Abstract
Disparities in the progression of ten chronic overlapping pain conditions (COPC) and four comorbid mental disorders among traditionally underrepresented populations have rarely been reported in the literature. In this study, we conducted retrospective cohort analyses using All of Us Research Program data from 1981 to 2022 to address this issue. Separate cohorts were created to assess the disparities by sex, race, and ethnicity. Logistic regression models, controlling for demographic variables, were employed to identify significant disparity factors associated with the progression from one COPC or mental condition to another. Among 122 frequent disease pairs, we identified demographic disparities in nine progression pathways. For example, we found an increased likelihood of subsequent depression after irritable bowel syndrome in Hispanics versus non-Hispanic individuals (OR:3.13, p<10−10). To our knowledge, this is the first study to unveil demographic disparities in COPC progression. These findings underscore the importance of examining health disparities among individuals affected by multiple chronic pain conditions.
Author Summary Understanding the disparities in chronic overlapping pain conditions (COPC) can aid patients and clinicians in managing and optimizing therapy. Although disparities have been extensively reported for individual COPCs, progression disparity between conditions has been rarely studied. This study addresses this gap by identifying multiple disparities related to sex, race, and ethnicity in COPC progression, providing valuable insights for patients and clinicians to help prevent the exacerbation of other COPCs.
Introduction
Chronic overlapping pain conditions (COPCs), defined as a collection of chronic pain syndromes that often co-occur (1), significantly impact lives worldwide. COPCs can impede a person’s daily activities, mental health, and overall quality of life. In this study, we focus on ten chronic pains, commonly referred to as COPCs in literature: chronic low back pains (CLBP), migraine (MHA), irritable bowel syndrome (IBS), temporomandibular disorders (TMD), fibromyalgia (FM), chronic fatigue syndrome (CFS), chronic tension-type headache (CTTH), endometriosis (ENDO), vulvodynia (VVD), and urologic chronic pelvic pain syndrome (UCPPS). They are closely related to mental health conditions including anxiety, depression, bipolar disorder, and schizophrenia (2).
The experience of pain is not uniform but profoundly shaped by the interplay of demographic factors, including sex, gender, age, race, ethnicity, and socioeconomic status (e.g., income) (3, 4), resulting in disparities known to particularly affect marginalized communities (5). Previous research reported a strong association between demographic background (such as racial minority groups) (3) and pain-related health outcomes (6), as well as evidence indicating that people with socioeconomic challenges also encounter greater obstacles in accessing pain management (7). However, there has been limited work investigating demographic disparities in the progression of COPCs and their comorbid mental health disorders. To address this gap, we studied the demographic disparities in COPCs based on a large-scale longitudinal cohort from the All of Us Research Program (AoURP) (8), which by design, has prioritized underrepresented populations during recruitment.
The primary objective of the research is to discover statistically significant disparity factors associated with one of the ten COPC conditions and four comorbid mental health conditions. We sought to reveal the progression patterns among these COPC and mental conditions, which were of special interest for their well-documented interactions with COPCs (9, 10). The study represents the first of its kind that applied rigorous statistics to identify health disparities and socioeconomic predictors in COPCs, specifically among progression patterns. Our results contribute to understanding the disparate, nuanced clinical experiences in COPCs, advocating for both accessible and individualized chronic pain care that fits each patient’s unique context (7, 11).
Results
Of the 242,835 patients with valid electronic health records in the AoURP v7 dataset, 61,383 (25.3%) had at least one COPC condition. Table 1 shows the demographic information of the COPC cases and their comorbid mental disorders.
Table 2 demonstrates the elevated risk of specific COPCs in females and in non-White races, based on multiple logistic regression studies of these variables using demographically matched cohorts (see Materials and Methods). Significance is defined as a corrected p-value below the 0.05 threshold, adjusted for the socioeconomic factor of income levels. For example, fibromyalgia has an odds ratio of 5.81 in females (p=6.9 ×10−10), indicating that females are five times more likely to be diagnosed with fibromyalgia than males. Similarly, for migraine headaches (MHA), females exhibit an odds ratio of 2.52 (p=2.33×10−11), suggesting females are more than twice as likely to have MHA compared to males. Along the race variable, Black or African American persons have an odds ratio of 1.18 in developing chronic lower back pain (CLBP, p<9.18×10−15) even after adjusting for income levels.
The incidence of mental disorders also showed significant demographic associations. Females present higher odds ratios of 1.39 and 1.33 with adjusted p-values of 1.13 ×10−3 and 1.19 ×10−2 respectively in developing anxiety and depression. For schizophrenia, a significant odds ratio of 2.16 (p<10−10) was observed in Black or African American persons, which corroborates the literature (12, 13).
Fig 1 demonstrates how sex, race, and ethnicity might be associated with the likelihood of developing a second COPC or mental condition following a first for 122 frequent (at least 21 occurrences) pairs of conditions that consist of at least one COPC in the pair. The connections between nodes indicate how strongly these conditions are associated with each other in sequence, by using regression analysis with White, non-Hispanic, or males as the reference group. CLBP exhibits the strongest disparity among the COPC conditions: Black or African American persons with migraine, anxiety, or depressive disorders are more likely to develop CLBP than White persons (shown in black arrows). The odds ratios of developing CLBP following migraine, anxiety, depressive, and bipolar disorders are 1.46, 1.46, 1.38, and 1.26, all statistically significant (adjusted p≤0.005), as compared to White persons. These contribute to the overall enrichment of CLBP in Black or African American persons (OR=1.18, adjusted p<10−10, Table 2).
Each node represents either a specific pain condition or mental disorder. Arrows represent significant trajectories of sequential occurrences of conditions based on reference demographics (white, non-Hispanic, or male). Solid edges with hollow triangle arrows represent further enrichment for specific demographic subgroups that are color-coded: each Black arrow indicates that for the specific directed pair of conditions Black or African American persons have significantly higher risk in comparison to White persons. Likewise, orange arrows for Hispanic or Latino persons as compared to non-Hispanic persons, purple arrows for females compared to males, and dark purple arrows for females vs. males or African Americans vs. Whites. Dashed edges represent significant trajectories among the reference population groups.
The Hispanic population was found more likely to develop anxiety after TMD (OR=3.00, adjusted p=0.017) or IBS (OR=2.42, adjusted p<10−10), and develop depressive disorder after IBS (3.13, adjusted p<10−10), as compared to non-Hispanic populations (shown in orange arrows). The findings indicate certain COPCs and mental health progression patterns are disproportionate across ethnic groups. Lastly, we observed sex disparity in the progression of COPC comorbidities: females were more likely to develop FM (OR=3.00, adjusted p= 0.014), MHA (OR=1.86, adjusted p=0.013), and CLBP (OR=1.50, adjusted p=0.011) after depressive disorders (shown in dark purple arrows). Of note, income level was adjusted during the examination of disparity, and the effect was concordant to the literature.
Specifically, we found lower income is associated with an elevated risk of subsequently developing another COPC or mental disorder, as compared to the highest income group ($200,000 USD annually), with an OR≥1.09 and an adjusted p<0.05, except for a few cases that were underpowered.
Materials and Methods
We used the AoURP(8) version 7, which involved the 1981-2022 electronic health records (EHR) of over 250,000 participants. Only individuals sharing EHR data were included, and those whose data showed apparent quality issues were also excluded. Variables studied include incidence of each COPC and mental condition, sex at birth, age, ethnicity, race, gender, and income level. Then, statistical differences were derived from the comparison between an examined level of a disparity factor and its reference level (e.g., female sex at birth versus male) with respect to a COPC/mental condition outcome. The research protocol received a waiver from the Institutional Review Board (IRB) of the University of Arizona.
The determination of COPC/mental cases is as follows:
Identifying episodes of COPC and mental conditions: We cataloged COPC and mental health conditions using the ICD-10 (14) and ICD-9 (1) codes recognized by the research community. Utilizing terms from the literature and descriptions, we linked the conditions to the corresponding standard concept IDs in AoURP. These codes and concept IDs formed the criteria for identifying potential cases.
Establishing chronicity: A condition was deemed chronic if it was diagnosed on at least two separate episodes, a minimum of 90 days apart, aligning with standards accepted by the pain research community.
Determining the incidence, or earliest onset: To establish each patient’s order of distinct conditions, we identified the earliest onset or incidence for each condition based on diagnosis dates. Additionally, we mandated that the onset of each condition of a patient occurs at least one year after his/her first EHR record, to minimize bias from incomplete historical data.
The conditions were first analyzed for their disproportionate effect on underrepresented groups, then for elevated risk to progress to other conditions, using reference groups for comparison. We built a separate cohort for each disparity factor’s incidence of developing a COPC/mental disease or progressing to another COPC. The process of cohort selection for each analysis proceeded as follows:
Building a cohort for each disparity level in developing a COPC/mental condition: Each level (e.g., African American) of a disparity factor (e.g., race) was compared to a reference level (e.g., White). We constructed a cohort of participants based on the studied or the reference level, choosing the smaller level whenever applicable. The same number of participants were selected for each of the two levels, further matched by other demographic variables, such as sex at birth, birth year (categorized in decade intervals), and race. We conducted random sampling without replacement from the factor group that had more samples (usually the reference group, e.g., White), ensuring an equivalent number of participants from the larger group for each demographic combination. Infrequently, when there were not enough samples to achieve a complete demographic match for a particular demographic feature combination, we opted for partial matching. The hierarchy for matching was sex, followed by birth year group, and then race.
Building a cohort for each disease progression with at least 21 occurrences (per AoU policy): Starting from any COPC/mental condition, we divided the participants with EHR into two groups, those with and without the start condition. We then matched the patients with the starting COPC/mental condition from those without the condition, by the sex, age interval, and race, similar to the procedure above. Then, we examined the difference in the incidence of other conditions after the start of the COPC/mental condition between the two groups of participants. We determined the disparity effect of demographic factors as compared to their reference level concerning this specific cohort of disease progression.
After building the cohort for each study, we evaluated the elevated risk of developing the first or progressing to another condition. Each COPC/mental condition other than the start condition served as an outcome variable for the cohort. We applied a logistic regression model to assess the statistical significance of each disparity factor, specifically the odds ratio of developing or progressing to the outcome condition, relative to the reference level. An odds ratio exceeding 1 with statistical significance suggests a disparity in the condition. We also included a socioeconomic factor in the model: income level. In this analysis, birth year was treated as numerical data, whereas sex, race, ethnicity, and income level were handled as categorical variables. To account for the possibility of multiple comparison errors, we applied the Bonferroni correction. This approach is thorough and robust, leveraging longitudinal data to recapture pairwise chronic pain condition progression and demonstrate disparities.
Discussion
Our results recaptured previous evidence (15) on the disparity of COPC incidence, which may be attributed to subconscious biases within the healthcare system (16). Moreover, the study made a unique contribution by elucidating disparities linked to specific progression patterns of these conditions (17). For example, an increased risk for Hispanic populations to develop anxiety or depression after IBS. While demographic disparities in developing individual COPC conditions have been frequently reported, studies investigating disparities in their progression are rare. Not only that multiple COPCs are associated with more complex outcomes (18), but the nuanced progression disparities offer critical information on potentially optimizing chronic pain management and cross-specialty coordination.
Not all observed progressions were statistically significant across demographic variables (e.g., sex) for some COPCs. This outcome is likely due to the underpowered nature of certain cohort groups, limiting the ability to detect subtle differences in COPC progression. The limitation in statistical power may stem from the sample sizes for progression between some COPC pairs (19) (e.g., between CTTH and TMD) or in certain strata, particularly when the data is segmented by multiple dimensions such as age, sex, race, and socioeconomic status. Despite these limitations, the observed trends offer valuable insights into the complex interplay among these factors in shaping chronic pain manifestations.
Future research should be able to build on the discovered correlations, exploring the causes and effects of these correlations. Our progression analysis represents a promising approach to health disparity research, especially as more data becomes available. For example, there is limited data for studying chronic pain disparities experienced by intersex persons (20). Such research is crucial for developing effective pain management strategies to address the needs of diverse populations (21).
In conclusion, research on progressions in health disparities in individuals with chronic overlapping pain conditions (COPCs) remains limited. By leveraging rich longitudinal data in the All of Us Research Program, we conducted a novel study to evaluate the disparate risk across demographic groups for their progression patterns in developing one COPC or mental condition after another. Our results not only reinforced existing evidence, but also uncovered lesser-known associations, such as the higher risk of progressing from IBS to depressive disorder in Hispanic populations (OR:3.13, adjusted p<10−10). By recognizing the confluence of demographic and socioeconomic factors, we anticipate such data-driven research will inform more contextualized strategies in caring for our diverse populations affected by chronic pains.
Data Availability
The original data can be accessed from All of Us Research Workbench, after registered as a control tier user.
Acknowledgements
This work is funded by NIDCR/NIH R21DE031424. We gratefully acknowledge All of Us participants for their contributions, without whom this research would not have been possible. We also thank the National Institutes of Health’s All of Us Research Program for making available the participant data examined in this study.
References
