key: cord-0818206-mt4ey0xq
authors: Lynch, Kristine E; Shipherd, Jillian C; Gatsby, Elise; Viernes, Benjamin; DuVall, Scott L; Blosnich, John R
title: Sexual orientation-related disparities in health conditions that elevate COVID-19 severity
date: 2021-11-14
journal: Ann Epidemiol
DOI: 10.1016/j.annepidem.2021.11.006
sha: 195a16ce31b9d55023aae7badf956f97a8699b3c
doc_id: 818206
cord_uid: mt4ey0xq

PURPOSE: The Veterans Health Administration (VA) is the largest single integrated healthcare system in the US and is likely the largest healthcare provider for people with minoritized sexual orientations (e.g., gay, lesbian, bisexual). The purpose of this study was to use electronic health record (EHR) data to replicate self-reported survey findings from the general US population and assess whether sexual orientation is associated with diagnosed physical health conditions that may elevate risk of COVID-19 severity among veterans who utilize the VA. METHODS: A retrospective analysis of VA EHR data from 10/01/1999-07-01-2019 analyzed in 2021. Veterans with minoritized sexual orientations were included if they had documentation of a minoritized sexual orientation within clinical notes identified via natural language processing. Veterans without minoritized sexual orientation documentation comprised the comparison group. Adjusted prevalence and prevalence ratios (aPR) were calculated overall and by race/ethnicity while accounting for differences in distributions of sex assigned at birth, age, calendar year of first VA visit, volumes of healthcare utilization, and VA priority group. RESULTS: Data from 108,401 veterans with minoritized sexual orientation and 6,511,698 controls were analyzed. After adjustment, veterans with minoritized sexual orientations had a statistically significant elevated prevalence of 10 of the 11 conditions. Amongst the highest disparities observed were COPD (aPR:1.24 [95% confidence interval:1.23-1.26]), asthma (1.22 [1.20-1.24]), and stroke (1.26 [1.24-1.28]). CONCLUSIONS: Findings largely corroborated patterns among the general US population. Further research is needed to determine if these disparities translate to poorer COVID-19 outcomes for individuals with minoritized sexual orientation.

A concordance of evidence about sexual orientation-related health disparities has developed in the decade following the 2011 National Academy of Medicine report about the health of lesbian, gay, bisexual, and transgender (LGBT) populations. 1 Health risk behaviors (e.g., smoking, 2 substance use 3 ) and social stressors (e.g., violence, 4 homelessness 5 ) are well-documented amongst key health determinants for people who have minoritized sexual orientations. However, disparities in health outcomes such as diagnosed medical conditions are still developing in the literature due to lags in availability of sexual orientation fields in electronic health record (EHR) data. 6 Differences in diagnosed physical health conditions for individuals with minoritized sexual orientations have gained increased urgency in the wake of the coronavirus disease (COVID- 19) pandemic. 7, 8 Several diagnosed medical conditions increase severity of COVID-19 symptoms, vulnerability to complications, and mortality. 9, 10 For example, hospitalizations are three times higher for those with type one or type two diabetes 11 and intensive care admissions are similarly higher among those with clinically active asthma. 12 Although not yet wholly understood, routinely collected healthcare data have enabled research on the impact of preexisting conditions on COVID-19 among certain marginalized populations, such as racial and ethnic minority groups. 13, 14 Similar research is non-existent for individuals with minoritized sexual orientations owing to the lack of data on sexual orientation in healthcare environments, including COVID-19 tracking systems. 8, 15, 16 Heretofore, most knowledge about physical health profiles of minoritized sexual orientation groups typically relies on self-reported survey data. For example, individuals with minoritized sexual orientations are more likely to report acute respiratory illnesses and asthma than their heterosexual peers. [17] [18] [19] [20] Data from the National Health and Nutrition Examination Survey indicated greater risk profiles for cardiovascular disease among people with minoritized sexual orientations, especially women. 21, 22 Most recently, Heslin and Hall used data from the Centers for Disease Control and Prevention Behavioral Risk Factor Surveillance System (BRFSS), finding consistent sexual orientation differences in ten physical health conditions that may worsen COVID-19 severity and increase risk of death: asthma, cancer, heart disease, chronic obstructive pulmonary disease (COPD), diabetes, hypertension, kidney disease, stroke, obesity, and smoking. Although self-reported survey data, such as BRFSS, advanced the field on health outcomes among the sexual minority community, some evidence is equivocal. For instance, a study of the Nurses' Health Study detected greater prevalence of type 2 diabetes among women with minoritized sexual orientations than heterosexual women, 23 other studies of survey data failed to detect sexual orientation-related differences in self-reported diagnoses of diabetes. [24] [25] [26] Additionally, questions remain about the accuracy of estimating disease prevalence given differing levels of agreement between self-report and medical record data. [27] [28] [29] Replication studies using routinely collected EHR data to determine objective provider diagnosis can clarify disparities, specifically for health outcomes. The purpose of this study was to replicate the recent Morbidity and Mortality Weekly Report 7 about sexual orientation differences in COVID-19-related medical comorbidities using data from the Veterans Health Administration (VA). We hypothesized that disparities in diagnosed physical health conditions are significantly influenced by sexual orientation.

We used EHR data from VA that had been transformed into the Observational Medical Outcomes Partnership (OMOP) common data model. 30 The VA OMOP data are sourced from the VA Corporate Data Warehouse (CDW), 31 a data repository that receives nightly updates from VA's EHR platform Veterans Health Information Systems and Technology Infrastructure. The CDW contains clinical data from inpatient, outpatient, and emergency settings as well as services provided externally but paid for by VA.

The study cohort for the present analysis considered veterans who enrolled in VA after fiscal year 1999 and had documentation in their EHR indicating a minoritized sexual orientation at any point in time between October 1, 1999 and July 1, 2019; the former date marks the beginning of the VA EHR and the latter the date of data extraction. Data were analyzed in 2021.

Documentation in this context consisted of records in both unstructured clinical notes, extracted via natural language processing (NLP), and, to a much lesser extent, structured administrative coding indicating a minoritized sexual orientation (e.g., ICD10 code Z72.52: high risk homosexual behavior). Details of the NLP extraction process are described elsewhere. 32,33 A comparison group of veterans without minoritized sexual orientation documentation who enrolled in VA between October 1, 1999 and July 1, 2019 was selected as a proxy for heterosexual veterans (hereinafter referred to as veterans with non-minoritized sexual orientation). This study was reviewed and approved by the University of Utah Institutional Review Board.

All physical health conditions presented by Heslin and Hall 7 were analyzed in the present study including asthma, cancer (excluding skin cancer and benign tumors), heart disease, COPD, type 1 diabetes, type 2 diabetes, hypertension, chronic kidney disease (CKD), obesity, smoking, and stroke. With the exception of obesity and smoking, each condition was defined using International Classification of Disease (ICD) Clinical Modification 9 th and 10 th Revision diagnosis codes and operationalized as a binary variable (ever/never according to VA data).

Diabetes was assessed separately for Type 1 and Type 2; gestational diabetes was excluded. 

To evaluate disparities in the prevalence of physical health conditions between veterans with minoritized sexual orientations and veterans with non-minoritized sexual orientations, we calculated adjusted prevalence and prevalence ratios with corresponding 95% confidence intervals for each of the eleven health conditions overall and stratified by race/ethnicity. Race and ethnicity were combined and categorized to align with Heslin and Hall's groupings 7 as non-Hispanic Black, non-Hispanic White, non-Hispanic other (including non-Hispanic Asian, non-Hispanic American Indian/Alaskan Native, non-Hispanic Pacific Islander/Native Hawaiian), and Hispanic. Additionally, because the VA patient population has a majority of individuals assigned male sex at birth, the analyses were repeated and stratified by data in the birth sex field. Initial adjusted analyses included age at first VA visit, year of first VA visit, and sex assigned at birth (male or female). Unlike Heslin and Hall, sexual orientation identity was not measured via self-report but rather defined by documentation within VA health records. Our previous research showed documentation is not immediate upon entering VA and the majority of documentation occurs in mental health settings. 35 Thus, volume and type of healthcare visits can potentially confound the relationship between sexual orientation and outcomes. 36 To address the possibility that the two groups differed systematically with respect to healthcare utilization ( Figure 1 ) we calculated adjusted prevalence ratios (aPRs) accounting for the number of primary care and mental health encounters, each as a separate categorical variable (0, 1-9, 10+) in the model. Due to similar potential effects on utilization and morbidity, we investigated the impact of further adjusting for VA priority group. All veterans are assigned to 1 of 8 enrollment priority groups after applying to receive VA health care. The allocation is based on a variety of factors including serviceconnected disability level, income, and history of military service. Veterans were categorized as no service-connected disability (priority groups 7 and 8), low income (priority group 5), low/moderate disability (priority groups 2, 3, and 6), or high disability (priority groups 1 and 4).

Groupings were selected based on prior VA research. 37 All estimates were calculated using logistic regression with the margins and nonlinear combination of estimators (nlcom) postestimation commands using Stata version 15. Individuals missing data on ethnicity, race (except when ethnicity was recorded as Hispanic), sex assigned at birth, year of birth, or had year of birth inconsistent with a plausible age (<18 or >115) were excluded. 

Of the 8,312,455 million veterans who enrolled in VA after 10/01/1999, 115,911 had documentation indicating a minoritized sexual orientation, of which 108,401 (93%) were eligible for analysis based on non-missing covariate data. There were 8, 196 ,544 veterans with nonminoritized sexual orientation of which 6,511,698 (80%) were eligible for further analysis. The primary reason for exclusion in both cohorts was missing data on race/ethnicity. The majority of veterans were White non-Hispanic (76.6%), followed by Black non-Hispanic (14.4%), Hispanic (6.4%), and Other non-Hispanic (2.5%). Those with minoritized sexual orientations were significantly more likely to have been assigned female at birth compared to those without minoritized sexual orientations (31.6% vs 6.9%). Additionally, the minoritized sexual orientation cohort was significantly younger than those without minoritized sexual orientations and started in VA care at younger ages (41 vs 54 years of age, respectively). See Table 1 for complete comparison of study groups. In the initial adjusted analyses accounting for sex, age, and year, except for obesity, individuals with minoritized sexual orientations were 11-75% more likely to experience each of the conditions ( Table 2 , aPR1). Accounting for healthcare utilization reduced the magnitude of differences, but the patterns largely persisted except for obesity. After accounting for healthcare utilization, the prevalence of obesity was 8% lower among veterans with minoritized sexual orientations than veterans with non-minoritized sexual orientations (aPR2= 0.92, 95% CI= 0.91-0.93). Results were virtually unchanged when priority group was added to the models ( Table 2, aPR3).

Overall, when stratified by race/ethnicity, there was some variation in the differences observed in the overall analysis ( Table 2 ). The sexual orientation differences in outcomes in the overall analysis that were also observed among racial and ethnic groups included asthma, cancer, COPD, CKD, smoking, and stroke. For example, among individuals who were Black non-Hispanic, those with minoritized sexual orientations were almost 50% more likely to be a current smoker than those with non-minoritized sexual orientations (aPR1=1.48, 95% CI=1.44-1.51). Adjusting for healthcare utilization narrowed these disparities slightly, but significant differences between sexual orientation groups remained within each racial/ethnic category.

Varied findings emerged for heart disease, type 1 and type 2 diabetes, hypertension, and obesity.

In adjusted analyses, there were no significant differences in prevalence of type 2 diabetes between veterans with minoritized sexual orientation and non-minoritized sexual orientations among Black non-Hispanic, Other race, non-Hispanic, or Hispanic veterans ( Minoritized sexual orientation differences -Among sex assigned at birth

In analyses stratified by sex assigned at birth, conditions were consistently more frequent for veterans with minoritized sexual orientations than non-minoritized veterans assigned female at birth ( Figure 2 ). Specifically, smoking and COPD were 1.4 and 1.3 times higher, respectively, among those assigned female at birth with minoritized sexual orientations compared to 1.1 and 1.2 for those assigned male at birth. Obesity, in particular, stood out because minoritized sexual orientation had a negative association with obesity for veterans assigned male at birth (aPR=0.88, 95% CI= 0.87-0.88), but a null association for veterans assigned female at birth (aPR=0.99, 95% CI= 0.97-1.00. Figure 2 . Adjusted prevalence ratios of physical health conditions among veterans with minoritized sexual orientations and non-minoritized sexual orientations stratified by birth sex

Using nationwide EHR data from the VA, we found veterans with minoritized sexual orientations were disproportionally burdened by many physical health conditions. This is the first study to empirically describe the physical health of veterans with minoritized sexual orientations using EHR data. Greater prevalence of health conditions among people with minoritized sexual orientations is a logical epidemiologic sequela given the large literature documenting greater prevalence of both health risk behaviors and minority stress among individuals with minoritized sexual orientations than non-minoritized identities. 1 Although we largely replicated the methods of Heslin and Hall and had similar overall findings, the prevalence of physician-diagnosed conditions in the present results were much larger than those of the BRFSS self-reported data. For instance, adjusted prevalence of cancer and heart disease among veterans with minoritized sexual orientations was considerably greater (34.4%

and 67.3%,) than Heslin and Hall's findings (9.2% and 8.0%). There are several possible explanations for these differences. First, the present study consisted of individuals who were engaged with a healthcare system and is consequently comprised of patients with poorer health status than the general population. 41 Second, although the demographic profile of the VA is shifting younger and more diverse, historically veterans tend to be older and have worse health status and more health risk behaviors than their civilian counterparts. 42, 43 Population prevalence differences notwithstanding, these clinical data from a nationwide healthcare system aligned with the overall population disparity patterns of self-report data from a national survey.

The Empirically describing potential COVID-19 risk factors among subgroups is especially important for healthcare systems to better understand patient needs. In the context of the health disparities framework proposed by Kilborne et al, 44 these results raise important questions for future research into factors driving these disparities and proposing and testing solutions to overcome them. For instance, we found similarly high rates of hypertension and type 2 diabetes between sexual orientation groups, but whether disease management is similar between groups is unclear.

Uncontrolled diabetes is an established risk factor for both ischemic and hemorrhagic stroke, 45 while sustained hypertension is a leading cause of impaired glomerular filtration and CKD. 46 Among the largest disparities observed in this study were stroke (aPR=1.26) and CKD (aPR=1.20). Recent research on asthma supports that notion that COVID-19 severity is more strongly related to current disease status (active vs inactive) rather than simply disease history (ever/never). 12 Future research is therefore warranted to understand differential disease control between sexual orientation groups. Preventing disease from developing and adequately treating disease for those in whom it has manifested are realistic goals for any healthcare system and should not be impacted by a patient's minority identity.

The data presented in this study are not without limitations. For the present study sexual orientation was extracted from clinical notes and outcomes were extracted from administrative coding. Ideally, sexual orientation was self-reported by patients and subsequently recorded in their EHR by a clinician, but we cannot confirm this sequence had occurred for all patients.

Providers may document based on their assumptions or interpretations, both which may not align with patient's self-identified sexual orientation. EHRs are not subject to the same degree of recall bias as survey data, yet there are other possible sources of bias that should be considered. First, the degree and context in which a patient interacts with a healthcare system can impact both the volume and type of information that are recorded in EHRs and subsequently available to researchersa term coined by Goldstein and colleagues as informed presence bias, 36 a type of ascertainment bias specific to EHR data. To mitigate the extent to which this type of bias could distort estimated prevalence ratios, we simultaneously adjusted for the number of mental health and primary care visits. However, residual confounding could remain, which could potentially overestimate the true association between sexual orientation and health outcomes. Second, the selection process used to identify patients for this analysis could have impacted the findings.

Specifically, we were only able to assess the prevalence of physical health conditions among veterans with minoritized sexual orientations who used VA services. As such, results may not be generalizable to all veterans with minoritized sexual orientations. Related, veterans with minoritized sexual orientations who did not disclose this information to a VA provider or if the provider did not document the disclosure would have been erroneously included in the nonminoritized sexual orientation group. While possible, veterans with minoritized sexual orientations have reported less need for confidentiality and more willingness to disclose sexual orientation to VA providers since the repeal of Don't Ask Don't Tell, a policy that barred gay military members from serving openly and the addition of a VA policy in 2014 47 assuring affirming care at VA. 48 Therefore, we do not believe this type of misclassification would have significantly affected our findings. Lastly, physical health conditions were defined by ICD codes which also has limitations. Using administrative data for research is common, yet the accuracy is variable across diseases, healthcare settings, and calendar time. 49, 50 However, such errors, if present, would likely be nondifferential and bias results towards a null association.

The present study of EHR data largely corroborates findings from a recent study of self-report survey data showing sexual orientation-related disparities in medical diagnoses associated with higher risk for COVID-19 severity and mortality. Further research is needed to determine if disparities in these conditions, which can antagonize COVID-19 symptoms, result in poorer outcomes for individuals with minoritized sexual orientations.

This work was supported using resources and facilities at the VA Salt Lake City Health Care

System and the VA Informatics and Computing Infrastructure (VINCI), VA HSR RES 13-457.

This work described has not been published previously, nor is it under consideration for publication elsewhere.

The views expressed in this article are those of the authors and do not represent the views of the U.S. Department of Veterans Affairs.

The Health of Lesbian, Gay, Bisexual, and Transgender People: Building a Foundation for Better Understanding

Tobacco use among sexual minorities in the USA

Substance use in lesbian, gay, and bisexual populations: an update on empirical research and implications for treatment

The prevalence of sexual assault against people who identify as gay, lesbian, or bisexual in the United States: A systematic review

A review of the literature on LGBTQ adults who experience homelessness

A rapid review of gender, sex, and sexual orientation documentation in electronic health records

Sexual orientation disparities in risk factors for adverse COVID-19-related outcomes, by race/ethnicity-Behavioral Risk Factor Surveillance System

Sexual and gender minority health in the COVID-19 pandemic: Why data collection and combatting discrimination matter now more than ever

Preliminary estimates of the prevalence of selected underlying health conditions among patients with coronavirus disease 2019-United States

Comorbidities and the risk of severe or fatal outcomes associated with coronavirus disease 2019: A systematic review and meta-analysis

COVID-19 severity is tripled in the diabetes community: A prospective analysis of the pandemic's impact in type 1 and type 2 diabetes

Asthma disease severity, COPD, and COVID-19 severity in a large multiethnic population

Hospitalization and mortality among black patients and white patients with Covid-19

Characteristics associated with racial/ethnic disparities in COVID-19 outcomes in an academic health care system

Still in the dark regarding the public health impact of COVID-19 on sexual and gender minorities

The needs for COVID-19 LGBTQ-specific data

Disparities in smoking and acute respiratory illnesses among sexual minority young adults

Asthma disparities and within-group differences in a national, probability sample of same-sex partnered adults

The association of asthma, sexual identity, and inhaled substance use among US adolescents

Asthma diagnosis among individuals in same-sex relationships

CVD risk among men participating in the National Health and Nutrition Examination Survey (NHANES) from 2001 to 2010: differences by sexual minority status

A population-based study of cardiovascular disease risk in sexual-minority women

Risk of type 2 diabetes among lesbian, bisexual, and heterosexual women: Findings from the Nurses' Health Study II

Health inequalities among sexual minority adults: evidence from ten US states

A population-based study of sexual orientation identity and gender differences in adult health

Health disparities among lesbian, gay, and bisexual older adults: Results from a population-based study

Sensitivity of household reported medical conditions in the medical expenditure panel survey

Validation of self-reported chronic conditions and health services in a managed care population

Agreement between self-report questionnaires and medical record data was substantial for diabetes, hypertension, myocardial infarction and stroke but not for heart failure

Incrementally transforming electronic medical records into the observational medical outcomes partnership common data model: a multidimensional quality assurance approach

Insights from advanced analytics at the Veterans Health Administration

The utility of clinical notes for sexual minority health research

Using enriched samples for semi-automated vocabulary expansion to identify rare events in clinical text: sexual orientation as a use case

Validating smoking data from the Veteran's

Affairs Health Factors dataset, an electronic data source

Variation in sexual orientation documentation in a national electronic health record system

Controlling for informed presence bias due to the number of health encounters in an electronic health record

Virtual care expansion in the Veteran Health Administration during the COVID-19 pandemic: clinical services and patient characteristics associated with utilization

Overweight and obesity among sexual minority adults in the United States

Health disparities among sexual minority women veterans

Overweight and obesity in sexual-minority women: evidence from population-based data

A data quality assessment guideline for electronic health record data reuse

Are patients at Veterans Affairs medical centers sicker? A comparative analysis of health status and medical resource use

The health status of veteran employees compared to civilian employees in Veterans Health Administration

Advancing health disparities research within the health care system: a conceptual framework

Diabetes and stroke: epidemiology, pathophysiology, pharmaceuticals and outcomes

Hypertension and prehypertension and prediction of development of decreased estimated GFR in the general population: a meta-analysis of cohort studies

Perceived stigma, discrimination, and disclosure of sexual orientation of sexual orientation among a sample of lesbian veterans receiving care in the Department of Veterans Affairs

Department of Veterans Affairs: Provision of health care for Veterans who identify as lesbian, gay or bisexual

Measuring diagnoses: ICD code accuracy

Comparing chronic condition rates using ICD-9 and ICD-10 in VA patients FY2014-2016

The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.