residency training programs should continue to tailor their curricula to help medical students and physicians in training develop a standardized approach to patients’ social determinants of health.57 By doing this, we can incorporate education on implicit biases and mitigate these disparities, fostering a more standardized approach to diagnosis that minimizes the influence of racial, ethnic, and gender biases. Additionally, clinicians who manage adults with chronic back pain can apply the recently developed referral recommendations, which highlight the individual characteristics that increase the probability of axSpA.58
Limitations of this study should be acknowledged. Due to the use of data from a single institution, we had a relatively small sample size that limited the precision of our effect estimates and limits generalizability to other populations. Patients were also required to be within the EHR system for a minimum of 1 year to be included in the analysis, potentially excluding those who had their diagnoses before the 1-year inclusion criteria. Despite using OHDSI to standardize cohorts and reduce errors in our data, we were unable to perform formal validation, and we were limited by the use of an EHR database rather than a dataset that includes more detailed characterization of axSpA, thus allowing validation. The way our study was designed did not allow us to explore previous duration of symptoms or previous visits to other healthcare systems. Although our cohort was representative of the source population, we were limited in race and ethnicity reporting in our EHR system; thus, 31.1% of the included patients were categorized as other or unknown race, hindering our ability to explore these underrepresented groups further. Whereas our study aimed to explore SDOH as 1 of the reasons for diagnostic delay, other complex challenges, including potential misdiagnoses, missed appointments, delays in obtaining certain tests, incorrect referrals before arriving at the rheumatology service, or concomitant visits to another healthcare system, could not be assessed owing to the limitations of the EHR system. We were unable to assess the effect of individual-level (eg, related to healthcare providers) and system-level (eg, related to healthcare access) factors associated with diagnostic delay in our study. Additionally, our results do not reflect the actual total length of time before r-axSpA diagnosis since we were unable to assess the total duration of patients’ symptoms prior to seeking care. Missing data may have also influenced the accuracy of our findings, including race and ethnicity information and responses to the THRIVE questionnaire, potentially missing the true extent of social disparities in the study population. There is no current consensus for systematic extraction of social needs information in patients with rheumatologic conditions9 and this may account for potential unmeasured social needs.
The strengths of our study include our ability to leverage OHDSI tools and the OMOP CDM to analyze a large, standardized dataset from an EHR database spanning over 2 decades, enabling a detailed analysis of diagnostic delays in r-axSpA within a diverse patient population. We also included data from a validated questionnaire, THRIVE, which allowed for a more comprehensive assessment of patients’ social needs, as these are often undercoded in other structural EHR data.
In conclusion, our findings of longer time to axSpA diagnosis among those with a documented social need indicates an ongoing need to develop strategies for prompt recognition and referral of patients presenting with suggestive symptoms of r-axSpA, particularly in underrepresented populations in which diagnostic delays have historically existed. Accounting for social disparities as one of the reasons for a diagnostic delay in r-axSpA only adds a layer of complexity when measuring the total time between symptom onset and definitive diagnosis, potentially indicating that some patients may experience an even longer delay to diagnosis than that described in the literature. This work adds to the growing body of research addressing social disparities in r-axSpA and underscores the necessity of more inclusiveness in clinical research to be representative of the population of patients with axSpA. The use of tools like OHDSI and OMOP that allow access to data across healthcare systems has the potential to enhance the study of disparities and promote equity in expedient access to health care, diagnosis, therapy, and, ultimately, axSpA outcomes.
Footnotes
CONTRIBUTIONS
RFE: writing – original draft, writing – review & editing, visualization; GR: data curation, investigation, formal analysis; SG: project administration, methodology, data curation; BM: data curation, formal analysis, visualization; RS: review & editing; SRJ: formal analysis, software, validation; JWL: writing – review & editing, supervision, conceptualization; MD: conceptualization, supervision, methodology, project administration.
FUNDING
This publication was supported by the National Center for Advancing Translational Sciences, National Institutes of Health (NIH), through Boston University Clinical & Translational science Institute (BU-CTSI) grant no. 1UL1TR001430. This work was supported by the Arthritis Foundation and NIH through grant no. P30 AR072571. JWL was supported by the Rheumatology Research Foundation Investigator Award, a Spondyloarthritis Research and Treatment Network pilot grant, and the Spondylitis Association of American Jane Bruckel Early Career Investigator Award. MD was supported by Rheumatology Research Foundation and the BU-CTSI. BM was supported by the NIH (grant no. T32 GM140972). The contents of this work are solely the responsibility of the authors and do not necessarily represent the official views of the NIH.
COMPETING INTERESTS
The authors declare no conflicts of interest relevant to this article.
ETHICS AND PATIENT CONSENT
This project received approval from the Institutional Review Board at Boston University Medical Center. The ethics approval number is H-42725. Patient consent was not required for this study as it used deidentified data from our database.
- Accepted for publication October 21, 2024.
- Copyright © 2025 by the Journal of Rheumatology
