PreprintPDF Available

Best Practices for Collecting Gender and Sex Data

Authors:
  • American Statistical Association
Preprints and early-stage research may not have been peer reviewed yet.

Abstract

The measurement and analysis of human sex and gender is a nuanced problem with many overlapping considerations including statistical bias, data privacy, and the ethical treatment of study subjects. Traditionally, human gender and sex have been categorized and measured with respect to an artificial binary system. The continuation of this tradition persists mainly because it is easy to replication and not, as we argue, because it produces the most valuable scientific information. Sex and gender identity data is crucial for many applications of statistical analysis and many modern scientists acknowledge the limitations of the current system. However, discrimination against sex and gender minorities poses very real privacy concerns when collecting and distributing gender and sex data. As such, extra thoughtfulness and care is essential to design safe and informative scientific studies. In this paper, we present statistically informed recommendations for the data collection and analysis of human subjects that not only respect each individual's identity and protect their privacy, but also establish standards for collecting higher quality data.
Best Practices for Collecting Gender and Sex Data
Word Count: 6027
1
arXiv:2103.09647v1 [stat.ME] 17 Mar 2021
Suzanne Thornton
Assistant Professor
Mathematics and Statistics Department, Swarthmore College
500 College Ave, Swarthmore, 19081, PA, United States
email: sthornt1@swarthmore.edu
Dooti Roy
Principle Methodology Statistician, Boehringer Ingelheim
Danbury, 06810, CT
email: dootiroy@gmail.com
Stephen Parry
Statistical Consultant, Cornell University
Ithica, 14850, NY, United States
email: sp2332@cornell.edu
Donna LaLonde
Director of Strategic Initiatives and Outreach, American Statistical Association
Alexandria, 22314, VA, United States
email: DonnaL@amstat.org
Wendy Martinez
Director, Mathematical Statistics Research, Bureau of Labor Statistics
Washington, DC, 20212, United States
email: martinez.wendy@bls.gov
Renee Ellis
Family Demographer, US Census Bureau
Washington, DC, 20002, United States
email: reneerellis@gmail.com
David Corliss
Lead data scientist, manager of the Data Science Center of Excellence, Stellantis
Auburn Hills, 48309, MI, United States.
email: davidjcorliss@gmail.com
2
March 18, 2021
Corresponding author. E-mail: sthornt1@swarthmore.edu
3
Abstract
The measurement and analysis of human sex and gender is a nuanced problem with many over-
lapping considerations including statistical bias, data privacy, and the ethical treatment of study
subjects. Traditionally, human gender and sex have been categorized and measured with respect to
an artificial binary system. The continuation of this tradition persists mainly because it is easy to
replication and not, as we argue, because it produces the most valuable scientific information. Sex
and gender identity data is crucial for many applications of statistical analysis and many modern
scientists acknowledge the limitations of the current system. However, discrimination against sex
and gender minorities poses very real privacy concerns when collecting and distributing gender and
sex data. As such, extra thoughtfulness and care is essential to design safe and informative scientific
studies. In this paper, we present statistically informed recommendations for the data collection
and analysis of human subjects that not only respect each individual’s identity and protect their
privacy, but also establish standards for collecting higher quality data.
1. INTRODUCTION
When teaching introductory statistics, many instructors encounter the question: Why do we typi-
cally use a significance level of 0.05 (or 0.01 or 0.10)? The answer is simple but not very gratifying:
this is a matter of convention. Helpful guidelines are well-established but there are infinite possible
significance levels a researcher could decide to use. A similar conundrum occurs when collecting
data on socially-defined attributes such as race or gender. The typical ways of categorizing these
variables have been passed down from statisticians to medical practitioners and other researchers
more often due to their ease of replication rather than their actual scientific value. However, quali-
ties such as these represent social constructs which, by their very nature, are not fixed but are fluid
concepts with evolving definitions. Merriam-Webster defines the term social construct as “an idea
that has been created and accepted by the people in a society”. Scientific endeavors must recognize
that many qualitative human characteristics fit this definition and as researchers, we are responsible
for practicing extra thoughtfulness and care when considering these traits as variables of interest
in our studies. Ultimately, we cannot adequately provide scientific or medical benefits to socially
4
marginalized people if we are not collecting the right data.
The goal of this document is to present statistical practitioners with guidelines on how to collect
quality data on sex and gender for adult subjects. This work is meant to serve as an introduction
to data quality concerns and ethical concerns from a perspective of gender inclusivity, a perspective
that has not traditionally been emphasized in the history of data science. We believe it is time to
challenge convention and strive for better practices when collecting and analyzing human-centric
data. This document is not a definitive source for understanding the complexities of gender identity
but does provide an introduction to the nuances of gendered language from a statistical perspective
and outlines a framework for researchers to adopt to better devise useful to address this specific
challenge in quantifying the human experience. These recommendations are meant to be revisited
and updated as our society evolves.
2. BACKGROUND INFORMATION AND TERMINOLOGY
While a thorough introduction to the language of sexual and gender identities is beyond the scope of
this paper, we have listed some terms and definitions in Figure 1 which will be referenced throughout
the remainder of the text. Many other reputable resources elaborate upon these terms and present
more terminology than we do, e.g. Bates et al. [2019]; Bauer et al. [2019]; and Badgett and Team
[2009]. For additional resources, see Appendix A.
5
Term Definition
Cisgender An adjective used to describe a person whose gender identity aligns
with the gender assigned to them at birth.
Gender
affirming
surgery
Surgery that is performed to modify a person’s body to be more
aligned with that person’s gender identity. These may include chest
and genital surgeries, facial feminization, and/or hair removal.
Gender
expression
The many ways in which someone may express their gender identity.
This can include clothing, behavior, and mannerisms which may be
intentionally or unintentionally ambiguous.
Gender identity
One's innermost concept of self as male, female, a blend of both or
neither – how individuals perceive themselves and what they call
themselves. One's gender identity can be the same or different from
their sex assigned at birth. Gender identity cannot reliably be inferred
from gender expression.
Gender or sex
minority
A catch-all phrase that includes transgender, non-binary, other gender
non-conforming, and intersex people.
Different Sexual
Development
(DSD)
See Intersex.
Intersex
An adjective that describes someone who has chromosomes,
anatomy, or other sex characteristics that don’t fit the typical
definitions of female or male. This can include people with genetic
predispositions to androgen sensitivity or people with hormones and
genetics that indicate one sex but an anatomy that is typical of
another sex.
Misgender
A verb that describes the action of referring to a person by a pronoun
or other gendered term (e.g., Ms./Mr.) that incorrectly indicates that
person’s gender identity. This is a significant source of emotional
harm for transgender individuals.
Non-binary
A term for people whose gender identity exists outside of the gender
binary of male or female. Many non-binary people identify as
transgender since their gender identity is likely to be different from the
sex assigned to them at birth.
Sex assigned at
birth
The sex (typically male or female but occasionally a neutral X)
assigned to a newborn. Most often, this designation is based on the
infant’s anatomical characteristics.
Sexual
orientation
An inherent or immutable enduring emotional, romantic or sexual
attraction to other people.
Transgender
An adjective used to describe a person whose gender identity is
different from the gender they were assigned at birth. Note that
although intersex and transgender identities may overlap, they are not
necessarily co-occuring identities.
Transition
The often lengthy process whereby a transgender person begins to
acknowledge their true gender identity in their private and/or public
life.
Figure 1: Common terms relevant to gender and sex minorities. Definitions are derived from the
National LGBTQIA+ Health Education Center and the Human Rights Campaign.
6
It is important to note the many possible intersections of LGBTQ+ identities. Naturally, the
full scope of people’s identities extends beyond any one particular categorization. Human beings
exist as unique intersections of multiple identities, some of which are ascribed by others and some
of which are personal and may not be readily identifiable to others. Figure 2 presents a simple
diagram that illustrates the possible (in)congruencies among gender and sex characteristics.
Sex Recorded
at Birth
Primary and
Secondary Sex
Characteristics
Gender Identity
Figure 2: A Venn Diagram show three different aspects of sex and gender identity. For many
people, all three characteristics align; however, for gender and sex minorities one or more of these
characteristics may not align in the typical manner.
2.1 Sexuality
We are not discussing sexuality in this paper but we mention it here because sexual preference is
often confused with the concepts of gender identity and sex. While sex and gender are descriptors
of one’s self and personal identity, sexuality is a descriptor of one’s romantic attraction to others.
For example, a transgender woman may identify as straight if she is only romantically interested
in men; a cisgender man may identify as gay if he is attracted to other men who may or may not be
cisgender. It is important for data scientists to be aware of this distinction because potential biases
7
can be introduced in a data set by focusing on one identity but using general, catch-all terms like
“LGBTQ+” or “queer” — neither of which specify identify gender identity or sexuality.
2.2 Gender
Though a nuanced understanding of gender may seem like a new trend to some, the modern recog-
nition of gender as a continuum actually has roots that stretch across the globe and into the ancient
past; c.f. e.g. Wiesner-Hanks [2011] and Herdt [2020]. For modern scientists to be advocating for
an understanding of gender identity that extends beyond a binary classification system, is actually
an example where modern science is only just now catching up to global historical and cultural
wisdom.
The enornaminity of gender identity is an immutable factor of the human experience; the lan-
guage with which we describe that experience however is non-constant and changes with generations.
For example, although the Chicago Manual of Style renounced the formal use of “they” as a generic
third-person singular pronoun in 2017, two years later the American Psychological Association en-
dorsed the formal use of “they” to refer to a person whose gender identity is unknown or irrelevant
to the context of the usage. Regardless of the social norms of the times, the purpose of data collec-
tion has always been to understand a particular population by capturing information in bite-sized,
digestible pieces. Although information loss is an inevitable byproduct of this process when applied
to social constructs such as gender, our goal as statistical practitioners is to design surveys and
experiments to capture the most relevant bits of information as possible.
The concept of gender in modern society can be broken into two components, gender expression
and gender identity. Respecting someone’s gender identity is an ethical practice that honors the
individual’s sense of self. Gender identity is a personal understanding of oneself; it comes from
within, not without. Gender expression refers to many possible expressions of gender including
attributes such as clothing, hair, mannerisms, postures, etc. Gender expression is like art in that it
may be formed by an artist who wants to communicate something particular but is subject to the
interpretation of the viewers who may or may not understand the artist’s intent. For any individual,
gender identity and gender expression can change with time. For example, a transgender woman
may not have transitioned until later in her life and thus in her younger years may have adopted
masculine or androgynous gender expression for various reasons including her personal safety and
8
comfort. The process of transitioning in one’s public and private life takes years and can be a
non-linear process. Some transgender men for example, may identify as non-binary as they begin
to question the gender identity imposed upon them from birth and then realize their identity as
men years later. Gender expression can be intentionally ambiguous or it can challenge the status
quo. For all of these reasons, it is best to never assume any individual’s gender identity from their
gender expression alone.
A common way to communicate gender identity is to share one’s pronouns, similar to sharing
a surname or a title. Asking someone which pronouns they use and providing your own is a way
to actively respect diverse gender identities. Pronouns can indicate information regarding how one
identifies themselves along the gender spectrum, from masculine to feminine and beyond. Some
people use gendered pronouns such as she/her/hers or he/him/his, other people may only use the
gender neutral singular they/them/theirs, and others still may use different gender-neutral pronouns
such as ze/zim/zir. In line with the American Psychological Association, we recommend the usage
of the singular “they” as an appropriate and respectful way to refer to people whose gender identity
is unconfirmed or non-binary.
2.3 Sex
Despite the overwhelming tendency to treat biological sex as a binary variable, the term “biological
sex” is not quite so cut and dry. When sexual characteristics of individuals are a variable of interest in
a study this can refer to primary or secondary sexual attributes, genetic information, gender assigned
at birth, or many other sexual characteristics. The most common use of the phrase “biological sex”
seems to refer either to a person’s chromosomes or to their genitalia at birth, however, neither of
these attributes are actually binary [Ainsworth, 2018].
A consensus statement published in 2005 endorsed the terminology disorded/different sex de-
velopment (DSD) to describe individuals with congenital conditions in which the development of
chromosomal, gonadal, or anatomic sex is atypical. However, the term intersex is preferred by
many activists and advocates since it is not affiliated with a derogative connotation like the word
“disorder” which implies something that needs to be “fixed” [Hughes et al., 2006, interACT, 2020].
In this article we will therefore refer to anyone with these congenital conditions as intersex.
Intersex people can be born with the differences mentioned above, or they may develop these
9
differences later in childhood. This makes pinning down the incidence rate very difficult. Most
experts agree that the prevalence of intersex births is about 2% of the population which, for com-
parison, is similar to the prevalence of red hair; c.f. e.g. Onyiriuka et al. [2016] or Fausto-Sterling
[2000]. People with ambiguous or atypical sexual characteristics are often coerced into changing
their bodies, usually at a very young age. Most surgeries to change intersex traits happen in infancy
and it is possible that an adult may not be aware of this procedure in their medical history as noted
by Greenberg and Stam [2012].
We bring up these points to illustrate that when one is conducting a study that merely asks
participants for their sex, one risks losing important information by making the assumption that
all participants understand “biological sex” in the same way. This is true not just for intersex
participants but also can apply to transgender patients who may have undergone gender confirming
genital reconstructive surgery. When collecting data on attributes that can have different meanings,
it is best for the researcher to be as explicit as possible about their working definitions to avoid
misclassification bias.
3. A MULTIFACETED ARGUMENT FOR INCLUSIVE DATA
Among the many reasons why researchers may want to be more inclusive in their data collection
process, here we focus on two: the ethical treatment of human subjects and the collection of better
quality data. While we do not claim to be experts in ethics, data is our specialty, and we have
found that these two arguments overlap. Any ethically responsible researcher or clinician collecting
data on human participants should be aware of the language needed to correctly identify different
segments of the population. When options outside of the gender binary are not provided in surveys,
for example, the researcher overlooks an entire group of people from their analysis. Or when
individuals do not see their identities respected in the data collection process, they may experience
feelings of confusion, depression, or failure as a result. On a larger scale, exclusion of gender and
sex minorities means that research results may be unable to identify smaller sub-populations who
may be the most vulnerable to certain illnesses.
For instance, previous research has shown that transgender individuals might be at a higher
risk than the cisgender population for a variety of health issues such as stress, blood pressure,
10
and heart attacks [Alzahrani et al., 2019, Dubois, 2012, Irwig, 2018]. Mental health issues are
also disproportionately prevalent among gender and sex minorities, and, as is widely acknowledged
within the LGBTQ+ community, the main factors contributing to these issues stem from perceived
and realized threats to personal safety and social inclusion; c.f. e.g. Olson and McLaughlin [2016];
Clements-Nolle et al. [2006]; or Pascoe and Richman [2009]. By collecting relevant information in
a way that respects gender identities, researchers can help mitigate the crippling barriers to health
care than many transgender and non-binary people face. For example, a 2018 study found that
transgender people whose pronouns were respected by all or most people in their lives attempted
suicide at half the rate of those who did not have their pronouns respected [Russell et al., 2018].
The fundamental goal of statistics is to articulate information about some population through
the collection and analysis of data. The ethical statistician must use methodology and data that
are relevant and appropriate, without favoritism or prejudice, in a manner that produces valid,
interpretable, and reproducible conclusions [Association, 2018]. There is much information to be
learned from gender and sex qualities of human subjects provided these qualities are researched in
a way that respects the variability of identities and experience.
As is true for data collection at large, when gender and sex are attributes collected without
forethought, the data can be easily and irreparably corrupted by statistical biases. A study is
susceptible to non-response bias, for example, when survey participants do not see their identities
represented in the questions or answers. In such situations, participants may become disgruntled
and not complete the survey and therefore the conclusions from statistical analyses of the data
may not apply to important subgroups of the population. Of course misclassification bias is also
possible when surveys are not designed to acknowledge different identities. If a clinician assumes
a participant’s gender identity based on their appearance, voice, or mannerisms, the information
encoded in the survey may be irrelevant, or worse, inaccurate.
With both response and non-response bias, the observed data fail to accurately represent the
population of interest. Ultimately this can lead to a loss in the signal of the response variable.
Although it will take time and effort to transition to more inclusive data collection practices for
gender and sex, these efforts will eventually result in better quality data and more useful analyses.
With more informative data, signal detection may become easier and resulting statistical models
11
may be useful for a wider range of the population. Example 1 considers a setting in which these
biases can be hazardous to the health of transgender patients in particular.
Example 1
Setting: Healthcare providers often hesitate to prescribe drugs citing that there is no information
about the safety of the drug on the specific population on the drug label. If there are no data on
a specific group of people and the healthcare providers suspect that the subgroup population may
react differently than the rest, they may be reluctant to prescribe the drug. A number of studies have
highlighted the increased risk of mental illnesses (e.g. depression, anxiety, and suicidality) in gender
and sex minorities, e.g. Su et al. [2016]. For clinical trials that are focused on treatments for mental
illnesses, it is therefore crucial to identify individuals who may have a higher risk of certain adverse
events. Furthermore, transgender and non-binary patients may take large doses of hormones as part
of their transition and hormone therapy may significantly influence and impact the safety or efficacy
of a particular drug. For these reasons, it is especially relevant for these trials to be able to identify
transgender patients to ensure that a drug is safe and effective across all populations.
Problem: Most modern clinical trials are not designed to identifying transgender patients,
even if the patient voluntarily self-identifies. Also, within a clinical trial setting, researchers are
required to justify and explain why certain information is collected and needed for the study.
Proposed solution: Collect both “sex assigned at birth” and “gender identity” as the baseline
characteristics during the conduct of the clinical trial. If these two fields do not align, this indicates
a patient who is potentially transgender. Researchers can then examine these sub-populations to
determine if there is an adequate representation within the sample and to see if there may be reason
to suspect that the efficacy and safety of the test drug is different for this sub-population. This
solution can help clinicians determine if transgender subjects may have a different response to the
test drug.
Another consideration when incorporating gender or sex information into a statistical model is
to be aware of the potential biases from imputing missing data. There are many programs that
attempt to infer a person’s gender using data mining techniques. The possible biases resulting from
these methods depend on the basis with which these packages determine gender and on whether
gender identity or sex at birth (or something else) is the information being imputed. The gender
12
R package for example, categorizes a person as “male” or “female” based historical data of gendered
names [Mullen, 2018]. Therefore, as acknowledged in the package documentation, in no way can
this package determine a person’s gender identity or can it determine any specific details about a
person’s sexual characteristics. Furthermore, it can only partially identify subjects from historically
marginalized gender roles as it does not measure transgender or other gender-non-conforming indi-
viduals. Another tactic that is often used to impute subjects’ missing gender or sex classifications
is to use crowdsourcing techniques. Most often, these methods rely on people who are paid to look
at names and sometimes pictures of the subjects and then make a personal judgment to classify the
subject as male or female. Hence, this method is subject to both personal and cultural bias and
prone to misclassification errors.
Collecting data in a way that honors different gender identities will contribute to an increase in
the general knowledge of and respect for various gender identities in the greater scientific community.
The results of these efforts could have an enormous effect on the quality of life of gender and sex
minorities both by setting better scientific standards for the ways in which human subjects are
treated with respect and by providing better targeted health care through more representative
data analyses. Since knowledge of gender and sex demographics can better inform a clinician in
their own practice, the statistical argument for more informative data and the usefulness of the
resulting statistical models is highly connected to the ethical argument for inclusive language and
data collection procedures.
4. STATISTICAL CONSIDERATIONS AND PRACTICAL GUIDELINES
Here we present a three-step approach for statistical practitioners to consider whenever collecting
data on human sex and gender. Broadly speaking, the steps we encourage each researcher to take
are to 1) identify the information relevant to the study; 2) respect the study participants’ identities
with inclusive language; 3) protect the data and participants’ identities. We discuss statistical
considerations that arise within these steps and offer some solutions for common issues that may
arise, and apply these suggestions in two real examples.
13
4.1 Identify relevant information
Gender and sex information does not necessarily need to be included in every study. Before re-
searchers begin a project, it is important to determine whether it possible that gender or sex may
affect the efficacy of the treatment. If researchers do not believe that gender or sex can moderate an
effect, then they may not need to ask for this sensitive information. However, in many clinical set-
tings in particular, the researcher may expect gender identity or sex to significantly affect the efficacy
of a treatment. In such cases, it is important for the researcher to explicitly state the underlying
mechanisms. For instance, if sex may affect the efficacy, which sexual characteristics specifically
are hypothesized to affect the treatment? If someone is receiving hormone therapy treatment, is
this an important factor in the efficacy of the treatment under study? If someone has undergone
gender-affirming surgery, is this an important factor in the efficacy of the treatment? Ultimately,
asking more specific questions about bodily attributes rather than using gender or sex as a catch-all
variable will lead to more reliable and inclusive data. For example, cisgender women, transgender
women, and transgender men may all need breast cancer screening. Therefore, a survey that asks
patients for their gender identity only will exclude transgender men and conversely, a survey that
asks patients for their sex assigned at birth only, will likely exclude transgender women.
Often, researches use the words “sex” and “gender” as interchangeable characteristics. This
is a dangerous practice in that it will most certainly lead to some type of misclassification bias.
If gender or sex are relevant features of a statistical model, the researcher must dig deeper and
ask is “What is relevant to my study, gender or sex or both?” This is merely an example of the
more general statistical best practice, when collecting data from which to construct a statistical
model, the researcher must carefully consider what are the variables of interest. Although there are
many statistical techniques for variable selection, there is no equal substitute for domain-specific
knowledge when it comes to variable selection. Example 2 provides an example of a clinical study
where gender and sex information were conflated, influencing the conclusions of the data analysis.
Example 2
Setting: One of the authors (ST) served as a methodologist for a research project that developed
a predictive model for a certain type of drug-resistant epilepsy. The researchers were interested in
determining the efficacy of an anti-epileptic drug on people with different types of epilepsy [Choi
14
et al., 2020]. To predict whether or not this drug will reduce seizures in various types of patients,
the researchers conducted a retrospective study and developed a predictive model based on data that
had been collected many years prior across two different clinical groups. Useful predictive variables
included age; different seizure types; whether or not there was a change in seizure frequency with a
patient’s menstrual cycle (called Catamenial epilepsy); and the sex of the patient (which was stated
without explanation and only marked as male or female).
Problem: The most important predictive variables are heavily related to one another. In
particular, Catamenial epilepsy and sex contain overlapping information. The solution proposed by
the statistician was to develop a new, three-level categorical predictor variable that indicated both
sex and the occurrence of Catamenial epilepsy rather than to use each of these variables separately.
Upon further reflection, the statistician started to realize the limitations of assessing signals from
ambiguous data collection procedures regarding sex and gender and realized that the data collection
procedure regarding the sex of the patients in this study was insufficient for capturing what appears
to be the most relevant information about this variable.
Proposed solution: For future studies, at the data collection stage, rather than asking for a
patient’s sex, more informative data could be easily obtained by instead asking patients both for their
gender identity and whether or not they regularly experienced menses. This simple adjustment to
the data collection stage would prevent the exclusion or missclassification of transgender men (who
may still experience menses) and cisgender women who may not experience menses due to age or
medical procedures (such as a hysterectomy) or having non-ovary producing internal genitalia.
In a clinical trial setting where a group of many typical characteristics of human sex development
is of interest, an inclusive question may ask participants for their sex assigned at birth. This will
typically be either assigned male at birth (AMAB) or assigned female at birth (AFAB). However, it is
worth considering how important information about intersex people may be lost in this classification
system or how important information about transgender people may be lost if gender identity is
not also specified. Furthermore, in the US and in other countries, gender neutral birth markers (X)
are an increasingly popular option for new parents.
Research has shown that individuals from the LGBTQ+ population tend to experience higher
rates of depression and suicidal ideation as well as perceived and experienced threats to personal
15
safety [Almeida et al., 2009]. If mental health could possibly moderate the effect of a treatment,
then sexual orientation and gender identity might be important variables to collect.
Researchers may be interested in gender identity or sexual characteristics of participants, but
must ensure binary sex is not being relied upon as a proxy for another more precise variable such
as the presence of breast tissue or type of hormonal balances. Imagine, for instance, designing a
study to identify determinants of prostate cancer. Rather than ask patients for their sex, one may
instead simply ask patients if they have (or have ever had) a prostate. This type of gender-neutral
re-wording of survey questions is much more specific and can collect more relevant information in
a way that respects individuals’ identities. These measures can help avoid the confusion of dealing
with transient variables such as sex or gender, both of which may change with time depending
on how they are defined. Furthermore, this type of gender-neutral questioning avoids causing any
unnecessary mental trauma to transgender and gender-non-conforming individuals who may feel
outted or experience intense gender dysphoria as a result of being forced to identify with a binary
gender that does not match their own identity. Figure 3 outlines general steps for evaluating what
information is relevant for differently motivated studies.
4.2 Center inclusivity and respect
The purpose of institutional review boards is to protect the rights and welfare of human subjects.
Human welfare is intimately connected with perceived respect, especially for those that identify
with various socially and politically marginalized groups [Durwood et al., 2017]. The discretization
of social constructs for the purpose of statistical analysis is not new and neither, unfortunately, are
data collection processes that disregard individual identities. For example, it was not until 1960 that
individuals could select their own racial identity in the US Census [Brown, 2020]. As a consequence
of the National Research Act being signed into law in 1974, the Belmont Report identified “respect
for persons” as a basic ethical principle of research. The report defines “respect for persons” based on
two fundamental ethical convictions: first, that individuals should be treated as autonomous agents,
and second, that persons with diminished autonomy are entitled to protection [HEW, 1979]. It is
critical for the researcher to take extra measures to ensure the respectful treatment of human subjects
is an intrinsic feature of any human-centric scientific endeavor. Unfortunately, the considerations
necessary for the respectful treatment of gender and sex minorities are too often overlooked.
16
Is your study concerned with socially created ideas (like gender expression, gender inequality, or gender identity)?
Is your study also concerned with biological
considerations?
Build questions that directly ask
about study concern rather than
using sex or gender as a filter
question. For example: Do you have
a prostate? Or Are you
breastfeeding?
Build protections into your project for small subgroups. Think about how you will approach statistical issues resulting from small
sample sizes and fluidity/uncertainty.
Include identifying questions for the comfort of your respondents or patients. For example: What is the preferred name you would
like us to use? What are your pronouns?
Include questions
on pertinent social
constructs.
Consider all
categories that may
be needed to be
inclusive.
Build questions that directly ask
about study concern rather than
using sex or gender as a filter
question. For example: Do you have
a prostate? Or Are you
breastfeeding?
Include questions on pertinent
social constructs. Consider all
categories that may be needed to
be inclusive.
Yes
Yes
No
No
Figure 3: A flow-chart outlining how to identify the relevant gender and sex information to reference
in a particular study.
17
One important measure that expresses respect to participants is the explicit communication
of the purpose behind collecting sensitive data. Participants have the right to understand why
their information is being collected and how it is going to be used and their identities protected.
As discussed in the previous section, another important consideration is to recognize that questions
about sex and gender may cause an undue amount of stress or confusion for transgender and gender-
non-conforming respondents. If it is necessary to question a study participant about their sex, the
researcher should clearly define what is meant by “sex” in a way that does not coerce the participant
to adopt an inappropriate gender identity but instead emphasized specific sexual characteristics
without gendered language. For example, in many non-clinical studies, gender is recorded because
the variable of interest is whether or not an individual belongs to a historically marginalized group.
In such cases, it would be misleading to discount the marginalization of transgender men, say,
by grouping them together with cisgender men. In this case, the question about gender could
be reworded to instead ask participants if they are cisgender men, cisgender women, transgender
(men or women) or otherwise gender-non-conforming. As another example, if the outer genitalia a
person currently has is a key feature of interest, it’s best to ask this explicitly rather than ask the
participant if they are male or female. On the other hand, if the genitalia a person was born with
is the key factor, then the question can be easily adapted to adjust for this without challenging the
participant’s gender identity.
In situations where the data analyst has no control over the data collection procedure, such as
in retroactive studies, the analyst is responsible for noting any ambiguities or possible sources of
bias in the data at hand. If gender information was collected but there were no options for people
who did not identify as a man or a woman, this is important to note because the study is limited by
not applying to gender-non-comforming populations. If the retroactive study collected information
on the sex of subjects, the data analyst should strive to understand this variable in the context of
the original study. In the resulting analysis, the researcher can still demonstrate respect for the
human subjects by including a discussion on how sex was coded and, following the suggestions
in the previous section, by discussing what particular sex characteristics were probably of interest
using gender-neutral language. The importance of this careful and thoughtful data collection is
only growing as many states (and countries) begin to offer gender-neutral identifiers on official
18
documentation such as birth certificates and drivers’ licenses [Fitzsimons, 2019].
Even if the data analyst can control the manner in which the data is collected, creating additional
categories such as “non-binary” for gender or “intersex” for sex can create difficulties in the statistical
analysis if small sample size considerations are not taken into account. Whether or not a particular
category is a valid option for a variable in a statistical model often comes down to an issue of sample
size. To avoid this small sample size problem that often results in wasted data, the researcher must
take time before collecting the data to plan the study with carefully developed sampling strategies
and/or with explicitly defined questions that take into account the topics discussed in the previous
section. Obviously these measures are important for studies that focus on the LGBTQ+ community.
These considerations are especially important for gender and sex minorities, c.f. e.g. Lau et al.
[2020]; Goldstein et al. [2017]; and Gupta et al. [2016].
Some additional considerations for surveys that take into account the points discussed above.
When asking about participants’ sex, consider including intersex as an option or if possible, consider
rewording the question to ask about the specific sexual characteristics of interest. When asking about
gender identity for classification purposes, include at least a third option for non-binary or gender
non-conforming individuals. Similarly, if asking for participant pronouns be sure to include a gender
neutral option such as the singular “they” in addition to the typical binary options of she/her/hers
and he/him/is. When it is important to identify transgender and other gender and sex minority
groups, we recommend using the two-step approach, asking participants both for their sex assigned
at birth and for their gender identity [Badgett and Team, 2009].
4.3 Protect the participant and the data
In our day-to-day activities, we regularly find ourselves being asked to provide our sex for events
such as doctor’s appointments, filling out the Census, registering a profile on a social media page,
or applying for a job. For cisgender people, these questions may not cause a second through but for
transgender and gender-non-conforming people, these questions can elicit not just mild discomfort
but feelings of dread and fear for personal safety. There are many reasons why a person may not
want to publicly reveal that their sex assigned at birth does not align with their gender identity;
c.f. e.g., Safer et al. [2016] and Wirtz et al. [2020]. From a statistical perspective, is important to
recognize these concerns from the perspective of gender and sex minorities. If data protection and
19
personal privacy cannot be guaranteed to these groups where the degree of potential harm from
data misuse is great, there may be a severe non-response bias. More importantly, an individual’s
safety may be under threat. Of course from an ethical perspective, scientists recognize that data
collection should not cause undue levels of stress to participants [Bennouna et al., 2017]. There are
two levels of data security to consider when collecting information on a person’s gender identity
or sex. The security of this information is crucial when collecting the data and when analyzing or
sharing the data.
In practice, data privacy is complicated by a lack of general or consistent standards. Differing
approached to privacy can be illustrated by considering the parts of speech used. The usual method
for protecting data privacy is through nouns: creating a list of data objects that qualify for protection
as Personally Identifying information (PII). This results in conflicting standards as each regulating
organization may have a different list. For example, the US Department of Homeland Security
has a list of nineteen items, six of which are considered PII alone and nine are considered PII if
paired with another item on the list [DHS, 2012]. However, the US Department of Health and
Human Services has a single tier of 18 items. Only thirteen items are on both lists. Neither protects
gender; sexual orientation qualifies as Sensitive PII for Homeland Security but is left unprotected
by HHS. Modern analytic algorithms pose an additional problem: identifying persons through a
multi-point match of data objects appearing on no list. A different approach to data privacy is
suggested Australian Data Privacy act, established in 2008 and updated it in 2014 and 2019 [Morris
and Cravigan, 2020]. This act takes a verb-based approach to privacy, placing restrictions where
an individual “reasonably identifiable”. Taking this notion a step further, we can better understand
data privacy as an adverb: the manner in which things are done. Assuring data, personal identify,
and privacy are protected with adverbs - securely, transparently, ethically - achieves the desired
intent.
Socially typical safety standards may not be adequate especially for transgender individuals [Rof-
fee and Waling, 2017]. As noted, the US HHS standard conflicts with others and affords no protec-
tion to sexual identity data. The unfortunate reality is that many gender-non-conforming children
may not feel safe to reveal their gender identities if a parent or legal guardian is present; c.f. e.g.
Riggs and Bartholomaeus [2018] and Catalpa and McGuire [2018]. Although we are limiting our
20
discussion in this article to data collection for adults, note that extra considerations are needed if
the participants of a study are minors. These considerations for safety of the subject participants
are relevant for adults as well. If a person is verbally asked to identify their sex and/or gender
identity in say a doctor’s waiting room, a transgender patient may feel vulnerable to harassment to
others in the room. Thus, it is important that security and privacy measures are taken into account
at the data collection stage by minimizing the opportunities for discrimination and harassment of
the participants.
The privacy of personal information, especially in small studies, can leave minorities vulnerable
to discovery if the information collected is too specific. For example, a transgender patient’s identity
may be identifiable from the data if it includes particular hospital information or information about
a specific type of treatment. It is possible that gender and sex minorities may be identified even if a
study adheres to HIPPA requirements [Corliss, 2015]. While a full discussion of data identifiability
concerns is beyond the scope of this article, a tried-and-true statistical recommendation is to plan
studies to avoid small sample size problems. This may require budgeting time or money to implement
carefully designed sampling procedures. According to a 2018 Gallup poll, approximately 5% of the
US adult population identify as LGBTQ+. Because gender and sex minorities represent a small
proportion of the overall population, any survey collecting gender or sex data is likely to face the
issue of unavoidably small sample sizes. In such instances, we recommend analyzing aggregated
information rather than rely on a single survey for analysis. There are many statistical methods
for conducting meta-analyses on aggregated data and the benefits of such methods may include
higher statistical power in addition to stronger data privacy [Saha-Chaudhuri and Weinberg, 2017,
Gatewood et al., 2001, Abowd, 2018].
5. DISCUSSION
LGBTQ+ individuals have always been part of society, even if these identities have not been encoded
in previous data collection efforts. Much research concerns response variables that can be moderated
by gender, sexual characteristics, or both. To prevent the continued erasure of minority populations
from scientific research, it is time for the standards of data collection procedures to include options
outside of the traditional gender and sex binary. Another important result of refining how gender
21
and sex data is collected will be a decrease in measurement error and bias in the results of statistical
analyses.
A. APPENDIX
For the interested reader, we have compiled some additional material that may be of use for im-
plementing more inclusive data collection procedures. Example of inclusive demographic survey
questions can be found on many survey sites such as Alchemer (https://www.alchemer.com/resou
rces/blog/how-to-write-better-demographic-questions). There are also many consulting and train-
ing resources offered by companies such as Speaking of Transgender (https://speakingoftransgend
er.com/transgender-training-and-course) and Sylveon Consulting (https://www.sylveonconsulting.
com). Writing tools and guidance are provided by many freelance professionals and professional
groups such as transgender author Julia Serano (https://www.patreon.com/juliaserano/overview
), The Society of Professional Journalists (https://www.spj.org/race-gender-hotline.asp), Radical
Copy Editor
(https://radicalcopyeditor.com/2017/08/31/transgender-style-guide)
A wealth of additional educational material can be found on the websites for the Federal Com-
mittee on Statistical Methodology (https://nces.ed.gov/FCSM/SOGI.asp), Planned Parenthood (
https://www.plannedparenthood.org/learn/gender-identity), The LGBT Foundation
(https://lgbt.foundation/research-guidance), and The National LGBTQIA+ Health Education Cen-
ter (https://www.lgbthealtheducation.org/resources/type/publication).
REFERENCES
J. M. Abowd. The us census bureau adopts differential privacy. Proceedings of the 24th ACM
SIGKDD International Conference on Knowledge Discovery & Data Mining, pages 2867–2867,
2018.
C. Ainsworth. Sex redefined: The idea of 2 sexes is overly simplistic. Scientific American, 22, 2018.
J. Almeida, R. M. Johnson, H. L. Corliss, B. E. Molnar, and D. Azrael. Emotional distress among
lgbt youth: The influence of perceived discrimination based on sexual orientation. Journal of
youth and adolescence, 38:1001–1014, 2009.
22
T. Alzahrani, T. Nguyen, A. Ryan, A. Dwairy, J. McCaffrey, R. Yunus, J. Forgione, J. Krepp,
C. Nagy, R. Mazhari, and J. Reiner. Cardiovascular disease risk factors and myocardial infarction
in the transgender population. Circulation: Cardiovascular Quality and Outcomes, 12(4), 2019.
American Statistical Association. Ethical guidelines for statistical practice. Available at https://
www.amstat.org/asa/files/pdfs/EthicalGuidelines.pdf , 2018. Committee on Professional Ethics.
M. Badgett and Sexual Minority Assessment Research Team. Best practices for asking questions
about sexual orientation on surveys. UCLA: The Williams Institute, 2009.
N. Bates, Y. A. García Trejo, and M. Vines. Are sexual minorities hard-to-survey? insights from
the 2020 census barriers, attitudes, and motivators study (cbams) survey. Journal of Official
Statistics, 35(4):709–729, 2019.
G. Bauer, A. Devor, M. Heinz, Z. Marshall, A. Pullen Sansfaçon, and J. Pyne. Cpath ethical guide-
lines for research involving transgender people & communities. Canadian Professional Association
for Transgender Health, 2019.
C. Bennouna, H. Mansourian, and L. Stark. Ethical considerations for children’s participation in
data collection activities during humanitarian emergencies: A delphi review. Conflict and Health,
11(1):1–15, 2017.
A. Brown. The changing categories the u.s. census has used to measure race. Pew Research Center,
2020.
J. M. Catalpa and J. K. McGuire. Family boundary ambiguity among transgender youth. Family
Relations, 67(1):88–103, 2018.
H. Choi, K. Detyniecki, C. Bazil, S. Thornton, P. Crosta, H. Tolba, and M. Muneeb et al. De-
velopment and validation of a predictive model of drug-resistant genetic generalized epilepsy.
Neurology, 95(15):e2150–e2160, 2020.
K. Clements-Nolle, R. Marx, and M. Katz. Attempted suicide among transgender persons: The
influence of gender-based discrimination and victimization. Journal of Homosexuality, 51(3):
53–69, 2006.
23
D. J. Corliss. Analytics and protected healthcare information. Presentation for Michigan Association
of Community Mental Health Boards https://www.peace-work.org/peace-work-studies, 2015.
US DHS. Handbook for safeguarding sensitive personally identifiable information, 2012.
L. Z. Dubois. Associations between transition-specific stress experience, nocturnal decline in ambu-
latory blood pressure, and c-reactive protein levels among transgender men. American Journal
of Human Biology, 24(1):52–61, 2012.
L. Durwood, K. A. McLaughlin, and K. R. Olson. Mental health and self-worth in socially transi-
tioned transgender youth. Journal of the American Academy of Child and Adolescent Psychiatry,
56(2):116–123, 2017.
A. Fausto-Sterling. Sexing the body: Gender politics and the construction of sexuality. Basic Books,
2000.
T. Fitzsimons. N.j. to become fourth state with gender-neutral birth certificate option. NBC News,
2019.
George Gatewood et al. A monograph on confidentiality and privacy in the us census. US Bureau
of the Census, Washington, 2001.
Z. Goldstein, T. A. Corneil, and D. N. Greene. When gender identity doesn’t equal sex recorded at
birth: the role of the laboratory in providing effective healthcare to the transgender community.
Clinical chemistry, 63(8):1342–1352, 2017.
J. A. Greenberg and R. Stam. Intersexuality and the Law : Why Sex Matters. 2012.
S. Gupta, K. L. Imborek, and M. D. Krasowski. Challenges in transgender healthcare: the pathology
perspective. Laboratory medicine, 47(3):180–188, 2016.
G. Herdt. Third sex, third gender: Beyond sexual dimorphism in culture and history. Princeton
University Press, 2020.
HEW. The belmont report: Ethical principles and guidelines for the protection of human subjects
of research. Available at http://www.nuhresearch.nhg.com.sg/obr/Belmont%20Report.pdf, 1979.
24
I. A. Hughes, C. Houk, S. F. Ahmed, P. A. Lee, and Lawson Wilkins Pediatric Endocrine So-
ciety/European Society for Paediatric Endocrinology. Consensus statement on management of
intersex disorders. Journal of pediatric urology, 2(3):148–162, 2006.
interACT. Statement on intersex terminology. Available at https://interactadvocates.org/
interact-statement-on-intersex-terminology/, 2020.
M. S. Irwig. Cardiovascular health in transgender people. Reviews in Endocrine and Metabolic
Disorders, 19(3):243–251, 2018.
F. Lau, M. Antonio, K. Davison, R. Queen, and A. Devor. A rapid review of gender, sex, and
sexual orientation documentation in electronic health records. Journal of the American Medical
Informatics Association, 27(11):1174–1783, 2020.
M. Morris and E. Cravigan. The privacy, data protection and cybersecurity law review: Australia.
https://thelawreviews.co.uk/title/the-privacy-data-protection-and-cybersecurity-law-review/
australia, 2020.
L. Mullen. gender: Predict gender from names using historical data. Available at https://github.
com/ropensci/gender, 2018. R package version 0.5.2.
K. R. Olson and K. A. McLaughlin. Mental health of transgender children who are supported in
their identities. Pediatrics, 137(3), 2016.
A. N. Onyiriuka, U. Kuhnle-Krahl, E. Sadoh, and J. B. E. Elusiyan. External genital anomalies in
newborns in two nigerian hospitals: A pilot study of the birth prevalence. International Journal
of Child and Adolescent Health, 9(2):187–193, 2016.
E. A. Pascoe and L. Smart Richman. Perceived discrimination and health: A meta-analytic review.
Psychological Bulletin, 135(4):531–554, 2009.
D. W. Riggs and C. Bartholomaeus. Gaslighting in the context of clinical interactions with parents
of transgender children. Sexual and Relationship Therapy, 33(4):382–394, 2018.
25
J. A. Roffee and A. Waling. Resolving ethical challenges when researching with minority and
vulnerable populations: Lgbtiq victims of violence, harassment and bullying. Research Ethics, 13
(1):4–22, 2017.
S. T. Russell, A. M. Pollitt, G. Li, and A. H. Grossman. Chosen name use is linked to reduced
depressive symptoms, suicidal ideation, and suicidal behavior among transgender youth. The
Journal of adolescent health : official publication of the Society for Adolescent Medicine, 63(4):
503–505, 2018.
J. D. Safer, E. Coleman, J. Feldman, R. Garofalo, W. Hembree, A. Radix, and J. Sevelius. Barriers
to health care for transgender individuals. Current opinion in endocrinology, diabetes, and obesity,
23(2):168–168, 2016.
P. Saha-Chaudhuri and C. R. Weinberg. Addressing data privacy in matched studies via virtual
pooling. BMC Medical Research Methodology, 17(1), 2017.
D. Su, J. A. Irwin, C. Fisher, A. Ramos, M. Kelley, D. A. Rogel Mendoza, and J. D. Coleman.
Mental health disparities within the lgbt population: A comparison between transgender and
nontransgender individuals. Transgender Health, 1(1):12–20, 2016.
M. E. Wiesner-Hanks. Gender in history: Global perspectives. John Wiley & Sons, 2011.
A. L. Wirtz, T. C. Proteat, M. Malik, and N. Glass. Gender-based violence against transgender
people in the united states: A call for research and programming. Trauma, Violence, & Abuse,
21(2):227–241, 2020.
26
ResearchGate has not been able to resolve any citations for this publication.
Article
Full-text available
As a stigmatized and vulnerable population, sexual minorities are often assumed to also be a hard-to-survey population. Despite this implicit assumption, there is little empirical evidence on the topic. Using a nationally representative survey that included sexual orientation (the Census Barriers, Attitudes, and Motivators Survey), we examine level of effort, the Census Bureau’s Low Response Score (LRS), and stated intent to respond to the 2020 Census as proxy measures to explore this assumption. We found no evidence that sexual minorities required higher levels of effort to secure participation in the survey. Additionally, we found that compared to straight respondents, lesbians, gays, and bisexuals had a higher intent to respond to the 2020 Census. We surmise the current social climate in the United States may be a contributing factor to these findings.
Article
Full-text available
Objective: The lack of precise and inclusive gender, sex, and sexual orientation (GSSO) data in electronic health records (EHRs) is perpetuating inequities of sexual and gender minorities (SGM). We conducted a rapid review on how GSSO documentation in EHRs should be modernized to improve the health of SGM. Materials and methods: We searched MEDLINE from 2015 to 2020 with terms for gender, sex, sexual orientation, and electronic health/medical records. Only literature reviews, primary studies, and commentaries from peer-reviewed journals in English were included. Two researchers screened citations and reviewed articles with help from a third to reach consensus. Covidence, Excel, and Atlas-TI were used to track articles, extract data, and synthesize findings, respectively. Results: Thirty-five articles were included. The 5 themes to modernize GSSO documentation in EHRs were (1) creating an inclusive, culturally competent environment with precise terminology and standardized data collection; (2) refining guidelines for identifying and matching SGM patients with their care needs; (3) improving patient-provider relationships by addressing patient rights and provider competencies; (4) recognizing techno-socio-organizational aspects when implementing GSSO in EHRs; and (5) addressing invisibility of SGM by expanding GSSO research. Conclusions: The literature on GSSO documentation in EHRs is expanding. While this trend is encouraging, there are still knowledge gaps and practical challenges to enabling meaningful changes, such as organizational commitments to ensure affirming environments, and coordinated efforts to address technical, organizational, and social aspects of modernizing GSSO documentation. The adoption of an inclusive EHR to meet SGM needs is a journey that will evolve over time.
Article
Full-text available
Objective To develop and validate a clinical prediction model for antiepileptic drug-resistant (AED-resistant) genetic generalized epilepsy (GGE). Method We performed a case-control study of patients with and without drug-resistant GGE, nested within ongoing longitudinal observational studies of AED response at 2 tertiary epilepsy centers. Using a validation dataset, we tested the predictive performance of 3 candidate models, developed from a training dataset. We then tested the candidate models' predictive ability on an external testing dataset. Results Of 5,189 patients in the ongoing longitudinal study, 122 met criteria for AED-resistant GGE and 468 met criteria for AED-responsive GGE. There were 66 GGE patients in the external dataset, of whom 17 were cases. Catamenial epilepsy, history of a psychiatric condition, and seizure types were strongly related with drug-resistant GGE case status. Compared to women without catamenial epilepsy, women with catamenial epilepsy had about a four-fold increased risk for AED-DR. The calibration of 3 models, assessing the agreement between observed outcomes and predictions, was adequate. Discriminative ability, as measured with area under the ROC curve (AUC) ranged 0.58–0.65. Conclusion Catamenial epilepsy, history of a psychiatric condition, and the seizure type combination of generalized tonic clonic, myoclonic, and absence seizures are negative prognostic factors of drug-resistant GGE. The AUC of 0.6 is not consistent with truly effective separation of the groups, suggesting other unmeasured variables may need to be considered in future studies to improve predictability.
Article
Full-text available
Background: As of 2016, ≈1.4 million people in the United States identify as transgender. Despite their growing number and increasing specific medical needs, there has been a lack of research on cardiovascular disease (CVD) and CVD risk factors in this population. Recent studies have reported that the transgender population had a significantly higher rate of CVD risk factors without a significant increase in overall CVD morbidity and mortality. These studies are limited by their small sample sizes and their predominant focus on younger transgender populations. With a larger sample size and inclusion of broader age range, our study aims to provide insight into the association between being transgender and cardiovascular risk factors, as well as myocardial infarction. Methods and Results: The Behavioral Risk Factor Surveillance System data from 2014 to 2017 were used to evaluate the cross-sectional association between being transgender and the reported history of myocardial infarction and CVD risk factors. A logistic regression model was constructed to study the association between being transgender and myocardial infarction after adjusting for CVD risk factors including age, diabetes mellitus, hypertension, hypercholesterolemia, chronic kidney disease, smoking, and exercise. Multivariable analysis revealed that transgender men had a >2-fold and 4-fold increase in the rate of myocardial infarction compared with cisgender men (odds ratio, 2.53; 95% CI, 1.14–5.63; P=0.02) and cisgender women (odds ratio, 4.90; 95% CI, 2.21–10.90; P<0.01), respectively. Conversely, transgender women had >2-fold increase in the rate of myocardial infarction compared with cisgender women (odds ratio, 2.56; 95% CI, 1.78–3.68; P<0.01) but did not have a significant increase in the rate of myocardial infarction compared with cisgender men. Conclusions: The transgender population had a higher reported history of myocardial infarction in comparison to the cisgender population, except for transgender women compared with cisgender men, even after adjusting for cardiovascular risk factors.
Article
Full-text available
This review examines the relationship between exogenous sex steroids and cardiovascular events and surrogate markers in trans (transgender) people. Data from trans populations is compared to data from postmenopausal women and hypogonadal men when appropriate. In an age-adjusted comparison with cisgender people, trans people appear to have an increased risk for myocardial infarction and death due to cardiovascular disease. It is uncertain whether hormone therapy in trans people affects their risk of stroke. In studies that followed trans people on hormone therapy, the rates of myocardial infarction and stroke were consistently higher in trans women than trans men. There is strong evidence that estrogen therapy for trans women increases their risk for venous thromboembolism over 5 fold. Extrapolating from studies of hormone therapy in postmenopausal women, transdermal estrogen likely carries a lower risk for venous thromboembolism than oral estrogen. Regarding red blood cells, testosterone therapy increases hemoglobin in trans men, and lowering testosterone in trans women has the opposite effect. Regarding blood pressure, the effects of hormone therapy on systolic blood pressure in trans women are inconsistent, with most studies showing an increase. In trans men, testosterone therapy consistently increases systolic blood pressure and may increase diastolic blood pressure. For lipids, hormone therapy may increase triglycerides in both trans women and men. In trans men, testosterone therapy also may increase LDL-cholesterol and decrease HDL-cholesterol.
Article
Full-text available
Understood as a type of identity-related abuse, gaslighting is a form of manipulation where the perpetrator attempts to convince someone that their thoughts, perceptions, or beliefs are mistaken. In the clinical context, gaslighting is often seen as part of a broader constellation of abuse and violence between adults. However it can also happen in more subtle ways, and can present in ways that are difficult to detect. This paper explores instances where gaslighting may potentially occur in clinical interactions involving parents of transgender children. Three fictionalised case studies are presented drawing on the first author’s clinical work, and demonstrate three overarching forms of gaslighting: 1) deferred action, 2) intentional forgetting, and 3) placing an emotional burden on the child. Having presented the three cases, the paper concludes by discussing how clinicians might identify and respond to gaslighting, emphasising 1) speaking with children on their own, 2) speaking on behalf of children to parents, 3) modelling advocacy to parents, 4) correcting misgendering in the clinical context, and 5) using peer supervision to discuss cases. In sum, the paper argues that whilst clinicians should not rush to “diagnose” gaslighting, clinicians should also not overlook its powerful effects upon transgender young people.
Article
Full-text available
Background Data confidentiality and shared use of research data are two desirable but sometimes conflicting goals in research with multi-center studies and distributed data. While ideal for straightforward analysis, confidentiality restrictions forbid creation of a single dataset that includes covariate information of all participants. Current approaches such as aggregate data sharing, distributed regression, meta-analysis and score-based methods can have important limitations. Methods We propose a novel application of an existing epidemiologic tool, specimen pooling, to enable confidentiality-preserving analysis of data arising from a matched case-control, multi-center design. Instead of pooling specimens prior to assay, we apply the methodology to virtually pool (aggregate) covariates within nodes. Such virtual pooling retains most of the information used in an analysis with individual data and since individual participant data is not shared externally, within-node virtual pooling preserves data confidentiality. We show that aggregated covariate levels can be used in a conditional logistic regression model to estimate individual-level odds ratios of interest. Results The parameter estimates from the standard conditional logistic regression are compared to the estimates based on a conditional logistic regression model with aggregated data. The parameter estimates are shown to be similar to those without pooling and to have comparable standard errors and confidence interval coverage. Conclusions Virtual data pooling can be used to maintain confidentiality of data from multi-center study and can be particularly useful in research with large-scale distributed data.
Conference Paper
The U.S. Census Bureau announced, via its Scientific Advisory Committee, that it would protect the publications of the 2018 End-to-End Census Test (E2E) using differential privacy. The E2E test is a dress rehearsal for the 2020 Census, the constitutionally mandated enumeration of the population used to reapportion the House of Representatives and redraw every legislative district in the country. Systems that perform successfully in the E2E test are then used in the production of the 2020 Census. Motivation: The Census Bureau conducted internal research that confirmed that the statistical disclosure limitation systems used for the 2000 and 2010 Censuses had serious vulnerabilities that were exposed by the Dinur and Nissim (2003) database reconstruction theorem. We designed a differentially private publication system that directly addressed these vulnerabilities while preserving the fitness for use of the core statistical products. Problem statement: Designing and engineering production differential privacy systems requires two primary components: (1) inventing and constructing algorithms that deliver maximum accuracy for a given privacy-loss budget and (2) insuring that the privacy-loss budget can be directly controlled by the policy-makers who must choose an appropriate point on the accuracy-privacy-loss tradeoff. The first problem lies in the domain of computer science. The second lies in the domain of economics. Approach: The algorithms under development for the 2020 Census focus on the data used to draw legislative districts and to enforce the 1965 Voting Rights Act (VRA). These algorithms efficiently distribute the noise injected by differential privacy. The Data Stewardship Executive Policy Committee selects the privacy-loss parameter after reviewing accuracy-privacy-loss graphs.
Article
Purpose: This study aimed to examine the relation between chosen name use, as a proxy for youths' gender affirmation in various contexts, and mental health among transgender youth. Methods: Data come from a community cohort sample of 129 transgender and gender nonconforming youth from three U.S. cities. We assessed chosen name use across multiple contexts and examined its association with depression, suicidal ideation, and suicidal behavior. Results: After adjusting for personal characteristics and social support, chosen name use in more contexts was associated with lower depression, suicidal ideation, and suicidal behavior. Depression , suicidal ideation, and suicidal behavior were lowest when chosen names could be used in all four contexts. Conclusion: For transgender youth who choose a name different from the one given at birth, use of their chosen name in multiple contexts affirms their gender identity and reduces mental health risks known to be high in this group.
Article
Objective To explore family boundary ambiguity in the parent–child relationships of transgender youth. Background Transgender youth may perceive a lack of clarity about whether parents will accept their authentic gender expression, continue to support them physically and emotionally, and regard them as a member of the family. Uncertainty about being in or out of the family and whether family relationships endure is stressful and can lead to psychological distress, a sense of ambiguous loss, and frozen grief. Method Ethnographic content analysis was conducted based on interviews with 90 transgender youth recruited from community centers in 10 regions across 3 countries. Results Narratives revealed that transgender youth experienced family boundary ambiguity related to relational ambiguity, structural ambiguity, and identity ambiguity. Each experience of ambiguity obscured whether participants remained in the family and interpersonally connected to their parents. Conclusion Transgender youth actively navigated complex and ambiguous parent–child relationships whereby participants attempted to reconcile their need for authentic gender expression combined with their need for family connectedness and acceptance. Implications Family clinicians, educators, and policymakers are urged to consider family and transgender resilience through a lens of ambiguous loss and to promote a gender‐affirmative life‐span approach to clinical care for transgender individuals and their families.