Utilization of electronic health record data to evaluate the association of urban environment with systemic lupus erythematosus symptoms

Dear Editor, The aetiology of SLE is thought to be dependent on the interplay of genetics and environment. The heterogeneity in disease presentation suggests environmental and social factors may contribute. Living in an urban rather than rural environment (‘urbanicity’) has become increasingly associated with non-communicable disease due to an increase in risk factors such as traffic-related pollution [1]. We hypothesize in this study that urbanicity and other measures of social stressors are associated with hallmarks of SLE, which we measure through satisfaction of SLE classification criteria.

We used an electronic health record (EHR) data subset from the Electronic Medical Records and Genomics (eMERGE) network that included patients from Northwestern Medicine, the Kaiser Permanente Washington Health Research Institute, Marshfield Medical Center, Columbia University Irving Medical Center and Mayo Clinic from 2000 to 2020 [2]. The cohort for this study includes eMERGE cohort subjects with addresses geocoded in 2015 [3, 4]. Only patients 18–85 years old were included due to the eMERGE inclusion criteria. Because age of first lupus diagnosis was not available for all patients in the dataset, age of last health care encounter was used as a proxy as there was a strong correlation in our data (R² = 0.81, Supplementary Fig. S1, available at Rheumatology online).

Cases were defined as patients who completely fulfilled the SLICC classification criteria [4, 5]; that is, a patient must meet at least 4 of the 17 classification criteria and at least one clinical and one immunological criterion (Supplementary Table S1, available at Rheumatology online). Controls were patients who did not fulfil the SLICC classification criteria. All procedures involving human subjects were approved by the Institutional Review Board at Northwestern University.

We categorized patients’ residence as urban or rural using Rural Urban Community Area (RUCA) codes, which are based on population density, urbanization and commuting patterns [6]. We categorized race as White, Black or African American, and an aggregate Other Non-White patient group due to sample size. The area deprivation index (ADI) measures average socioeconomic disadvantage experienced by a resident of a specific area. ADI varies continuously from 0 to 1, with larger values reflecting greater deprivation [7]. We performed multivariate logistic regression to determine the association between (i) urbanicity, (ii) self-reported race, and (iii) ADI and satisfying SLICC classification criteria, adjusting for distance from the medical centre, sex and age in the final model. Analyses were performed with R version 4.1.0 (R Foundation for Statistical Computing, Vienna, Austria).

In our cohort, 1122 people fulfilled the SLICC classification criteria (cases) for SLE and 6650 did not (controls) (Supplementary Table S2, available at Rheumatology online). Living in an urban census tract relative to a rural census tract was associated with an increased odds of fulfilling the SLICC criteria of 1.55 (95% CI: 1.31, 1.85). Black or African American and Other Non-White race were both significantly associated with fulfilling the SLICC classification criteria (odds ratio [OR] 2.32 [95% CI: 1.81, 2.97] and OR 5.45 [95% CI: 4.32, 6.87], respectively). Increased ADI was also associated with fulfilling SLICC criteria (OR 1.21 [95% CI: 1.14, 1.29]) (Table 1).

Cases lived further from the participating medical centres, with a raw mean distance from of 120.96 km for cases and 53.39 km for controls. Cases had a greater proportion of patients who were Black or African American (12.6%) and Other Non-White (17.6%) compared with controls (3.9% and 3.0%, respectively) (Supplementary Table S2, available at Rheumatology online).

Our study adds to the existing knowledge of factors associated with SLE symptoms by considering measures of social stress. We find that urbanicity, a measure of a person’s lived environment, is associated with satisfying the SLICC criteria. Urbanicity captures specific exposures that may contribute to SLE symptom development including an abundance of traffic-related particulate pollution and greater stress relative to rural areas, which may exacerbate SLE symptoms [1, 8]. Greater deprivation as measured by ADI is also associated with satisfying SLICC criteria, suggesting that socioeconomic status has a relationship with SLE symptom presentation.

Our project had several limitations. The eMERGE cohort is not a population-representative sample, as there is selection bias towards patients who opt into eMERGE participation. We used SLICC classification criteria because they are clinically relevant and inclusive, but they do not equate to a SLE diagnosis [5]. Controls could have a lupus diagnosis with criteria not present at time of EHR review and we were unable to ascertain the SLICC criterion of biopsy-confirmed lupus nephritis from our data, so our estimate for cases is conservative. The limitations described above likely result in our observation of a lower female to male ratio (∼2:1) than the expected 9:1 female to male ratio of actual SLE diagnosis (Supplementary Table S3, available at Rheumatology online).

These findings provide opportunities for future investigation of environmental elements that may contribute to SLE symptom development. Future work could also explore how systemic barriers may impact patients of marginalized identities. Identifying factors associated with SLE cases may help identify aetiological pathways driving this disease and elucidate avenues for mitigating those factors.

Supplementary data

Supplementary data are available at Rheumatology online.

Data availability

The data used and analysed during the current study are available from the corresponding author on reasonable request. Note that row-level data are access controlled and cannot be provided publicly due to data use restrictions. Algorithms for identification of classification criteria attributes can be found in the Phenotype KnowledgeBase (PheKB), https://phekb.org/.

Funding

This work was supported by the National Human Genome Research Institute through the following grants: U01HG008657 (Kaiser Permanente/University of Washington), U01HG006828 (Cincinnati Children’s Hospital Medical Center), U01HG006379 (Mayo Clinic), U01HG8673 (Northwestern University), U01HG006389 (Marshfield Clinic Research Foundation), U01HG008680 (Columbia University Health Sciences), and U01HG8701 and U01HG006385 (Vanderbilt University Medical Center serving as the Coordinating Center).

Disclosure statement: T.W. receives unrelated research funding from Gilead Sciences. A.K. is an advisor to Datavant. The other authors declare no conflicts of interest.

Acknowledgements

K.F.M. receives salary support from funding awarded to Vanderbilt University Medical Center from GE Healthcare for unrelated work. R.R.G. receives unrelated research funding from Gallagher Research Professor of Rheumatology/Professor of Medicine.

References

Author notes