State-Level Policy Stigma and Non-Prescribed Hormones Use among Trans Populations in the United States: A Mediational Analysis of Insurance and ...
←
→
Page content transcription
If your browser does not render page correctly, please read the page content below
ann. behav. med. (2021) XX:1–13
https://doi.org/10.1093/abm/kaab063
REGULAR ARTICLE
State-Level Policy Stigma and Non-Prescribed Hormones Use
among Trans Populations in the United States: A Mediational
Analysis of Insurance and Anticipated Stigma
Downloaded from https://academic.oup.com/abm/advance-article/doi/10.1093/abm/kaab063/6352463 by guest on 17 August 2021
Landon D. Hughes, BA,1,4, ∙ Kristi E. Gamarel, PhD, EdM1,4, ∙ Wesley M. King, EdM, MPH1,4 ∙ Tamar Goldenberg,
PhD, MPH2, ∙ James Jaccard, PhD3 ∙ Arline T. Geronimus, ScD1,4
Published online: 14 August 2021
Published by Oxford University Press on behalf of the Society of Behavioral Medicine 2021. This work is written by (a) US Government
employee(s) and is in the public domain in the US.
Abstract
Background Medical gender affirmation (i.e., hormone Results Among trans adults using hormones, we found
use) is one-way transgender (trans) people affirm their that healthcare policy stigma was positively associ-
gender and has been associated with health benefits. ated with NPHs use and operated through insurance
However, trans people face stigmatization when ac- coverage and anticipating stigma in healthcare settings.
cessing gender-affirming healthcare, which leads some to The effect sizes on key predictor variables varied signifi-
use non-prescribed hormones (NPHs) that increase their cantly between those who use supplemental NPHs and
risk for poor health. those who only use NPHs suggesting the need to treat
Purpose We examined whether healthcare policy stigma, NPHs use as distinct from those who use supplemental
as measured by state-level trans-specific policies, was as- NPHs.
sociated with NPHs use and tested mediational paths Conclusions Our work highlights the importance of
that might explain these associations. Because stigma- healthcare policy stigma in understanding health in-
tizing healthcare policies prevent trans people from equities among trans people in the USA, specifically
participation in healthcare systems and allow for dis- NPHs use.
crimination by healthcare providers, we hypothesized
that healthcare policy stigma would be associated with
Keywords Transgender ∙ Hormone use ∙ Gender
NPHs use by operating through three main pathways:
affirmation ∙ Structural stigma ∙ Policy ∙ Insurance
skipping care due to anticipated stigma in healthcare set-
tings, skipping care due to cost, and being uninsured.
Methods We conducted analyses using data from the Introduction
2015 U.S. Transgender Survey. The analytic sample in-
cluded trans adults using hormones (N = 11,994). We Gender affirmation is the social process by which one’s
fit a multinomial structural equation model to examine gender identity, expression, or role is recognized and af-
associations. firmed [1]. Transgender (trans) individuals, including
gender nonbinary people, experience gender affirmation
in many ways. Gender affirmation is comprised of four
Landon D. Hughes
landonh@umich.edu different but interconnected dimensions: social, psy-
chological, legal, and medical [2]. Specifically, Reisner
1
School of Public Health, University of Michigan, 1415 et al. [2] describe social affirmation as interpersonal rec-
Washington Heights, Ann Arbor, MI 48109, USA ognition (e.g., using the correct name and pronouns),
2
Carolina Population Center, University of North Carolina at psychological affirmation as the internal felt sense of
Chapel Hill, Chapel Hill, NC, USA self-actualization (e.g., validation of self), legal affirm-
3
Silver School of Social Work, New York University, New ation as the recognition by legal systems of one’s gender
York, NY, USA (e.g., legal name and gender marker changes), and med-
4
Institute for Social Research, University of Michigan, Ann ical affirmation as the use of medical technologies to af-
Arbor, MI, USA firm one’s gender (e.g., hormones, gender affirmation2 ann. behav. med. (2021) XX:1–13
surgery, and puberty blockers). The majority of research Stigma as a Social Determinant of Health
has focused on medical gender affirmation, specifically
hormone use [3, 4]. Hormone use, like other forms of There has been an increasing recognition that stigma
gender affirmation, has been associated with a range of is a fundamental cause of population health inequities
positive health outcomes, including reductions in sui- among trans populations [10, 20, 21]. Hatzenbuehler
cidal ideation, binge-drinking, drug use, anxiety, and de- et al. [22] define stigma as “the cooccurrence of labeling,
pression and an increase in quality of life among trans stereotyping, separation, status loss, and discrimination
people who use them for medical gender affirmation in a context in which power is exercised.” In the USA,
[4–7]. the dominant and pervasive ideology on gender is that
Downloaded from https://academic.oup.com/abm/advance-article/doi/10.1093/abm/kaab063/6352463 by guest on 17 August 2021
Hormone use needs and receipt vary within trans men and women are biologically distinct and inherently
populations. For example, 20% of participants in a possess certain psychological and behavioral traits de-
large national survey indicated they did not want hor- rived from reproductive functions [23]. This ideology
mones, and among those who wanted hormones, only conflates gender with sex, creating what we refer to as
half had ever accessed them [8]. Notably, many people the gender/sex fallacy [23]. The gender/sex fallacy alien-
are unable to access hormones from a licensed med- ates people whose gender identity or expression are dis-
ical professional and turn to non-prescribed hormones cordant with the gender typically aligned with their sex
(NPHs). Not being able to access prescribed hormones assigned at birth, or whose gender identity or expression
(PHs) can force people to go without or to access does not align with the man-woman binary. Further, the
NPHs by purchasing them online, obtaining them from gender/sex fallacy provides a rationale for stigmatization,
friends, or acquiring them via some other non-licensed promoting the discrimination and stereotyping of trans
source [9–11]. Furthermore, merely having access to a people [10].
doctor does not guarantee access to hormones, as doc- The majority of research regarding stigma in trans
tors may refuse to prescribe hormones, insurance may populations has focused on interpersonal or individual
refuse to cover hormone prescriptions, or people may forms of stigma, such as victimization (e.g., physical or
be unable to afford hormones due to out-of-pocket emotional abuse a trans person encounters), internal-
costs or lack of insurance [8]. Moreover, structural ized stigma (e.g., internalizing negative societal messages
stigma may affect the availability of hormones by about oneself as a trans person), or anticipating and
operating as an impediment to accessing PHs, which is avoiding stigma (e.g., the presumption that one might be
discussed below [12]. victimized and avoids instances where victimization may
Given the aforementioned barriers, trans people have be a threat) [21, 24]. While interpersonal and individual
developed alternative ways to access the healthcare they stigma is critical to understanding the health of trans
need, including hormones; however, some of these al- people, these are not the only means by which stigma im-
ternatives may be risky [13]. Access to PHs is important pacts health. White Hughto et al. [10] argue that we must
because NPHs significantly increase the risk of poor consider how stigma operates across multiple levels,
health outcomes due to improper dosing and the lack including structural forms of stigma, such as policies
of monitoring [14, 15]. While the long-term effects of that limit the resources, opportunities, and wellbeing of
any hormone use are unclear, some studies have shown trans people. For example, stigmatizing policies may act
an increased risk for adverse cardiometabolic indicators as structural impediments that constrain trans peoples’
after beginning hormone therapy [16]; therefore, the cur- access to hormones by mandating that Medicaid cannot
rent medical guidelines recommend that doctors closely cover trans-related care, even if a doctor deems medical
monitor their patients’ cardiometabolic health while interventions necessary [25]. Furthermore, religious ex-
taking hormones [17]. For example, some formulations emption laws allow doctors to deny trans people any
of oral estrogen increase the risk of venous thrombo- healthcare services so long as they claim this exemption
embolism and are therefore no longer prescribed by most [26]. Religious exemption laws not only affect access to
clinicians; however, trans people who use non-prescribed hormones but also to any healthcare service for trans
estrogen often take high dosages of these formulations, people. Together, these policies result in healthcare policy
increasing their risk for venous thromboembolism [18]. stigma, which we conceptualize as stigma resulting from
Furthermore, some people may use high doses of NPHs policies that govern healthcare systems and demean, de-
in conjunction with PHs because they believe this will value, and restrict the healthcare of trans people.
achieve faster results, placing them at risk of adverse Thus, healthcare policy stigma is a specific form of
health effects [15]. Researchers have also speculated that structural stigma that may constrain the ability of trans
NPHs may increase the risk of HIV infection due to people to access care that meets their gender affirmation
sharing needles or parenteral administration, although needs by operating through two pathways: anticipated
no study has formally linked these two [19]. stigma and cost. Healthcare policy stigma may allowann. behav. med. (2021) XX:1–133
violence and discrimination in medical settings to go un-
checked, increasing individuals’ fear or anticipation of
encountering stigma in healthcare contexts and driving
healthcare avoidance [27]. Additionally, healthcare
policy stigma may increase the out-of-pocket cost for ac-
cessing hormones by allowing insurers to refuse to cover
hormone-related care. Lastly, healthcare policy stigma
may influence trans people’s insurance rates as some may
choose not to participate in a healthcare system that is
Downloaded from https://academic.oup.com/abm/advance-article/doi/10.1093/abm/kaab063/6352463 by guest on 17 August 2021
not built to meet their needs [13]. Thus, healthcare policy
stigma may be a critical factor for understanding why
Fig. 1. Multinomial model predicting non-prescribed hormone
people use NPHs.
use. Note: Model controlled for gender identity, race/ethnicity,
age, education, Census region, unemployment, sex work, physical/
Purpose and Hypotheses verbal abuse, engagement with other trans people, experiencing
homelessness, and family support. Medicaid expansion was in-
cluded as a control when predicting uninured.
The purpose of this paper is to examine whether
healthcare policy stigma is associated with using NPHs
and test possible mediational pathways in a sample of and skipping care due to cost will increase the likelihood
trans people who use hormones. Previous research of both supplemental NPHs use and only using NPHs
demonstrates associations between state-level policies compared to those who only use PHs.
and health among transgender populations [20, 28,
29] including findings that demonstrate that state-level
policy stigma is associated with decreases in hormone use Materials and Methods
for medical gender affirmation [12]. However, this study
builds on this work to demonstrate, for the first time, the
This study is a secondary data analysis of the 2015 U.S.
mechanisms through which state-level policy stigma may
Transgender Survey (USTS), conducted among a na-
work to influence NPHs use. Importantly, the literature
tional sample of trans people in the USA and spon-
on NPHs has predominantly treated NPHs use as a di-
sored by the National Center for Transgender Equality
chotomous outcome: any NPHs use versus no NPHs use
[8]. Data were collected in August and September of
[9, 30]. Simply treating NPHs use as a dichotomous out-
2015. The National Center for Transgender Equality
come may not capture people who supplement their PHs
worked with over 400 organizations across the USA to
with NPHs, suggesting a third group [15]. Given that risk
recruit nearly 28,000 respondents via social media and
factors for only using NPHs may be different from those
email. While these data were collected in 2015, they re-
who supplement their PHs with NPHs, this paper seeks
main the largest source of information on NPHs use in
to understand whether healthcare policy stigma is differ-
trans populations in the USA. Surveys were completed
entially associated with exclusive NPHs use or supple-
on web-enabled devices (e.g., computers, tablets, and
mental NPHs use.
smartphones) and were made accessible to respondents
We posit that healthcare policy stigma operates
with disabilities using screen readers. Surveys were avail-
through two pathways to contribute to any form of
able in English and Spanish. For more information on
NPHs use: skipping care due to cost and anticipating
methods, see the 2015 USTS report [8]. The original data
stigma. Figure 1 presents the conceptual model to be
collection was approved by the University of California
tested. First, we hypothesize that living in a state with
Los Angeles Institutional Review Board and the sec-
high levels of healthcare policy stigma will be associated
ondary analyses were ruled exempt by the University of
with skipping care due to anticipated stigma and cost,
Michigan Institutional Review Board.
which will increase the chances of using supplemental
NPHs and using only NPHs compared to those who
only use PHs. Sample
Second, we hypothesize that healthcare policy stigma
will be associated with a higher probability of being The National Center for Transgender Equality re-
uninsured and skipping needed healthcare due to cost. cruited 27,715 people for the project. Eligibility for the
Trans people may be less likely or able to participate in project included (a) identifying as trans or some other
a healthcare system that allows for discrimination and gender-diverse individual, (b) being at least 18 years
will be more likely to pay out of pocket for their care. of age, and (c) living in the USA. We then limited our
We posit that, in turn, increases in the uninsured rate analytic sample to those who reported currently using4 ann. behav. med. (2021) XX:1–13
hormones (n = 12,044). Respondents who identified as composite are (a) private insurance protections for trans
cross-dressers were removed from the sample because people, (b) whether or not Medicaid covers trans-specific
their experiences are fundamentally different than those healthcare, (c) state-wide nondiscrimination protections,
with other trans identities (n = 20). Respondents who and (d) religious exemption laws. This measure is adapted
lived on a military base or one of the U.S. territories from Goldenberg et al.’s state-level trans-specific policies
at the time of data collection were also removed from measure [12, 31]. These four policies were chosen because
the sample because we could not calculate a healthcare they are relevant to healthcare utilization in that they ei-
policy stigma score for these areas and the means of ac- ther stigmatize trans people, restrict access to healthcare
cessing hormones on a military base is different than in services, or provide legal protections in healthcare set-
Downloaded from https://academic.oup.com/abm/advance-article/doi/10.1093/abm/kaab063/6352463 by guest on 17 August 2021
the rest of the USA (n = 30). Our final overall sample tings. We gave supportive policies a score of minus one,
included 11,994 respondents. while we gave unsupportive policies a score of plus one,
while states that did not have an explicit policy were held
Measures unchanged. In total, the potential scores range from −2
to 2; however, the observed ranges for this variable in
NPHs use the data were −1 to 2. Higher scores indicate states with
stigmatizing policies toward trans people. One point was
Current NPHs use was coded into three nominal subtracted from a state’s score if that state had private
categories: currently using PHs, supplemental NPHs, insurance nondiscrimination policies or if that state had
and NPHs. Respondents were asked “Where do you cur- a state-wide nondiscrimination policy. States were given
rently get your hormones?” and selected one of three one point if that state restricted trans healthcare for
responses. Those who chose “I only go to licensed pro- Medicaid populations or had any religious exemption
fessionals (like a doctor) for hormones” were coded as laws. For a map of state-specific values for this variable,
PHs only. Those who chose “In addition to licensed pro- see Fig. 2.
fessionals, I also get hormones from friends, online, or
other non-licensed sources” were coded as supplemental
NPHs. And those who chose “I ONLY get hormones Mediators
from friends, online, or other non-licensed sources” were Skipped care due to anticipated stigma was coded as a
coded as NPHs only. dichotomous variable (i.e., “Was there a time in the past
12 months when you needed to see a doctor but did not
Healthcare policy stigma
because you thought you would be disrespected or mis-
Healthcare policy stigma is a cumulative measure of the treated as a trans person?”). Anyone who indicated “yes”
severity of policy-level factors that demean, devalue, and to the question was coded as one, while those who in-
restrict the care of trans people. We created the state- dicated “no” were coded as zero. Uninsured was coded
specific healthcare policy stigma variable by tallying the as a dichotomous variable, with those having no form
total number of policies that were supportive of trans of insurance (e.g., private insurance, Medicaid, and
people and those that were unsupportive in 2015, the Medicare) being coded as one and those with any insur-
year data were collected. The policies underlying this ance being coded as zero. Skipped care due to cost was
Fig. 2. Map of state-specific policy stigma values.ann. behav. med. (2021) XX:1–135
coded as a dichotomous variable (i.e., “Was there a time one for those who reported socializing with trans people
in the past 12 months when you needed to see a doctor in person and zero for those who reported not social-
but could not because of cost?”). Anyone who indicated izing with trans people in person. Family support was
“yes” to the question was coded as one, while those who coded into three categories: (a) those who are not out
indicated “no” were coded as zero. to their family, (b) those who reported their family was
unsupportive of their gender identity, and (c) those who
Covariates reported either not having a family, having a supportive
family, or a family that was neither supportive nor un-
While reporting current gender identity, respondents
supportive. Although there may be important differences
chose one of six options: cross-dresser, woman, trans-
Downloaded from https://academic.oup.com/abm/advance-article/doi/10.1093/abm/kaab063/6352463 by guest on 17 August 2021
in the third category of the family support variable, the
gender woman, man, transgender man, or nonbinary/
sample sizes were too small to analyze these groups sep-
genderqueer. We excluded respondents who chose
arately. Because a lack of family support and disclosure
“cross-dresser” and created a three-level variable: (a)
have both been associated with adverse outcomes, we
trans woman/woman; (b) trans man/man; and (c) gender
coded this variable to examine differences between those
nonbinary/genderqueer. Given the small number of per-
with negative family experiences and those who were not
sons of color (n = 2,063), the race was coded as a di-
out to their family and compared them participants with
chotomous variable for those who identify as “white”
more neutral or positive family experiences [34].
and those who identified as a person of color (i.e., 5%
To control for variation in NPHs use resulting from
Hispanic, 5% Biracial, 3% Black, 2% Asian/Pacific
state-level and geographic factors other than healthcare
Islander, and 1% Native American). While this approach
policies, we included census region and Medicaid expan-
is not ideal, the small cell sizes for NPHs use when cross-
sion as covariates. Census region was used to group states
tabulated by race made it impossible to include the
together by geographical location based on the tax-
multi-category covariate. Age was collected and used as
onomy used by the Census that classifies states into ei-
a continuous variable age in years at the time of data
ther the Midwest, Northeast, South, or West. Although
collection. Unemployment was coded as a dichotomous
imperfect, we included the Census region as a control
variable with those who were currently unemployed but
because states in similar regions tend to have similar pol-
looking for work being coded as one and all else being
itical and social climates. Medicaid expansion was one
coded as zero. The Highest level of education was coded
provision of the Affordable Care Act aimed at reducing
into four categories: less than high school, high school
the uninsured rate. This statute allowed for states to opt
graduate, some college, and college graduate.
into increasing the number of people eligible to receive
In addition to cost, lack of insurance, and anticipated
Medicaid in exchange for more federal funding [35].
stigma [9, 13, 27, 32], prior studies have also shown that
Medicaid expansion has been shown to significantly de-
NPHs use is correlated with lifetime sex work, experi-
crease the uninsured rates in states that have expanded
encing homelessness, verbal or physical victimization,
Medicaid [35]. We controlled for Medicaid expansion
having a network of other trans people who use hor-
using a categorical variable identifying whether a state
mones, and family rejection [9, 13, 19, 27, 30, 33]. To
had expanded Medicaid before the data were collected;
control for these additional factors, we relied on meas-
states were coded as one if they expanded Medicaid and
ures collected by the USTS that mapped onto these
zero if they did not.
constructs. Consistent with prior studies using this data
source to examine medical gender affirmation [31], re-
spondents were asked whether they had ever engaged in Statistical Analyses
sex or sexual activity for money or worked in the sex in-
dustry, such as erotic dancing, webcam work, or porn We tested the conceptual model outlined in Fig. 1
films. Individuals who responded “yes” were coded as (covariates are omitted to reduce clutter). The model
one for the variable sex work and zero if they responded was evaluated using the Mplus 8.0 software for struc-
“no.” The variable experiencing homelessness was coded tural equation modeling. The model was fit using robust
as one for those respondents who reported experiencing (Huber-White) maximum likelihood algorithms. The un-
homelessness in the past year and zero for those who re- insured skipped care due to cost, and skipped care due to
ported “no.” Respondents reported whether they experi- anticipated stigma mediators are dichotomous and were
enced physical or verbal abuse due to their gender identity estimated using a logit function. NPHs use was treated
in the past year. The variable physical or verbal abuse was as a three-level nominal outcome that was regressed
coded as one for those who had reported experiencing onto all variables, except Medicaid expansion, using a
either physical or verbal abuse due to their gender iden- multinomial logit function with numerical integration.
tity in the past year and zero for those who reported they The referent group for the multinomial equation was
did not experience either. Trans engagement was coded as those who only use NPHs. Multinomial equations yield6 ann. behav. med. (2021) XX:1–13
coefficients that estimate local odds whereas our interest our focus was on contrasts between substantively mean-
was with marginal probabilities for each of the three ingful predictor profiles.
NPHs use categories. We used the methods described
in Muthén, Muthén, and Asparouhov [36] to estimate
the relevant marginal probabilities where all covariates Results
were held constant at their respective mean values (i.e.,
we used a form of marginal effects analysis at the mean), Table 1 presents unadjusted tabulations of demo-
but where the component probabilities of the marginal graphics by hormone use. Among the respondents,
effects analysis at the mean were used to form relative 11,004 (92%) currently accessed hormones only from a
Downloaded from https://academic.oup.com/abm/advance-article/doi/10.1093/abm/kaab063/6352463 by guest on 17 August 2021
risk ratios rather than probability differences using the licensed doctor (PHs use), 255 (2%) currently accessed
MODEL CONSTRAINT command in Mplus. hormones only from some other source (NPHs use),
The initial fit of the model revealed global ill-fit due and 735 (6%) accessed hormones from both a licensed
to the need for correlated disturbances between skipping doctor and some other source (supplemental NPHs
care due to trans stigma and the other two mediators. We, use). Without adjusting for covariates, on average, as
therefore, added parameters to the model to reflect these age increased individuals were slightly more likely to use
covariances. No other localized sources of model ill fit NPHs. Compared to white people, people of color were
were noted. Missingness was not a major issue with these slightly more likely to use either supplemental NPHs or
data. Missing data were treated using the default full in- only NPHs. On average, those with higher levels of edu-
formation maximum likelihood methods in Mplus. Data cation were less likely to use either supplemental NPHs
were missing for the variables sex work (n = 9), skipped or only NPHs. Compared to trans women/women and
care due to anticipated stigma (n = 13), trans engagement nonbinary/genderqueer individuals, trans men/men were
(n =6), currently experiencing homelessness (n = 53), un- significantly less likely to use both supplemental NPHs
insured (n = 30), skipped care due to cost (n = 45), and and only NPHs. Table 2 reports the targeted predictor
family support (n = 22). profile contrasts. We discuss each set of contrasts, in
We report the results using profile analyses where we turn. To view the full results from the structural equation
varied selected values on key predictors while holding all model, see the Online Supplement.
other variables constant at their mean values. The ad-
vantage of this approach is that it allows us to focus on
probabilities and relative risk ratios, which are more in- Predicting the Probability of Being Uninsured
terpretable and less misleading than odds ratios. The esti-
mation algorithms do not permit the estimation of total On average, the uninsured rate was an estimated 4.5%
effects from traditional structural equation modeling, so lower in states that expanded Medicaid compared to
Table 1 Respondent Demographics by Hormone Use
PHs only Supplemental NPHs NPHs only
(n = 11,004) (n = 735) (n = 255)
n or M % or SD n or M % or SD n or M % or SD Significance
Age (in years) 35 14 35 13 38 14 F(2, 11,991) = 6.09; p = .002
Race
White 9,154 92% 576 6% 201 2% x2(2) = 14; p = .001
People of color 1,850 90% 159 8% 54 3%
Educationann. behav. med. (2021) XX:1–137 Table 2 Profile Analyses: Direct Effects Profile contrast Profile 1 probability Profile 2 probability Relative risk p values Outcome: uninsured ME(no) vs. ME(yes) 0.118 0.073 .616 (.521, 712)
8 ann. behav. med. (2021) XX:1–13
stigma had any significant direct effect on the probability in using supplemental NPHs and only using NPHs as
of only using PHs. operating through insurance coverage. Similarly, given
that healthcare policy stigma was statistically associ-
Predicting the Probability of Supplemental NPH Use ated with an increase in skipping care due to anticipating
stigma and skipping care due to anticipating stigma was
We found that those who were uninsured were more statistically associated with an increase in supplemental
likely to use supplemental NPHs than their insured NPH use and only using NPHs, under the property of
counterparts: 7% to 5% respectively (p = .001). Those the joint significance test, healthcare policy stigma was
who skipped care due to anticipated stigma in healthcare associated with an increase in using supplemental NPHs
Downloaded from https://academic.oup.com/abm/advance-article/doi/10.1093/abm/kaab063/6352463 by guest on 17 August 2021
settings were more likely to use supplemental NPHs than and only using NPHs as operating through anticipated
their counterparts who did not skip care due to antici- stigma. Lastly, given that healthcare policy stigma was
pated stigma: 8% to 4% respectively (p < .001). Those statistically associated with an increase of skipping care
who skipped care due to cost were more likely to use sup- due to cost and skipping care due to cost was statistic-
plemental NPHs than their counterparts who did not: ally associated with an increase in supplemental NPHs
7% to 4% respectively (p < .001). However, this did not use, under the property of the joint significance test,
reach statistical significance when analyzing local odds healthcare policy stigma was associated with an increase
(p = .079); thus, the results should be interpreted with in using supplemental NPHs as operating through an-
caution. Lastly, we found that healthcare policy stigma ticipated stigma. Again, this last mediational chain was
had a negative direct effect on the probability of supple- not statistically significant when analyzing local odds
mental NPHs, although statistical significance remained (p = .079). The pathway from healthcare policy stigma to
suspect (Relative Risk Ratio = .797, p = .077). using only NPHs was not significant.
Table 3 reports the predicted probabilities for the cu-
mulative effect of the best versus the worst outcomes
Predicting the Probability of Using Only NPHs from the full multinomial model using profile analyses.
The probabilities for the best-case group is the esti-
Those who were uninsured were more likely to only use mated probability of PHs use only, supplemental NPHs
NPHs than their insured counterparts: 4% to 0.7% re- use, and NPHs use only when the control variables are
spectively (p < .001). Those who skipped care due to an- mean-centered and the pathway variables are set to their
ticipated stigma in healthcare settings were more likely to most favorable values (e.g., uninsured, skipping care due
only use NPHs than their counterparts who did not skip to stigma, and skipping care due to cost are all equal to
care due to anticipated stigma: 1.4% to 0.7% respect- 0; Medicaid expansion is set to 1, and healthcare policy
ively (p < .001). We did not find that those who skipped stigma is set to −1). The probabilities for the worst-case
care due to cost were statistically more or less likely to group are the estimated probability of using PHs only,
only use NPHs than their counterparts who did not skip supplemental NPHs use, and NPHs use only when the
care due to cost. Lastly, we found a direct effect from control variables are mean-centered and the pathway
healthcare policy stigma to only using NPHs. Those in variables are set to their least favorable values (e.g.,
states with the greatest healthcare policy stigma were reverse-scored values from above).
more likely to only use NPHs than those in states with We found that the best-case probabilities were posi-
the least healthcare policy stigma: 1.1% to 0.7% respect- tively associated with desired outcomes and negatively
ively (p = .036). associated with undesirable outcomes. Of particular
note, compared to the best-case scenario, the worst-case
Testing the Mediational Chains From Healthcare Policy scenario showed an 18-fold increase in the probability of
Stigma to Hormone Use Type using NPHs only (0.5% to 8%) and a 3-fold increase in
using supplemental NPHs (4% to 13%). Each of these
The pattern of results for the profile analyses implies findings was statistically significant below a p value of
statistically significant mediation effects using the logic .001.
of joint significance tests as described in Fritz and
MacKinnon [37]; Fritz et al. [38]. For example, given
that healthcare policy stigma was statistically associ- Discussion
ated with an increase in the uninsured rate and being
uninsured was, in turn, statistically associated with Our findings are consistent with other studies that dem-
an increase in supplemental NPHs use and only using onstrate that structural stigma, specifically healthcare
NPHs, under the property of the joint significance test, policies, is associated with medical gender affirmation
healthcare policy stigma was associated with an increase practices of trans people in the USA [4, 31]. Our studyann. behav. med. (2021) XX:1–139
Table 3 Profile Analyses: Best- Versus Worst-Case
Profile contrast Best-case probability Worst-case probability Relative risk p values
Outcome: uninsured
0.038 0.117 3.10 (2.504, 3.696)10 ann. behav. med. (2021) XX:1–13
model does not account for all possible mediators be- Existing studies suggest the need for policies that
tween healthcare policy stigma and using only NPHs. protect trans people from discrimination in healthcare
One mediator that may be relevant to understanding the settings given the evidence that discrimination against
effect of healthcare policies on the use of NPHs use is trans people in healthcare settings is related to adverse
access to PHs. While the USTS does not specifically as- physical and mental health outcomes [43]. Our research
sess factors associated with accessing PHs, such as the builds on this work to demonstrate the importance of
ability to access a pharmacy, others have documented trans-specific public policy to not only addressing NPHs
inconsistent access to trans-competent pharmacological use but also to insuring trans people. In June 2020, the
care amongst trans individuals that may be influential in Department of Health and Human Services finalized a
Downloaded from https://academic.oup.com/abm/advance-article/doi/10.1093/abm/kaab063/6352463 by guest on 17 August 2021
understanding NPHs use [39]. Exploring potential me- rule that removed existing protections from healthcare
diators between healthcare policy stigma and using only discrimination for an estimated 1.4 million trans adults
NPHs may prove an important topic for future research. [44, 45]. This rule allows healthcare facilities, insurers,
Mixed-methods research may be particularly useful in and providers to deny care to trans people simply be-
exploring potential mediators (e.g., interpersonal inter- cause of their gender identity [45].
actions and intrapersonal factors such as cognitions, Our findings suggest that it is plausible a lack of trans-
preferences, and behaviors) that might be driving the use inclusive healthcare policies may increase the number of
of NPHs only as opposed to supplemental NPHs use. trans people who are uninsured, skip care due to stigma
A notable finding we had not hypothesized was how and cost, and who use NPHs. Trans-inclusive policies
the association between healthcare policy stigma and the that guarantee adequate access to safe, effective hor-
probability of being uninsured would compare to that mones are crucial to ensuring health equity for trans
of Medicaid expansion. Remarkably, compared to states people. While documenting the potential effects of
with the most healthcare policy stigma, states with the harmful policies is an important step, it is by no means
least healthcare policy stigma have a lower predicted the last. Public health practitioners must work to create
uninsured rate of 3.5%; while states that have passed interventions that meaningfully reduce structural stigma
Medicaid expansion have a lower predicted uninsured and build political coalitions to enact policies that pro-
rate of 4.5%. This finding suggests that trans-specific tect trans people.
healthcare policies are nearly as influential at insuring Beyond practical implications, this study also suggests
trans people than gender-blind policies like Medicaid a few implications for researchers working with categor-
expansion. Our findings suggest healthcare systems, ical variables and cross-sectional data. Chiefly, our find-
including state policies, that are not explicitly designed to ings highlight the importance of thinking critically about
protect trans people (e.g., do not cover gender-affirming how to operationalize categorical variables. Researchers
care or protect from discrimination and victimization) ought to carefully examine their categorical variables
may result in avoidance of care or trans people may be and exhaust combinations relevant to their research
shut out from participation. This finding supports prior topic. Our analyses also showcase the ability to conduct
research by Glick et al. [13] that trans people often go preliminary mediational research in a cross-sectional
outside of mainstream healthcare services for their care, setting. While we are unable to “prove” causality in this
like accessing hormones from non-licensed sources, if study due to its cross-sectional nature, we were able to
they are not supported by mainstream institutions. test whether our conceptual model is plausible given our
Notably, we found that nonbinary adults were at data. This is an important first step in examining our con-
higher risk for NPHs use compared to trans men and ceptual model and testing our hypotheses, especially in
trans women. It is plausible that nonbinary individ- situations where it is unethical to conduct cause-probing
uals may turn to NPH because the current World studies (e.g., randomized control trials). In this way,
Professional Association for Transgender Health cross-sectional mediational analyses allow for testing the
(WPATH) Standards of Care may be too restrictive and plausibility of mediational models without the unethical
reinforce normative binary conceptualizations of gender methods required to “proving” them.
conceptualization of gender identity and expression [40].
WPATH’s Standards of Care are currently being updated Limitations
to be more inclusive of nonbinary patients [41]. It is also
plausible that providers may not have knowledge or These findings should be interpreted within the context
competency regarding proper care for nonbinary people, of the following limitations. The USTS is a conveni-
which may reinforce binary conceptualizations of gender ence sample, which limits generalizability. The study
[42]. These findings suggest future research is warranted also relies on self-reported data of sensitive topics (e.g.,
to better understand NPHs use among nonbinary indi- NPHs use, sex work) such that there may be social de-
viduals to help inform clinical practice and training. sirability bias. Furthermore, there is reason to believeann. behav. med. (2021) XX:1–1311
the number of people reporting NPHs use and being in- NPHs use as at least three categories: PHs use only, sup-
sured may be smaller in our sample than in the overall plemental NPHs use, and NPHs use only.
population given that the study recruited some partici- Our findings also demonstrate the importance of
pants via medical centers; thus, these individuals may trans-inclusive policies and insurance coverage among
be actively engaging in mainstream healthcare settings. trans populations. We found that stigmatizing pol-
Furthermore, the USTS sample is predominantly non- icies were associated with an increase in the likelihood
Hispanic white, which made it impossible to conduct of trans people being uninsured. This suggests that to
analyses on specific racial and ethnic categories. The lower the uninsured rates of trans people, states cannot
small number of people of color also limited our ability simply enact gender-blind policies aimed at insuring en-
Downloaded from https://academic.oup.com/abm/advance-article/doi/10.1093/abm/kaab063/6352463 by guest on 17 August 2021
to conduct interaction analyses between race/ethnicity tire populations, such as Medicaid expansion, but must
and gender identity groups. When predicting the prob- also consider trans-specific protections. Finally, this
ability of supplemental NPHs use, some have shown study also connects individual-level forms of stigma,
that body satisfaction, or how happy a person is with such as avoiding healthcare services due to fear of dis-
their physical body, may be a key indicator of supple- crimination, with structural forms of stigma, such as
mental NPHs use [46]. Our inability to control for in- states’ healthcare policy environments. Policies that
dividuals’ body satisfaction maybe masking differences stigmatize trans people are highly associated with how
or acting as a confounder in our analyses. Future re- trans people navigate healthcare; the more stigmatizing
search should consider other risk factors for those who a state’s policies are, the more likely trans people may
use supplemental NPHs: this is particularly important be to go without the care they need. In this way, policies
given that, at least in this sample, more individuals use “get under the skin” because they may lead people to use
supplemental NPHs than rely on NPHs alone. Logistic NPHs, which can have serious consequences for their
and multinomial logistic modeling have limitations in health. Future research using longitudinal designs must
that the estimated effects are dependent on values of the consider the limits of trans individuals’ health behaviors
covariates given the nonlinear nature of the modeling in the presence of pernicious forms of structural stigma,
[47]. The generalizability of the results must therefore be such as stigmatizing healthcare policies, that constrain
viewed cautiously. Additionally, formal mediational and their ability to access safe hormones.
total effects analysis is difficult with nominal outcomes
and dichotomous mediators. Future research would
benefit from developing continuous measures of these Supplementary Material
constructs for use in more traditional structural equa-
tion models, such as how often a participant uses NPHs. Supplementary material is available at Annals of Behavioral
Lastly, as the data are cross-sectional, causality cannot Medicine online.
be determined from this study.
Acknowledgments Landon Hughes was supported by the
Rackham Merit Fellowship, the National Institute on Aging (T32
AG000221), and the Eunice Kennedy Shriver National Institute
Conclusions of Child Health and Development (T32 HD00733931). We thank
the U.S. Trans Survey (USTS) team and all of the individuals who
These findings demonstrate a pathway from healthcare participated in the study.
policy stigma to NPHs use, with more inclusive policies
being protective against NPHs use. We found that this Compliance With Ethical Standards
association is partially mediated by insurance coverage,
Authors’ Statement of Conflict of Interest and Adherence to Ethical
skipping care due to anticipated stigma, and skipping Standards Authors Landon D. Hughes, Kristi E. Gamarel, Wesley
care due to cost. However, our research also demon- M. King, Tamar Goldenberg, James Jaccard, and Arline T.
strates that these mediational factors vary in import- Geronimus declare that they have no conflict of interest. All pro-
ance when predicting supplemental NPHs use versus cedures, including the informed consent process, were conducted
predicting NPHs use only. This highlights the import- in accordance with the ethical standards of the responsible com-
mittee on human experimentation (institutional and national) and
ance of tailoring interventions to address the specific with the Helsinki Declaration of 1975, as revised in 2000.
needs of trans people who are using NPHs. For example,
interventions that focus on those using supplemental
NPHs may be best served by focusing on anticipated References
stigma, while interventions focusing on those who only
use NPHs need to consider how being uninsured limits 1. Sevelius JM. Gender affirmation: a framework for conceptu-
one’s ability to access PHs. Due to these kinds of dif- alizing risk behavior among transgender women of color. Sex
ferences, our work stresses the importance of treating Roles. 2013;68:675–689.12 ann. behav. med. (2021) XX:1–13
2. Reisner SL, Radix A, Deutsch MB. Integrated and gender- of male-to-female transsexual subjects: a review. Andrologia.
affirming transgender clinical care and research. J Acquir 2014;46:791–795.
Immune Defic Syndr. 2016;72(suppl 3):S235–S242. 19. Sanchez NF, Sanchez JP, Danoff A. Health care utilization,
3. King WM, Gamarel KE. A scoping review examining social barriers to care, and hormone usage among male-to-female
and legal gender affirmation and health among transgender transgender persons in New York City. Am J Public Health.
populations. Transgend Health. 2020:6(1):5–22. 2009;99:713–719.
4. White Hughto JM, Reisner SL. A systematic review of the ef- 20. Perez-Brumer A, Hatzenbuehler ML, Oldenburg CE,
fects of hormone therapy on psychological functioning and Bockting W. Individual- and structural-level risk factors
quality of life in transgender individuals. Transgend Health. for suicide attempts among transgender adults. Behav Med.
2016;1:21–31. 2015;41:164–171.
5. Keo-Meier C, Fitzgerald KM, Pardo ST, Babcock J. The ef- 21. Testa RJ, Habarth J, Peta J, Balsam K, Bockting W.
Downloaded from https://academic.oup.com/abm/advance-article/doi/10.1093/abm/kaab063/6352463 by guest on 17 August 2021
fects of hormonal gender affirmation treatment on mental Development of the gender minority stress and resilience
health in female-to-male transsexuals. J Gay Lesbian Ment measure. Psychol Sex Orientat Gend Divers. 2015;2:65–77.
Health. 2011;15:281–299. 22. Hatzenbuehler ML, Phelan JC, Link BG. Stigma as a funda-
6. Murad MH, Elamin MB, Garcia MZ, et al. Hormonal therapy mental cause of population health inequalities. Am J Public
and sex reassignment: a systematic review and meta-analysis Health. 2013;103:813–821.
of quality of life and psychosocial outcomes. Clin Endocrinol 23. West C, Zimmerman DH. Doing gender. Gend Soc.
(Oxf). 2010;72:214–231. 1987;1:125–151.
7. Wilson EC, Chen YH, Arayasirikul S, Wenzel C, Raymond HF. 24. Hendricks ML, Testa RJ. A conceptual framework for clin-
Connecting the dots: examining transgender women’s util- ical work with transgender and gender nonconforming clients:
ization of transition-related medical care and associations an adaptation of the Minority Stress Model. Prof Psychol Res
with mental health, substance use, and HIV. J Urban Health. Pract. 2012;43:460–467.
2015;92:182–192. 25. Bakko M, Kattari SK. Transgender-related insurance denials
8. James SE, Herman JL, Rankin S, Keisling M, Mottet L, as barriers to transgender healthcare: differences in experience
Anaf M. The Report of the 2015 U.S. Transgender Survey. by insurance type. J Gen Intern Med. 2020;35:1693–1700.
Washington, DC: National Center for Transgender Equality; 26. Perone AK. Protecting health care for transgender older adults
2016. amidst a backlash of U.S. federal policies. J Gerontol Soc
9. Clark K, Fletcher JB, Holloway IW, Reback CJ. Structural in- Work. 2020;63:743–752.
equities and social networks impact hormone use and misuse 27. Rotondi NK, Bauer GR, Scanlon K, Kaay M, Travers R,
among transgender women in Los Angeles County. Arch Sex Travers A. Nonprescribed hormone use and self-performed
Behav. 2018;47:953–962. surgeries: “do-it-yourself ” transitions in transgender com-
10. White Hughto JM, Reisner SL, Pachankis JE. Transgender munities in Ontario, Canada. Am J Public Health. 2013;103:
stigma and health: a critical review of stigma deter- 1830–1836.
minants, mechanisms, and interventions. Soc Sci Med. 28. Du Bois SN, Yoder W, Guy AA, Manser K, Ramos S. Examining
2015;147:222–231. associations between state-level transgender policies and trans-
11. White Hughto JM, Rose AJ, Pachankis JE, Reisner SL. gender health. Transgend Health. 2018;3:220–224.
Barriers to gender transition-related healthcare: identifying 29. Gleason HA, Livingston NA, Peters MM, Oost KM, Reely E,
underserved transgender adults in Massachusetts. Transgend Cochran BN. Effects of state nondiscrimination laws on trans-
Health. 2017;2:107–118. gender and gender-nonconforming individuals’ perceived
12. Goldenberg T, Reisner SL, Harper GW, Gamarel KE, community stigma and mental health. J Gay Lesbian Ment
Stephenson R. State‐level transgender‐specific policies, race/ Health. 2016;20:350–362.
ethnicity, and use of medical gender affirmation services 30. de Haan G, Santos GM, Arayasirikul S, Raymond HF. Non-
among transgender and other gender‐diverse people in the prescribed hormone use and barriers to care for transgender
United States. Milbank Q. 2020;98(3):802–846. women in San Francisco. LGBT Health. 2015;2:313–323.
13. Glick JL, Andrinopoulos KM, Theall KP, Kendall C. 31. Goldenberg T, Reisner SL, Harper GW, Gamarel KE,
“Tiptoeing around the system”: alternative healthcare navi- Stephenson R. State policies and healthcare use among trans-
gation among gender minorities in New Orleans. Transgend gender people in the U.S. Am J Prev Med. 2020;59:247–259.
Health. 2018;3:118–126. 32. Stroumsa D, Crissman HP, Dalton VK, Kolenic G,
14. Meriggiola MC, Gava G. Endocrine care of transpeople part Richardson CR. Insurance coverage and use of hormones
II. A review of cross-sex hormonal treatments, outcomes among transgender respondents to a national survey. Ann Fam
and adverse effects in transwomen. Clin Endocrinol (Oxf). Med. 2020;18:528–534.
2015;83:607–615. 33. Van Schuylenbergh J, Motmans J, Defreyne J, Somers A,
15. Moore E, Wisniewski A, Dobs A. Endocrine treatment of trans- T’Sjoen G. Sexual health, transition-related risk behavior and
sexual people: a review of treatment regimens, outcomes, and need for health care among transgender sex workers. Int J
adverse effects. J Clin Endocrinol Metab. 2003;88:3467–3473. Transgend. 2019;20:388–402.
16. Streed CG Jr, Harfouch O, Marvel F, Blumenthal RS, 34. Gamarel KE, Watson RJ, Mouzoon R, Wheldon CW, Fish JN,
Martin SS, Mukherjee M. Cardiovascular disease among Fleischer NL. Family rejection and cigarette smoking among
transgender adults receiving hormone therapy: a narrative re- sexual and gender minority adolescents in the USA. Int J
view. Ann Intern Med. 2017;167:256–267. Behav Med. 2020;27:179–187.
17. Martinez C, Rikhi R, Haque T, et al. Gender identity, hor- 35. Courtemanche C, Marton J, Ukert B, Yelowitz A, Zapata D.
mone therapy, and cardiovascular disease risk. Curr Probl Early impacts of the Affordable Care Act on health insurance
Cardiol. 2018;45(5):100396. coverage in Medicaid expansion and non-expansion states:
18. Asscheman H, T’Sjoen G, Lemaire A, et al. Venous thrombo- impacts of the Affordable Care Act. J Pol Anal Manage.
embolism as a complication of cross-sex hormone treatment 2017;36:178–210.ann. behav. med. (2021) XX:1–1313
36. Muthén B, Muthén L, Asparouhov T. Categorical dependent 43. Reisner SL, Hughto JM, Dunham EE, et al. Legal protections
variable. In: Regression and Mediation Analysis Using Mplus. in public accommodations settings: a critical public health
Los Angeles, CA: Muthén & Muthén; 2016. issue for transgender and gender-nonconforming people.
37. Fritz MS, Mackinnon DP. Required sample size to detect the Milbank Q. 2015;93:484–515.
mediated effect. Psychol Sci. 2007;18:233–239. 44. Flores AR, Herman JL, Gates GJ, Brown TNT. How Many
38. Fritz MS, Taylor AB, Mackinnon DP. Explanation of two Adults Identify as Transgender in the United States? Los
anomalous results in statistical mediation analysis. Multivariate Angeles, CA: The Williams Institute; 2016.
Behav Res. 2012;47:61–87. 45. Janssen A, Voss R, Written on behalf of the Gender
39. Lewis NJW, Batra P, Misiolek BA, Rockafellow S, Tupper C. Development Program at the Ann and Robert H. Lurie
Transgender/gender nonconforming adults’ worries and Children’s Hospital of Chicago. Policies sanctioning dis-
coping actions related to discrimination: Relevance to pharma- crimination against transgender patients flout scientific evi-
Downloaded from https://academic.oup.com/abm/advance-article/doi/10.1093/abm/kaab063/6352463 by guest on 17 August 2021
cist care. Am J Health Syst Pharm. 2019;76:512–520. dence and threaten health and safety. Transgend Health.
40. Castañeda C. Developing gender: the medical treatment of 2020;1:61–63.
transgender young people. Soc Sci Med. 2015;143:262–270. 46. Radix AE, Lelutiu-Weinberger C, Gamarel KE. Satisfaction
41. Radix A. Standards of Care Version 8 Update. Boston, MA: and healthcare utilization of transgender and gender non-
Fenway Health; 2019. conforming individuals in NYC: a community-based partici-
42. Korpaisarn S, Safer JD. Gaps in transgender medical education patory study. LGBT Health. 2014;1:302–308.
among healthcare providers: a major barrier to care for trans- 47. Norton EC, Dowd BE, Maciejewski ML. Odds ratios—current
gender persons. Rev Endocr Metab Disord. 2018;19:271–275. best practice and use. JAMA. 2018;320:84–85.You can also read
