State-level heterogeneity in associations between structural stigma and individual health care access: A multilevel analysis of transgender adults in the United States

Abstract

Objective

State-level variation in how restrictive policies affect health care access for transgender populations has not been widely studied. Therefore, we assessed the association between structural stigma and four measures of individual health care access among transgender people in the United States, and the extent to which structural stigma explains state-level variability.

Methods

Data were drawn from the 2015–2019 Behavioral Risk Factor Surveillance System and the Human Rights Campaign’s State Equality Index. We calculated weighted proportions and conducted multilevel logistic regression of individual heterogeneity and discriminatory accuracy.

Results

An increase in the structural stigma score by one standard deviation was associated with lower odds of health care coverage (OR = 0.80; 95% CI: 0.66, 0.96) after adjusting for individual-level confounders. Approximately 11% of the total variance for insurance coverage was attributable to the state level; however, only 18% of state-level variability was explained by structural stigma. Adding Medicaid expansion attenuated the structural stigma-insurance association and explained 22% of state-level variation in health insurance. For the remaining outcomes (usual source of care, routine medical check-up, and cost-related barriers), we found neither meaningful associations nor considerable between-state variability.

Conclusions

Our findings support the importance of Medicaid expansion and transgender-inclusive antidiscrimination protections to enhance health care insurance coverage. From a measurement perspective, however, additional research is needed to develop and validate measures of transgender-specific structural stigma to guide future policy interventions.

Keywords

structural stigma health care access transgender

Introduction

Transgender people report more restricted access to health insurance coverage and health care relative to their cisgender counterparts in the United States (US).^1–3 Evidence from the 2014–2017 Behavioral Risk Factor Surveillance System (BRFSS) indicated that the prevalence of having health insurance and not experiencing a financial barrier to care were six percentage points higher among cisgender than transgender respondents.¹ The TransPop survey, a population sample of transgender people drawn from the annual Gallup Poll, did not find differences in health insurance coverage between transgender and cisgender respondents but did identify cost-related barriers to care among transgender respondents.³ Some have posited that differences in findings between these populations may be related to sampling procedures, geographic coverage, and gender identification process (one-step vs two-step).³ Taken together, however, these data indicate that transgender individuals continue to experience unjust barriers to health care in the US.⁴

Several studies have attributed health care barriers among transgender people to interpersonal factors such as discrimination in medical care^5,6 and a lack of provider knowledge of transgender health needs.^7,8 At the structural level, these barriers to care may be further shaped by structural stigma, defined by Hatzenbuehler et al. as the ‘societal conditions that constrain an individual’s opportunities, resources, and wellbeing.’^9(p813) The structural stigma framework posits that, because these conditions can operate through government institutions to influence disease outcomes and risk factors for members of stigmatized communities, public policy can be both an important indicator of the presence of structural stigma and a potential lever for intervening on stigma as a fundamental cause of health disparities.¹⁰

Structural stigma against transgender people results from a rigid gender binary that labels non-cisgender gender identity and expression as non-normative and leads to a lack of resources, protection, and support for transgender people.^6,11 Forms of anti-transgender structural stigma that affect access to health care include insurance exclusions for gender-affirming health care¹¹ and a lack of transgender-inclusive nondiscrimination protections at the state level.^12–14 Several studies, for example, have indicated that living in a state with fewer nondiscrimination protections is associated with decreased health care access among transgender people.^12,13

While the literature around structural stigma and health care access among transgender people has been informative, few attempts have been made to quantify how the state context influences health care access and the degree to which between-state differences in health care access among the transgender population can be explained by structural stigma. Relying solely on average measures of association across states provides limited information on the heterogeneity that may underlie the effect of the state policy context.¹⁵ This hinders efforts to improve health care access among transgender people. Furthermore, research on structural stigma among transgender people has primarily relied on convenience samples.^12–14 Although there are merits to convenience sampling (e.g. cost-effectiveness), this approach represents a limitation in the literature for generating estimates that are generalizable to a larger population.

Our objectives in the current study are to address these limitations. First, we assessed the association between structural stigma – as measured by the Human Rights Campaign’s State Equality Index (SEI) (discussed below) – and four common indicators of health care access, using a population sample of BRFSS data from 38 states. Second, we quantified how much of the state-level variability in each outcome is explained by the socio-political environment as captured by the SEI.

By focusing on the state policy context, we aim to better understand how structural stigma may influence health care access among transgender people, which can help inform future state- and federal-level policy strategies to reduce inequities affecting the transgender population.

Methods

Study design and data source

We conducted a pooled cross-sectional analysis using data from the 2015–2019 BRFSS, a telephone-based health survey that is designed to be nationally representative of non-institutionalized adults (age 18 years and older) in the US. Details about BRFSS, including the sampling and weighting methodology, can be found elsewhere.¹⁶ Our analysis included data from residents of 38 states that implemented the gender identity question module at least once between 2015 and 2019. The module was first implemented in 2014 before the establishment of marriage equality throughout the US in 2015. Thus, inclusion of data for 2014 would have biased our results given the changes in the socio-political landscape and policy data collected by the Human Rights Campaign.

As part of the module, participants were asked, ‘Do you consider yourself to be transgender?’ with the following response options: (1) yes, transgender, male to female; (2) yes, transgender, female to male; (3) yes, transgender, gender non-conforming; and (4) no. The analytic sample was restricted to participants who self-identified as transgender. The study protocol was deemed exempt by the Drexel University Institutional Review Board.

Exposure

Structural stigma was derived from the SEI from 2015 to 2019. We used the SEI as a proxy for anti-transgender stigma as it is a global measure of discrimination against sexual and gender minority (SGM) populations. The SEI provides an annual score for all 50 U.S. states and Washington, DC, based on a comprehensive review of statewide laws and policies that affect SGM populations.¹⁷ SEI scores are generated based on six domains, including policies related to parental rights (e.g. second parent adoption), hate crimes and criminal justice (e.g. sodomy laws, HIV criminalization), nondiscrimination protections (e.g. nondiscrimination in employment, housing, public accommodations, and education), religious exemptions (e.g. state religious freedom restoration acts), protections for SGM youth (e.g. prohibition of conversion therapy), and health care protections (e.g. bans on insurance exclusion for gender-affirming health care). For each policy, a state may receive a score of −1 if the policy is harmful (i.e. restricts the resources and rights of SGM populations), a score of +1 if the policy is protective (i.e. protects the resources and rights of SGM populations), and a score of 0 if the policy does not exist. Final scores are generated by summing across all six domains.

For our analysis, we reverse-coded SEI scores to derive a measure of structural stigma, with higher scores representing a lower level of legal equality and thus a higher level of structural stigma. We also centered structural stigma (i.e. subtracted the mean stigma score from each individual stigma value) and divided by its standard deviation (SD). Centering at the grand mean facilitates the interpretation of the intercept (i.e. provides the expected value when structural stigma is at its mean value rather than zero) and often reduces challenges with model convergence.

Outcomes

We used four health care access measures in the core component of BRFSS for our analysis: health insurance, usual source of care, routine medical check-up, and cost-related barriers. Participants were asked at the time of their interview if they had any kind of health insurance coverage (yes/no). We dichotomized responses to having a usual health care provider as none or ≥1. Respondents also indicated whether they had a routine check-up and had encountered cost-related barriers to health care (both in the past 12 months; yes/no).

Covariates

Socio-demographic characteristics used to describe the analytic sample included age; race/ethnicity (non-Hispanic White, non-Hispanic Black, Latino/a/e/Hispanic, non-Hispanic Asian, and non-Hispanic other); gender identity (transgender women, transgender men, and gender non-conforming people); sexual orientation (heterosexual, gay/lesbian, bisexual, and other); education level (<high school, high school graduate, some college/tech school, and college/tech school graduate); employment (employed or self-employed; unemployed; homemaker, student, or retired; unable to work; and don’t know/refuse to answer); and annual income ($US) (<$15,000, $15,000–$24,999, $25,000-$34,999, $35,000-$49,999, ≥$50,000 and do not know/refuse to answer). At the state level, we used data from the Kaiser Family Foundation to assess whether a state had expanded its Medicaid program between 2010 and 2019 (i.e. as part of the expansion initiated by the Affordable Care Act). States that had done so were coded as ‘yes’.¹⁸

Statistical analyses

Weighted descriptive statistics for each socio-demographic characteristics and measures of health care access were calculated to describe the sample, and repeated measures one-way ANOVA was used to assess whether structural stigma varied over time. Because of the hierarchical structural and complex sampling design of BRFSS, we scaled sampling weights in regression analyses so that the new weights summed to the state’s sample size.¹⁹

Our main modeling strategy used a stepwise approach to fit four multilevel logistic models of individual heterogeneity and discriminatory accuracy for each individual outcome.²⁰ Model 1 used conventional logistic regression to regress each health care access outcome on survey year and individual age, while ignoring the clustering of individuals within states. Model 2 then used multilevel logistic regression with state-level random intercepts to decompose the total individual variance into the between- and within-state variance. Model 3 added state-level structural stigma to estimate its association with measures of individual health care access. Finally, Model 4 included Medicaid expansion as a potential state-level confounder. Individual-level covariates such as marital status and measures of adulthood socio-economic position (i.e. education level, employment, and income) were not included, as these are potential mediators of the association between structural stigma and health care access.

To accompany our measures of association, we calculated a series of estimates to quantify the between-state variation of health care access and the degree to which differences between states could be explained by the SEI-derived measure of structural stigma. Measures of between-state variability included the intraclass correlation coefficient (ICC) as well as median odds ratio (MOR) and 95% credible interval (CrI). These complementary measures quantify the effect of clustering; for instance, the ICC represents the percentage of total variance between individuals that is attributable to the state level, while the MOR estimates the probability (on the odds scale) of individual health care access given the random state-level effect.²⁰ Next, we calculated the proportional change in variance (PCV) between 2 successive models (e.g. Model 2–3)²⁰ to determine the amount of state-level variance explained in health care access with the addition of structural stigma and Medicaid expansion. We also calculated the proportion of opposed odds ratios (POOR) to assess the degree to which the overall associations between structural stigma and health care access were driven by the exposure itself or the residual variability between states.²⁰ The POOR is an estimate of the proportion of state-specific odds ratio (OR) with opposite direction to the overall OR; it can take on values between 0 and 50%.²⁰ This measure provides an indication of the heterogeneity of the exposure-outcome association across states, with higher POOR indicative of greater heterogeneity. Finally, we calculated the area under the receiver operator characteristics (AU-ROC) curve to assess the discriminatory accuracy of each model.²⁰

We conducted two sensitivity analyses. First, given the potential concerns with BRFSS weights producing inaccurate estimates for the transgender population,²¹ we re-estimated our models using unweighted data. Second, it is possible that higher levels of structural stigma may reflect aspects of the socio-political environment that affect health care access for cisgender and transgender people. We evaluated this assumption using a negative control approach²² and fitted our models to the cisgender population, estimating the exposure-outcome association and measures of state-level variability (i.e. PCV and ICC). Analyses were conducted in R Studio version 4.0.3²³ using the packages survey²⁴ and lme4.²⁵

Results

Participant characteristics and Medicaid expansion

The analytic sample consisted of 4335 transgender individuals living in 38 different states who completed the BRSS gender identity question at least once from 2015 to 2019 (Table 1). Of those, 1464 were transgender men, 1881 were transgender women, and 990 were gender non-conforming people. Eighty-one percent of respondents had some sort of health insurance coverage at the time of interview, 73.6% had access to a usual source of health care, and 70.4% had received a routine check-up in the past 12 months. However, 21.8% of respondents reported experiencing unmet health care needs because of cost in the past 12 months. Between 2010 and 2019, 66% (n = 33) of states adopted Medicaid expansion.

Table 1.

Socio-demographic characteristics and health care access among transgender adults in the US, 2015–2019 (n = 4335).

	n	Weighted% (SE)^a
Age (years)
18–24	586	25.9 (1.5)
25–34	598	18.9 (1.2)
35–44	475	13.3 (1.0)
45–54	621	12.9 (0.9)
55–64	856	14.0 (1.1)
≥65	1199	15.0 (0.9)
Race/ethnicity
White, non-Hispanic	2812	54.6 (1.5)
Black, non-Hispanic	438	13.8 (1.1)
Latino/a/e/Hispanic	497	20.6 (1.3)
Asian, non-Hispanic	153	5.1 (0.8)
Other, non-Hispanic	435	6.0 (0.5)
Gender identity
Transgender men	1464	32.6 (1.4)
Transgender women	1881	43.5 (1.5)
Gender non-conforming people	990	23.9 (1.3)
Sexual orientation
Heterosexual	2714	53.5 (1.5)
Gay/Lesbian	371	9.9 (0.9)
Bisexual	636	18.8 (1.2)
Other	614	17.8 (1.3)
Education level (n = 4316)
<High school	632	24.2 (1.5)
High school graduate	1483	32.4 (1.3)
Some college/tech school	1133	28.3 (1.3)
College/tech school graduate	1068	15.1 (0.8)
Employment (n = 4331)
Employed or self-employed	1996	49.6 (1.5)
Unemployed	308	9.2 (1.0)
Homemaker, student, or retired	1435	27.6 (1.3)
Unable to work	547	12.4 (1.0)
Don’t know/Refuse to answer	45	1.3 (0.3)
Annual income ($US) (n = 4319)
<$15,000	604	15.6 (1.1)
$15,000-$24,999	853	20.0 (1.2)
$25,000-$34,999	461	10.3 (0.6)
$35,000-$49,999	453	7.7 (0.6)
≥ $50,000	1241	28.7 (1.3)
Don’t know/Refuse to answer	707	17.7 (4.5)
Health care coverage	3763	81.2 (1.2)
Usual source of care	3438	73.6 (1.3)
Routine medical check-up, past 12 months	3268	70.4 (1.4)
Cost-related barriers, past 12 months	757	21.8 (1.3)

^aSE: design-adjusted standard error. Percentages were calculated based on sampling weights, meaning they reflect the population prevalence rather than the sample prevalence.

Structural stigma

The mean SEI-derived structural stigma score was −31.7 (SD = 32.2), with higher levels in Southeastern US states such as Alabama and lower levels in Western and Northeastern states such as California and New York (Figure 1). From 2015 to 2019, there was a slight decrease in structural stigma by roughly 5% (−29.1 to −30.4; p = 0.85) across all 38 states in the sample.

Figure 1.

Choropleth map of structural stigma over time across states 2015–2019. Structural stigma for states that have implemented the sexual orientation and gender identity (SOGI) question module, as measured by the Human Rights Campaign’s State Equality Index. Lighter shades of grey indicate states with lower levels of structural stigma, darker shades indicate higher levels of structural stigma. States with diagonal lines are those that did not implement the SOGI question module that year.

Association of structural stigma and health care access

In comparing Models 1 and 2, we found that health care access for transgender adults varied substantially across states (Tables 2 and 3). The ICC in Model 2 was highest for insurance coverage (11%) and lowest for cost-related barriers (1%), indicating that state-level variance (i.e. the general state-level effect) is important for understanding the individual odds of health care access, particularly for insurance coverage among transgender adults. In addition, the MOR ranged from 1.20 to 1.81 with all 95% CrIs excluding 1. The MOR of 1.81 of insurance coverage, for instance, can be interpreted as meaning that a transgender person’s median odds of having coverage would nearly double if they moved to another state with a higher prevalence of insured transgender persons.

Table 2.

Single (Model 1) and multilevel (Models 2–4) logistic regression analyses of not having health care coverage and access to a usual source of care among transgender adults in the US, 2015–2019.

	Not having health care coverage				Usual source of care
	Single-level model	Multilevel models			Single-level model	Multilevel models
	Model 1	Model 2	Model 3	Model 4	Model 1	Model 2	Model 3	Model 4
	OR (95% CI)	OR (95% CI)	OR (95% CI)	OR (95% CI)	OR (95% CI)	OR (95% CI)	OR (95% CI)	OR (95% CI)
Year	1.07 (1.00, 1.13)	1.07 (1.00, 1.15)	1.05 (0.98, 1.13)	1.08 (1.01, 1.16)	0.98 (0.93, 1.03)	0.97 (0.92, 1.03)	0.97 (0.92, 1.03)	0.99 (0.93, 1.05)
Age (years)
18–24	Ref	Ref	Ref	Ref	Ref	Ref	Ref	Ref
25–34	0.63 (0.50, 0.78)	0.62 (0.49, 0.78)	0.62 (0.49, 0.78)	0.63 (0.49, 0.79)	0.92 (0.76, 1.11)	0.93 (0.77, 1.14)	0.93 (0.77, 1.14)	0.95 (0.78, 1.16)
35–44	0.60 (0.47, 0.77)	0.57 (0.44, 0.75)	0.57 (0.44, 0.75)	0.58 (0.44, 0.76)	1.19 (0.95, 1.50)	1.19 (0.94, 1.52)	1.20 (0.94, 1.52)	1.19 (0.94, 1.52)
45–54	1.08 (0.82, 1.44)	1.07 (0.79, 1.45)	1.07 (0.79, 1.45)	1.11 (0.82, 1.50)	1.71 (1.35, 2.19)	1.73 (1.34, 2.24)	1.74 (1.34, 2.25)	1.78 (1.38, 2.31)
55–64	1.80 (1.33, 2.46)	1.88 (1.34, 2.64)	1.88 (1.34, 2.64)	1.90 (1.35, 2.67)	2.36 (1.84, 3.04)	2.47 (1.87, 3.24)	2.47 (1.88, 3.24)	2.49 (1.89, 3.28)
≥65	4.58 (3.08, 7.07)	4.43 (2.81, 7.00)	4.46 (2.82, 7.05)	4.49 (2.85, 7.09)	4.58 (3.43, 6.20)	4.65 (3.33, 6.49)	4.66 (3.34, 6.51)	4.73 (3.38, 6.62)
State-level characteristics
SEI-derived structural stigma (±SD.)			0.80 (0.66, 0.96)	0.99 (0.79, 1.24)			0.97 (0.83, 1.19)	1.03 (0.85, 1.24)
POOR			0.39	0.49			0.48	0.48
Medicaid expansion (yes/no)				2.22 (1.36, 3.62)				1.26 (0.83, 1.92)
Random effects
State-level variance		0.381	0.320	0.23		0.186	0.180	0.174
PCV		Ref	0.180	0.40		Ref	0.027	0.061
ICC		0.106	0.088	0.066		0.053	0.052	0.050
MOR (95% CrI)		1.81 (1.64, 2.02)	1.71 (1.65, 2.02)	1.58 (1.48, 1.77)		1.51 (1.39, 1.63)	1.50 (1.39, 1.62)	1.49 (1.39, 1.62)
AU-ROC	0.662	0.704	0.704	0.704	0.680	0.696	0.696	0.696

Models were estimated using survey-weighted logistic regression, modified to account for multiple survey years and hierarchical structure of individuals clustered in states.

OR: odds ratio; CI: confidence interval; Ref: the reference model, used to calculate the PCV; SEI: State Equality Index; SD: standard deviation; POOR: proportion of opposed odds ratios; PCV: proportional change in the variance; ICC: intraclass correlation coefficient; MOR: median odds ratio; CrI: credible interval; AU-ROC: area under the receiver operator characteristics curve.

Table 3.

Single (Model 1) and multilevel (Models 2–4) logistic regression analyses of not having a routine medical check-up and experiencing cost-related barriers in the past 12 months among transgender adults in the US, 2015–2019.

	Not having routine medical check-up				Cost-related barriers
	Single-level model	Multilevel models			Single-level model	Multilevel models
	Model 1	Model 2	Model 3	Model 4	Model 1	Model 2	Model 3	Model 4
	OR (95% CI)	OR (95% CI)	OR (95% CI)	OR (95% CI)	OR (95% CI)	OR (95% CI)	OR (95% CI)	OR (95% CI)
Year	1.10 (1.05, 1.16)	1.08 (1.02, 1.14)	1.07 (1.01, 1.13)	1.11 (1.05, 1.17)	1.08 (1.03, 1.14)	1.08 (1.02, 1.14)	1.08 (1.02, 1.15)	1.06 (1.00, 1.13)
Age (years)
18–24	Ref	Ref	Ref	Ref	Ref	Ref	Ref	Ref
25–34	0.91 (0.75, 1.10)	0.93 (0.76, 1.13)	0.93 (0.76, 1.13)	0.93 (0.76, 1.13)	1.22 (1.00, 1.49)	1.22 (1.00, 1.49)	1.22 (1.00, 1.49)	1.25 (1.02, 1.53)
35–44	1.03 (0.83, 1.29)	1.03 (0.82, 1.30)	1.03 (0.82, 1.30)	1.03 (0.82, 1.30)	0.89 (0.70, 1.19)	0.91 (0.71, 1.16)	0.91 (0.71, 1.16)	0.94 (0.74, 1.20)
45–54	1.47 (1.17, 1.86)	1.41 (1.10, 1.80)	1.41 (1.11, 1.81)	1.41 (1.11, 1.81)	0.62 (0.48, 0.80)	0.61 (0.46, 0.80)	0.61 (0.46, 0.79)	0.61 (0.47, 0.80)
55–64	1.92 (1.52, 2.44)	2.02 (1.56, 2.62)	2.02 (1.56, 2.63)	2.03 (1.57, 2.64)	0.46 (0.35, 0.60)	0.42 (0.31, 0.56)	0.42 (0.31, 0.56)	0.40 (0.30, 0.54)
≥65	4.16 (3.17, 5.53)	4.15 (3.03, 5.69)	4.16 (3.04, 5.71)	4.17 (3.05, 5.72)	0.32 (0.24, 0.42)	0.34 (0.25, 0.47)	0.34 (0.25, 0.47)	0.34 (0.25, 0.47)
State-level characteristics
SEI-derived structural stigma (±SD.)			0.93 (0.81, 1.07)	1.05 (0.89, 1.24)			1.12 (1.02, 1.22)	1.04 (0.94, 1.15)
POOR			0.44	0.46			0.26	NA
Medicaid expansion (yes/no)				1.61 (1.11, 2.33)				0.75 (0.60, 0.94)
Random effects
State-level variance		0.153	0.139	0.118		0.036	0.016	0.000
PCV		Ref	0.090	0.23		Ref	0.568	1.000
ICC		0.045	0.041	0.035		0.011	0.005	0.000
MOR (95% CrI)		1.45 (1.33, 1.54)	1.44 (1.32, 1.53)	1.39 (1.29, 1.47)		1.20 (1.16, 1.27)	1.13 (1.10, 1.17)	NA
AU-ROC	0.660	0.675	0.675	0.675	0.656	0.661	0.657	0.657

Models were estimated using survey-weighted logistic regression, modified to account for multiple survey years and hierarchical structure of individuals clustered in states.

In Model 3, after adjusting for survey year and individual age, we found that an increase in the structural stigma score by one standard deviation was associated with 20% lower odds of health insurance coverage (OR = 0.80; 95% CI: 0.66, 0.96; Table 2), and that the state-level variance decreased by 18% with the addition of structural stigma. The POOR was 39%, indicating that the odds of having insurance would be higher in states with higher structural stigma scores in 39% of pairwise state comparisons. Adjusting for Medicaid expansion in Model 4 attenuated the association between structural stigma and health insurance (OR = 0.99; 95% CI: 0.79, 1.24) and further explained 22% of the between-state variation in insurance coverage (total PCV = 40%). We did not find meaningful associations nor considerable between-state variation for usual source of care (OR = 0.97; 95% CI: 0.83, 1.19; ICC = 5%), routine medical check-up (OR = 0.93; 95% CI: 0.81, 1.07; ICC = 4%), and cost-related barriers (OR = 1.12; 95% CI: 1.02, 1.22; ICC = 1%) after adjusting for age and survey year in Model 3 or Medicaid expansion in Model 4.

Sensitivity analysis

Results from our sensitivity analyses using unweighted data indicated similar results to the main models in both the magnitude and direction of the association (Online Supplement Tables S1 and S2). In comparing transgender and cisgender adults (Online Supplement Table S3), the association of structural stigma and health care access differed in both magnitude and direction for health care coverage, usual source of care, and routine check-up (Online Supplement Figure S1). Furthermore, the state context had a greater influence on individual health care access for transgender people as compared to their cisgender counterparts (Online Supplement Figure S2), and the addition of structural stigma increased the between-state variability for cisgender populations (Online Supplement Figure S3), indicating that the SEI-derived structural stigma variable did not explain between-state differences in health care access outcomes for cisgender people.

Discussion

In this study, we demonstrated that, of the four health care access outcomes, state-level effects most strongly conditioned whether US transgender adults had insurance coverage from 2015–2019, and this association was partly explained by the socio-political context of the state as quantified by an SEI-derived variable measuring structural stigma. This finding is consistent with the current literature, which documents that living in a less protective state can negatively impact health care access for transgender people.^13,14 Adding Medicaid expansion, however, reduced the magnitude of association between structural stigma and insurance coverage. This underscores the importance of Medicaid expansion for addressing the structural inequities in health care access for transgender people, as more than half a million²⁶ were living in states that had not adopted Medicaid expansion at the end of 2019.

Critical to our findings is that the association between structural stigma and insurance coverage may be more dynamic in the US than initially anticipated. In similar analyses examining the influence of the socio-political context among transgender participants in the National Transgender Discrimination Survey¹⁴ and transgender patients identified using diagnostic codes in the Veterans Administration database,¹² investigators found that the between-state variance was lower than expected (ICC≤1%). In contrast, we found that roughly 11% of the total variability for insurance coverage was explained by the state context and that 18% of the state context was explained by structural stigma, after controlling for survey year and individual age. These findings provide two important insights: (1) the determinants of having health insurance may be more strongly influenced by the contextual environment of a state than has been indicated in previous studies, and (2) further investigation is needed to understand the degree of between-state variability in insurance access that could be theoretically eliminated with state and federal laws that seek to expand insurance access by mitigating structural stigma across the domains measured by the SEI. Thus, future research and policies should evaluate the effect of state-level clustering in these effects, as this can inform policies and help direct resources for health interventions that are influenced by the state-level context.¹⁵

Geographical variation in health insurance coverage among transgender people may reflect differences in the adoption of Medicaid expansion and antidiscrimination policies related to coverage exclusions for gender-affirming care.⁴ States that have adopted these policies, on average, have higher numbers of transgender people having health insurance.²⁶ Despite these reforms, challenges remain for transgender people seeking health insurance. For instance, not all states adopting Medicaid expansion guarantee coverage for gender-affirming care.²⁶ Furthermore, Canadian data indicate that, even with universal health insurance, transgender people continue to experience unmet health care needs, in part due to differences in local administration of health funds and policies for funding specialty clinics and surgeries.²⁷ Therefore, even in countries with universal health insurance, community-level structural stigma may continue to play a role in restricting insurance access for transgender people. However, studying this would require structural stigma indices that consider attitudes as well as formal laws, as most countries do not have substantial local variation in SGM-related laws.

Results indicated that 82% of the between-state variance in insurance coverage was not captured by the SEI-derived structural stigma measure. This unexplained variance may be attributable to (1) our use of SEI as a proxy measure of anti-transgender stigma, and (2) the difficulty of fully capturing pertinent dimensions of the socio-political environment of US states for transgender people. The SEI score, like similar composite measures of structural stigma,¹³ measures the presence or absence of policies but not the extent of their implementation (or lack thereof). The extent to which policies are implemented likely contributes to state-level variation in insurance access. Thus, our results indicate that future US-based research should also consider attitude-based measures of structural stigma and should incorporate dimensions not captured in the SEI, such as structural racism, economic, and gender inequality, and Medicaid expansion to further address the gap between laws on the books and the reality experienced by transgender people. This also underscores the need to adopt an intersectional framework²⁸ in advancing research on anti-transgender structural stigma.

We did not find associations between our measure of structural stigma and the other health care access indicators. That is, although there may be an association between the state environment and the outcomes of having a usual source of care, routine check-up, and experiencing cost-related barriers to care, between-state variability in these outcomes was low and not well explained by the SEI-derived measure of structural stigma. This suggests that mechanisms operating at other levels (e.g. interpersonal) might better explain variability in these outcomes. Prior work indicates that previous encounters with interpersonal discrimination in clinical settings may largely explain the absence of a usual primary care provider and delays in routine medical visits among transgender people.^5,11 Our findings do not rule out structural approaches to expanding health care access by addressing interpersonal barriers to care, such as requirements for provider training in transgender cultural and clinical competency, but rather encourage the use of a multilevel framework (e.g. socio-ecological models) to tailor policy interventions that can effectively address stigma at multiple levels.^6,29 They also do not discredit the importance of enacting protective policies for transgender people as a means of improving a variety of health care access outcomes, as research has shown how structural stigma in the form of hostile socio-political environments can reinforce individual-level stigma against transgender people.⁶

Limitations

Beyond those already described, it is important to interpret these data in the context of additional limitations. First, there have been recent concerns that BRFSS weights, which are based in part on a problematic binary sex variable, may produce inaccurate estimates for the US transgender population.³⁰ We addressed this concern by excluding weights from our models as part of our sensitivity analysis, which indicated that our findings were robust to the use of BRFSS survey weights.

Second, our sample does not include respondents from every US state. Although this may not be a major concern in terms of producing valid estimates, we are unable to generalize our results to the entire US transgender population.

Third, identification of transgender respondents in the BRFSS relies solely on a single-item measure rather than the recommended use of a two-step measure.^4,31

Finally, few states consistently used the BRFSS gender identity question over time. As such, we were unable to include a random slope that allows states to each have their own trajectory over time. This limits the scope of our inference to that of a cross-sectional association, precluding any interpretation of causality.

Conclusions

We found an association between higher structural stigma and lower odds of insurance coverage among transgender people. However, only 18% of the between-state variation was explained by our measure of structural stigma, and we did not find meaningful associations for the other three health care access indicators.

Taken together, these findings point to the need for future work to develop and validate measures of structural stigma experienced by transgender people that also consider factors such as the degree of policy implementation and the effects of structural racism. At the same time, our findings underscore the importance not just of assessing how structural stigma affects transgender people but also of developing public policy interventions that mitigate stigma by protecting transgender people from discrimination in health care and other areas of daily life.

Supplemental Material

Supplemental Material - State-level heterogeneity in associations between structural stigma and individual health care access: A multilevel analysis of transgender adults in the United States

Supplemental Material for State-level heterogeneity in associations between structural stigma and individual health care access: A multilevel analysis of transgender adults in the United States by Nguyen K Tran, Kellan E Baker, Elle Lett and Ayden I Scheim in Journal of Health Services Research & Policy

Footnotes

Author’s note

State Equality Index copyright 2020 Human Rights Campaign Foundation. All rights reserved. Reproduced with permission. Any further use without the express written consent of Human Rights Campaign is prohibited.

Declaration of conflicting interests

The author(s) declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.

Ethics approval

Ethics approval was not required for this original research article.

Funding

The author(s) received no financial support for the research, authorship, and/or publication of this article.

ORCID iDs

Nguyen K Tran

Elle Lett

Supplemental Material

Supplemental material for this article is available online.

References

Baker

. Findings from the behavioral risk factor surveillance system on health-related quality of life among US transgender adults, 2014-2017. JAMA Intern Med 2019; 179: 1141–1144.

Lett

Dowshen

Baker

. Intersectionality and health inequities for Gender Minority Blacks in the U.S. Am J Prev Med 2020; 59: 639–647.

Feldman

Luhur

Herman

, et al. Health and health care access in the US transgender population health (TransPop) survey. Andrology 2021; 9: 1707–1718.

Scheim

Baker

Restar

, et al. Health and health care among transgender adults in the United States. Annu Rev Public Health 2022; 43: 503–523.

Cruz

. Assessing access to care for transgender and gender nonconforming people: a consideration of diversity in combating discrimination. Soc Sci Med 2014; 110: 65–73.

White Hughto

Reisner

Pachankis

. Transgender stigma and health: a critical review of stigma determinants, mechanisms, and interventions. Soc Sci Med 2015; 147: 222–231.

Puckett

Cleary

Rossman

, et al. Barriers to gender-affirming care for transgender and gender nonconforming individuals. Sex Res Soc Pol 2018; 15: 48–59.

Lurie

. Identifying training needs of health-care providers related to treatment and care of transgendered patients: a qualitative needs assessment conducted in New England. Int J Transgend 2005; 8: 93–112.

Hatzenbuehler

Phelan

Link

. Stigma as a fundamental cause of population health inequalities. Am J Public Health 2013; 103: 813–821.

10.

Hatzenbuehler

. Structural stigma and health inequalities: research evidence and implications for psychological science. Am Psychol 2016; 71: 742–751.

11.

Seelman

Colón-DIaz

MJP

Lecroix

, et al. Transgender noninclusive healthcare and delaying care because of fear: connections to general health and mental health among transgender adults. Transgender Health 2017; 2: 17–28.

12.

Blosnich

Marsiglio

Gao

, et al. Mental health of transgender veterans in US States with and without discrimination and hate crime legal protection. Am J Public Health 2016; 106: 534–540.

13.

Goldenberg

Reisner

Harper

, et al. State policies and healthcare use among transgender people in the U.S. Am J Prev Med 2020; 59: 247–259.

14.

White Hughto

Murchison

Clark

, et al. Geographic and individual differences in healthcare access for U.S. Transgender adults: a multilevel analysis. LGBT Health 2016; 3: 424–433.

15.

Invited

Merlo J.

Commentary: multilevel analysis of individual heterogeneity--a fundamental critique of the current probabilistic risk factor epidemiology. Am J Epidemiol 2014; 180: 208–212.

16.

Center for Disease Control and Prevention . BRFSS questionnaires, https://www.cdc.gov/brfss/questionnaires/index.htm (2021, accessed 2 October 2021).

17.

Warbelow

Avant

Kutney

. 2019 state equality index, https://www.thehrcfoundation.org/professional-resources/state-equality-index-2019 (2020, accessed 2 October 2021).

18.

Kaiser Family Foundation . Status of State Medicaid Expansion Decisions: Interactive Map, https://www.kff.org/medicaid/issue-brief/status-of-state-medicaid-expansion-decisions-interactive-map/ (2021, accessed 2 October 2021).

19.

Carle

. Fitting multilevel models in complex survey data with design weights: recommendations. BMC Med Res Methodol 2009; 9: 49.

20.

Austin

Merlo

. Intermediate and advanced topics in multilevel logistic regression analysis. Stat Med 2017; 36: 3257–3277.

21.

Cicero

Reisner

Merwin

, et al. Application of behavioral risk factor surveillance system sampling weights to transgender health measurement. Nurs Res 2020; 69: 307–315.

22.

Lipsitch

Tchetgen

Cohen

. Negative controls: a tool for detecting confounding and bias in observational studies. Epidemiol 2010; 21: 383–388.

23.

R Core Team . R: a language and environment for statistical computing, https://www.r-project.org/ (2021, accessed 2 October 2021).

24.

Lumley

. Analysis of complex survey samples. J Stat Softw 2004; 9: 1–19.

25.

Bates

Mächler

Bolker

, et al. Fitting linear mixed-effects models using lme4. J Stat Softw 2015; 67: 1–48.

26.

Mallory

Tentindo

. Medicaid coverage for gender-affirming care, https://williamsinstitute.law.ucla.edu/wp-content/uploads/Medicaid-Gender-Care-Oct-2019.pdf (2019, accessed 2 October 2021).

27.

Scheim

Coleman

Lachowsky

, et al. Health care access among transgender and nonbinary people in Canada, 2019: a cross-sectional survey. CMAJ Open 2021; 9: E1213–E1222.

28.

Crenshaw

. Demarginalizing the intersection of race and sex: a Black Feminist Critique of Antidiscrimination Doctrine, feminist theory and antiracist politics. Univ Chic Leg Forum 1989; 1989: 139–167.

29.

Rao

Elshafei

Nguyen

, et al. A systematic review of multi-level stigma interventions: state of the science and future directions. BMC Med 2019; 17: 41.

30.

Lett

Everhart

. Considerations for transgender population health research based on US national surveys. Ann Epidemiol 2022; 65: 65–71.

31.

Bauer

Braimoh

Scheim

, et al. Transgender-inclusive measures of sex/gender for population surveys: Mixed-methods evaluation and recommendations. PLoS One 2017; 12: e0178043.

Supplementary Material

Please find the following supplemental material available below.

For Open Access articles published under a Creative Commons License, all supplemental material carries the same license as the article it is associated with.

For non-Open Access articles published, all supplemental material carries a non-exclusive license, and permission requests for re-use of supplemental material or any part of supplemental material shall be sent directly to the copyright owner as specified in the copyright notice associated with the article.

0.00 MB

0.67 MB