Abstract
Objective
Two-step questions to assess gender identity are recommended for optimizing care delivery for gender-diverse individuals. As gender identity fields are increasingly integrated into electronic health records, guidance is needed on how to analyze these data. The goal of this study was to assess potential approaches for analyzing 2-step gender identity questions and the impact of each on suicidal ideation.
Materials and Methods
A regional Youth Risk Behavior Survey in one Northeastern school district used a 2-step question to assess gender identity. Three gender measurement strategies (GMSs) were used to operationalize gender identity, (1) combining all gender-diverse youth (GDY) into one category, (2) grouping GDY based on sex assigned at birth, and (3) categorizing GDY based on binary and nonbinary identities. Mixed-effects logistic regression was used to compare odds of suicidal ideation between gender identity categories for each GMS.
Results
Of the 3010 participants, 8.3% were GDY. Subcategories of GDY had significantly higher odds (odds ratio range, 1.6-2.9) of suicidal ideation than cisgender girls regardless of GMS, while every category of GDY had significantly higher odds (odds ratio range, 2.1-5.0) of suicidal ideation than cisgender boys.
Conclusions
The field of clinical informatics has an opportunity to incorporate inclusive items like the 2-step gender identity question into electronic health records to optimize care and strengthen clinical research. Analysis of the 2-step gender identity question impacts study results and interpretation. Attention to how data about GDY are captured will support for more nuanced, tailored analyses that better reflect unique experiences within this population.
Keywords: gender identity, transgender persons, informatics, suicidal ideations, adolescent
INTRODUCTION
Gender identity is much more expansive than the binary “male” and “female” categories corresponding with external anatomy visualized at birth. This broader understanding of human experience, that is inclusive of transgender, nonbinary, and other gender-diverse identities, leads to challenges in reflecting gender identity in research.1–4 Gender-diverse people are those whose gender identity and sex assigned at birth do not fully align. Accurate assessment of gender identity that is reflective of the lived experience of gender-diverse people is critical given the significant health disparities this group faces.5–8 Since 2018, electronic health records (EHRs) have been required to have the capability to collect data on gender identity under Meaningful Use Criteria through the Centers for Medicare and Medicaid Services.9 Thus, collecting information about gender identity, most often alongside sexual orientation and deemed SOGI, for sexual orientation/gender identity, or GSSO, for gender, sex, and sexual orientation, is a growing area of informatics research.10 Despite the required inclusion of these data in the EHR, current strategies lack standardization with regard to what information is collected and how it should be used for research.9,10
Initially developed in 1997 by the Transgender Health Advocacy Coalition, the 2-step method of assessing gender identity asked about gender identity (eg, “What is your gender identity?”) and sex assigned at birth (eg, “What was your sex assigned at birth?”) separately.11 This 2-step model has since undergone significant evaluation and has been found to be superior to single questions (eg, “What is your gender identity? Female, male, transgender, other”) because it allows researchers to classify participants as gender diverse by identifying an incongruence in sex and gender identity even if individuals do not identify with specific terms like “transgender.”11,12 The 2-step method has been validated in gender-diverse youth (GDY).6,12–15 It has also been recommended by the Gender Identity in US Surveillance Group through the Williams Institute of the University of California, Los Angeles, the Center for Excellence in Transgender Health at the University of California, San Francisco, and the World Professional Association of Transgender Health Electronic Medical Record Working Group.15–19
Recent reports encourage careful examination of the research question to structure questions assessing gender identity. In addition, many studies and EHRs have begun to include a 2-step gender identity item.20–22 The 2-step gender identity questions remain an imperfect measure with controversy around question order,14 potential inclusion of “intersex” as an option for sex assignment,1 the number and content of gender identity response options,1 the “select all that apply” clause for gender identity,6,16 and if “other” with or without a free response option should be included.3,12,17 Despite these limitations, use of the 2-step question has been associated with higher yield in identification of gender-diverse individuals as well as with increased community acceptability.12,16
Despite this growing body of research on 2-step gender identity questions, there remains little guidance regarding how to operationalize and subsequently analyze the data collected as well as implications for future data collection through EHRs through use of SOGI/GSSO forms. While some authors challenge the inclusion of a multitude of gender identity options in noting that we can never include all possible responses,23 there also remains the ethical concern that by not reflecting the lived experiences of gender-diverse people, we will continue to exclude them from research, thus failing to address known and unknown health disparities.24 There is also the methodological concern that comparisons across studies are difficult due to lack of consistent response options. Typically, these concerns are raised when question responses are collapsed into gender identity categories in analyses. This may be done by grouping all individuals with nonbinary identities into a single category or collapsing them into existing assigned sex or binary categories. Although such collapsing of categories is common practice in research, there is no clear guidance on how to analyze these new groups. Some investigators choose to categorize participants based on their sex assignment,25,26 while others do so based on the binary vs nonbinary nature of participants’ gender identity.27–29 Additionally, analytic strategy is often born out of considerations of sample size and convenience, and thus fails to accurately reflect differences within groups due to how results are organized. The goals of this study were to (1) assess the ways in which various approaches for analyzing the 2-step gender question impact one outcome of interest (suicidal ideation in the prior 12 months) and (2) propose a framework for analyzing data generated from a 2-step gender identity question.
MATERIALS AND METHODS
Study design
A regionally modified version of the Centers for Disease Control and Prevention’s Youth Risk Behavior Survey was administered to all high school students in grades 9 to 12 in a Northeastern midsized city school district comprised of 13 high schools and approximately 5200 students. Paper surveys and bubbled answer sheets were provided, and students anonymously completed surveys in October 2018. Overall, 4730 students completed surveys (91% response rate).
Measures
Gender identity
Gender identity was assessed using the 2-step method by asking participants: (1) “What is your sex (the sex you were assigned at birth, on your birth certificate)?” (response options included “female” and “male”); and (2) “Which of the following best describes you? Select all that apply.” (response options included “girl,” “boy,” “trans girl,” “trans boy,” “genderqueer,” “nonbinary,” and “another identity”) (Figure 1). We coded participants’ gender identity using multiple gender measurement strategies (GMSs) in order to examine the effects of each approach on our outcome of interest, suicidal ideation. For all approaches described subsequently, participants who selected “female” assignment at birth and the singular response option of “girl” for gender identity were coded as cisgender girls. Participants who selected “male” sex assignment and the singular response option of “boy” as their gender identity were coded as cisgender boys. GDY were then coded using 3 overarching GMSs:
Figure 1.
Two-step gender identity question components.
GMS 1: Gender Diverse Youth as a Monolith. In this first GMS, we coded all participants with any incongruence in sex assignment and gender identity into a single category of GDY.
-
GMS 2: Gender Diverse Youth Based on Sex Assignment. In this GMS, we coded all GDY into 2 separate groups: (1) GDY who were assigned “female” at birth, which included people who reported a “female” sex assignment and had a gender identity of “boy,” “trans boy,” genderqueer,” “nonbinary,” and/or “another identity”; and (2) GDY who were assigned “male” at birth, which included people who reported a “male” sex assignment and had a gender identity of “girl,” “trans girl,” genderqueer,” “nonbinary,” and/or “another identity.”
This GMS highlighted an important incongruence that needed to be accounted for. A small number of participants indicated a gender-diverse identity (eg, “trans boy”), but with a congruent sex assigned at birth (eg, “male”). We suspect that these participants intentionally selected an alternative sex assignment that aligned with their gender identity and, based on their selection of “trans girl” or “trans boy,” included them in the gender-diverse category. To account for potential causes of this unexpected congruence, we included 3 total groupings in GMS 2 (GMS 2.1, GMS 2.2, and GMS 2.3). GMS 2.1 includes these groups of individuals (“assigned female trans girls [AFTGs]” and “assigned male trans boys [AMTBs]”) with their sex assigned at birth. In GMS 2.2, we created 2 additional categories for AFTGs and AMTBs. Based on our concern that this may have been an intentional misrepresentation of sex assignment, in GMS 2.3 we reversed the sex assignment provided by the respondent for AFTG and AMTB individuals to align with their designated gender identity.
GMS 3: GDY Based on Binary and Nonbinary Gender Identities. For this third GMS, we coded GDY based on binary and nonbinary gender identities. We created a variable that classified GDY as (1) exclusively binary, which included GDY who only selected gender response options of “girl,” “trans girl,” “boy,” or “trans boy”; (2) exclusively nonbinary, which included GDY who only selected gender response options of “genderqueer,” “nonbinary,” or “another identity”; and (3) both binary and nonbinary, which included GDY who selected at least 1 binary gender identity as well as at least 1 nonbinary identity as described previously. From this variable (denoted as GMS 3.1), we created 2 new variables, GMS 3.2 and GMS 3.3. The first, GMS 3.2, included both GDY who identified as exclusively nonbinary as well as those who identified as both binary and nonbinary. This variable was labeled binary exclusive or nonbinary inclusive. The second, GMS 3.3, included both GDY who identified as exclusively binary and those who identified as both binary and nonbinary. This variable was labeled nonbinary exclusive or binary inclusive.
Other demographic variables
Sexual orientation was assessed by the question, “Which of the following best describes you?” Response options were heterosexual (straight), mostly heterosexual (mostly straight), gay or lesbian, bisexual, queer, asexual, and not sure. For the purposes of this analysis, we considered all who responded with any answer except for “heterosexual (straight)” as a sexual minority. Age was assessed by the question, “How old are you?” Response options were 12 years old or younger, 13 years old, 14 years old, 15 years old, 16 years old, 17 years old, or 18 years old or older. Race and ethnicity were assessed by 2 questions. The first was, “Are you Hispanic/Latino?” Response items were “yes” or “no.” The second was, “What is your race? (Select one or more responses)” Response options were American Indian or Alaska Native, Asian, Black or African American, Native Hawaiian or other Pacific Islander, White, and Other.
Outcome variable
The outcome in this analysis was suicidal ideation in the prior year assessed by response to the following question: “During the past 12 months, did you ever seriously consider attempting suicide?” Response options were “yes” or “no.” This item was chosen because it has previously demonstrated good test-retest reliability and prior literature demonstrating disparities in suicidal ideation by gender identity.8,30
Analysis
Of the 4730 returned surveys, 243 were considered mischievous responders due to observable patterns in bubbled answers or completing fewer than 20 items, while 37 were deemed unreadable. These surveys were excluded from analysis. We further excluded 1197 surveys, as these participants were missing a response to at least 1 question item crucial to this study (sex assigned at birth, gender identity, sexual orientation, age, race/ethnicity, or suicidal ideation over the prior 12 months), rendering our total analytical sample to include 3010 respondents.
First, we described the demographics of the overall sample and the sample of GDY, and examined these differences in sexual orientation, race and ethnicity, and age using chi-square tests. Second, for each GMS we examined the prevalence of gender-diverse identity subgroups in both the total sample and among youth identified as GDY. Third, we used separate mixed-effects logistic regression models (accounting for the clustering of students within schools with a random intercept) using GMS subcategory as the primary predictor and suicidal ideation as the outcome. We adjusted for sexual orientation, age, and race and ethnicity as they have been associated31,32 with our primary exposure, gender identity, and outcome (suicidal ideation over the prior 12 months), in an effort to control for potential confounding.
Fourth, based on these multivariable models, we estimated the adjusted prevalence and 95% confidence intervals (CIs) of our suicidal ideation outcome for each GMS using marginal effect estimation. Models were initially fit with cisgender girls as the reference group, as prior literature suggests that cisgender girls have higher rates of suicidal ideation than cisgender boys,8 making it a more conservative comparison. We then ran the same models with cisgender boys as the reference group to assess the impact of reference group change on our primary outcome, suicidal ideation. A significance level of α = 0.05 was assumed, and no adjustments were made for multiplicity. All analyses were conducted in StataSE 15.1 (StataCorp, College Station, TX).
RESULTS
Prevalence of GDY
Of the 3010 respondents included in this analysis, 1585 (52.7%) were cisgender girls, 1176 (39.1%) were cisgender boys, and 249 (8.3%) were gender diverse (GMS 1). Aside from “girl” (55.6%) and “boy” (40.8%), the most selected identity was “trans boy” (1.5%), followed by “nonbinary” (1.3%), “trans girl” (1.1%), “genderqueer” (0.9%), and “another identity” (0.8%). While “another identity” was the least commonly selected, most who selected this (73.1%) did not select any other option. Gender identity was significantly associated with sexual orientation (P < .0001) and race (P < .001), such that GDY were most likely to be sexual minorities and least likely to be non-Hispanic White (Table 1). Age did not differ among cisgender girls, cisgender boys, and GDY (P = .296).
Table 1.
Demographics of total sample and by gender identity
| Total (N = 3010) |
Cisgender Girls (n = 1585) |
Cisgender Boys (n = 1176) |
Gender-Diverse Youth (n = 249) |
P Value | |||||
|---|---|---|---|---|---|---|---|---|---|
| n | % | n | % | n | % | n | % | ||
| Sexual orientation | |||||||||
| Heterosexual | 2067 | 68.67 | 987 | 62.27 | 970 | 82.50 | 110 | 44.18 | P < .001 |
| Sexual minoritya | 943 | 31.33 | 598 | 37.73 | 206 | 17.52 | 139 | 55.82 | |
| Race/ethnicity | |||||||||
| White | 1265 | 42.03 | 640 | 40.38 | 539 | 45.83 | 86 | 34.50 | P < .001 |
| Black | 928 | 30.83 | 539 | 34.01 | 306 | 26.02 | 83 | 33.33 | |
| Mixed race | 408 | 13.55 | 213 | 13.44 | 162 | 13.78 | 13 | 5.22 | |
| Hispanic | 271 | 9.00 | 125 | 7.89 | 114 | 9.69 | 32 | 12.85 | |
| Other race | 138 | 4.58 | 68 | 4.29 | 55 | 4.68 | 35 | 14.06 | |
| Age | |||||||||
| <14 y | 26 | 0.86 | 10 | 0.63 | 12 | 1.02 | 4 | 1.61 | P = .296 |
| 14 y | 554 | 18.41 | 294 | 18.55 | 215 | 18.28 | 45 | 18.07 | |
| 15 y | 732 | 24.32 | 386 | 24.35 | 278 | 23.64 | 68 | 27.31 | |
| 16 y | 769 | 25.55 | 410 | 25.87 | 307 | 26.12 | 52 | 20.88 | |
| 17 y | 740 | 24.58 | 391 | 24.67 | 289 | 24.57 | 60 | 24.10 | |
| 18+ y | 189 | 6.28 | 94 | 5.93 | 75 | 6.38 | 20 | 8.03 | |
Sexual minority was defined as selection of any sexual orientation other than “heterosexual (straight).”
When examining GDY based on sex assignment (GMS 2) (Table 2), we found that 50.6% of GDY were assigned female and 49.4% were assigned male at birth. Additionally, 26 individuals (10.4% of GDY) indicated a sex assigned at birth congruent with their gender identity (eg, “male” and “trans boy” or “female” and “trans girl”) and were classified as AMTB or AFTG. When examining GDY by binary and nonbinary identities, we found that 69.1% of GDY were exclusively binary, 22.1% were exclusively nonbinary, and 8.8% had both binary and nonbinary gender identities.
Table 2.
Prevalence of gender identity subgroups in total sample and among GDY
| GMS | Prevalence in Total Sample (N = 3010) |
Prevalence Among GDY (n = 249) |
||
|---|---|---|---|---|
| n | % | n | % | |
| GMS 1 | ||||
| Cisgender girls | 1585 | 52.70 | 0 | 0 |
| Cisgender boys | 1176 | 39.10 | 0 | 0 |
| GDY | 249 | 8.27 | 249 | 100 |
| GMS 2.1a | ||||
| Gender-diverse AFAB | 126 | 4.19 | 126 | 50.60 |
| Gender-diverse AMAB | 123 | 4.09 | 123 | 49.40 |
| GMS 2.2 | ||||
| Gender-diverse AFAB | 113 | 3.75 | 113 | 45.40 |
| Gender-diverse AMAB | 110 | 3.65 | 110 | 44.20 |
| Gender-diverse AFTGs | 13 | 0.43 | 13 | 5.20 |
| Gender-diverse AMTBs | 13 | 0.43 | 13 | 5.20 |
| GMS 2.3a | ||||
| Gender-diverse AFAB | 126 | 4.19 | 126 | 50.60 |
| Gender-diverse AMAB | 123 | 4.09 | 123 | 49.40 |
| GMS 3.1 | ||||
| Gender-diverse binary exclusive | 172 | 5.71 | 172 | 69.10 |
| Gender-diverse nonbinary exclusive | 55 | 1.83 | 55 | 22.10 |
| Gender-diverse both binary and nonbinary | 22 | 0.73 | 22 | 8.80 |
| GMS 3.2 | ||||
| Gender-diverse binary inclusive | 185 | 6.15 | 185 | 74.30 |
| Gender-diverse nonbinary exclusive | 64 | 2.13 | 64 | 25.70 |
| GMS 3.3 | ||||
| Gender-diverse binary exclusive | 172 | 5.55 | 172 | 69.10 |
| Gender-diverse nonbinary inclusive | 77 | 2.56 | 77 | 30.90 |
AFAB: assigned female at birth; AFTG: assigned female trans girl; AMAB: assigned male at birth; AMTB: assigned male trans boy; GDY: gender-diverse youth, GMS: gender measurement strategy.
GMS 2.1 and 2.3 are identical because there were equal numbers of AFTGs and AMTBs.
Suicidal ideation by gender identity
Of the total sample, 21.3% indicated that they had seriously considered attempting suicide in the prior 12 months. In GMS 1, GDY had higher odds of suicidal ideation compared with cisgender girls (odds ratio [OR], 1.55; 95% CI, 1.15-2.08) (Table 3). Additionally, the adjusted prevalence of suicidal ideation using GMS 1 was 15.2% for cisgender boys, 23.5% for cisgender girls, and 31.7% for GDY.
Table 3.
Gender identity differences in suicidal ideation: Results from multivariable logistic regression models
| Model | Gender Identity Subgroup | OR | 95% CI |
|---|---|---|---|
| GMS 1 | Cisgender girls | 1.00 | Reference group |
| Cisgender boys | 0.57a | 0.46-0.70a | |
| Gender-diverse youth | 1.55a | 1.16-2.08a | |
| GMS 2.1 | Cisgender girls | 1.00 | Reference group |
| Cisgender boys | 0.57a | 0.46-0.70a | |
| Gender-diverse AFABs | 1.92a | 1.30-2.83a | |
| Gender-diverse AMABs | 1.23 | 0.81-1.86 | |
| GMS 2.2 | Cisgender girls | 1.00 | Reference group |
| Cisgender boys | 0.57a | 0.46-0.70a | |
| Gender-diverse AFAB | 1.92a | 1.28-2.89a | |
| Gender-diverse AMAB | 1.16 | 0.75-1.82 | |
| Gender-diverse AFTGs | 1.87 | 0.58-6.00 | |
| Gender-diverse AMTBs | 1.78 | 0.58-5.39 | |
| GMS 2.3 | Cisgender girls | 1.00 | Reference group |
| Cisgender boys | 0.57a | 0.46-0.70a | |
| Gender-diverse AFAB | 1.94a | 1.32-2.86a | |
| Gender-diverse AMAB | 1.21 | 0.80-1.83 | |
| GMS 3.1 | Cisgender girls | 1.00 | Reference group |
| Cisgender boys | 0.57a | 0.46-0.70a | |
| Gender-diverse binary exclusive | 1.52a | 1.07-2.15a | |
| Gender-diverse nonbinary exclusive | 1.29 | 0.73-2.28 | |
| Gender-diverse both binary and nonbinary | 2.85a | 1.18-6.91a | |
| GMS 3.2 | Cisgender girls | 1.00 | Reference group |
| Cisgender boys | 0.57a | 0.46-0.70a | |
| Gender-diverse binary inclusive | 1.57a | 1.12-2.20a | |
| Gender-diverse nonbinary exclusive | 1.50 | 0.89-2.55 | |
| GMS 3.3 | Cisgender girls | 1.00 | Reference group |
| Cisgender boys | 0.57a | 0.46-0.70a | |
| Gender-diverse binary exclusive | 1.52a | 1.07-2.15a | |
| Gender-diverse nonbinary inclusive | 1.62a | 1.00-2.63a |
The data reflect responses to the question, “During the past 12 months, did you ever seriously consider attempting suicide?” All models adjusted for age, race/ethnicity, and sexual orientation.
AFAB: assigned female at birth; AFTG: assigned female trans girl; AMAB: assigned male at birth; AMTB: assigned male trans boy; CI: confidence interval; GMS: gender measurement strategy; OR: odds ratio.
P < .05.
When examining GDY based on sex assignment (GMS 2.1-2.3), all assigned-female-at-birth groups, regardless of AMTB or AFTG status, had significantly higher odds of seriously considering suicide compared with cisgender girls (GMS 2.1: OR, 1.92; 95% CI, 1.30-2.83; GMS 2.2: OR, 1.92; 95% CI, 1.28-2.89; GMS 2.3: OR, 1.94; 95% CI, 1.32-2.86).
When examining suicidal ideation inequities for GDY based on binary and nonbinary identities (GMS 3.1-3.3), there were notable differences within the 3 categorizations. In GMS 3.1, GDY with exclusively binary identities and those with both binary and nonbinary identities had a 1.52 and 2.85 times higher odds of considering suicide than cisgender girls (OR, 1.52; 95% CI, 1.07-2.15, and OR, 2.85; 95% CI, CI, 1.18-6.91), respectively. In GMS 3.2, only participants with inclusive binary identities considered suicide at significantly higher rates than cisgender girls (OR, 1.57; 95% CI, 1.12-2.20). In GMS 3.3, both the binary exclusive and nonbinary inclusive groups had considered suicide at significantly higher rates than cisgender girls (OR, 1.52; 95% CI, 1.07-2.15; and OR, 1.62; 95% CI, 1.00-2.63), respectively. Finally, adjusted predicted prevalence of suicidal ideation was then calculated for each GMS (Table 4).
Table 4.
Adjusted predicted prevalence of suicidal ideation by gender identity
| GMS | % | 95% CI |
|---|---|---|
| GMS 1 | ||
| Cisgender girls | 23.47 | 20.44-26.50 |
| Cisgender boys | 15.18 | 12.52-17.85 |
| Gender-diverse youth | 31.70 | 25.65-37.75 |
| GMS 2.1 | ||
| Gender-diverse AFAB | 36.16 | 27.68-44.63 |
| Gender-diverse AMAB | 27.13 | 19.33-34.94 |
| GMS 2.2 | ||
| Gender-diverse AFAB | 36.23 | 27.35-45.12 |
| Gender-diverse AMAB | 26.16 | 18.01-34.31 |
| Gender-diverse AFTGs | 35.64 | 10.48-60.82 |
| Gender-diverse AMTBs | 34.53 | 10.87-58.18 |
| GMS 2.3 | ||
| Gender-diverse AFAB | 36.42 | 27.93-44.91 |
| Gender-diverse AMAB | 26.83 | 19.05-34.62 |
| GMS 3.1 | ||
| Gender-diverse binary exclusive | 31.26 | 24.15-38.36 |
| Gender-diverse nonbinary exclusive | 28.05 | 17.12-38.98 |
| Gender-diverse both binary and nonbinary | 45.15 | 24.53-65.77 |
| GMS 3.2 | ||
| Gender-diverse binary inclusive | 31.94 | 25.02-38.86 |
| Gender-diverse nonbinary exclusive | 31.07 | 20.39-41.74 |
| GMS 3.3 | ||
| Gender-diverse binary exclusive | 31.27 | 24.15-38.39 |
| Nonbinary inclusive | 32.61 | 22.56-42.66 |
The data reflect responses to the question, “During the past 12 months, did you ever seriously consider attempting suicide?” All models adjusted for age, race/ethnicity, and sexual orientation.
AFAB: assigned female at birth; AFTG: assigned female trans girl; AMAB: assigned male at birth; AMTB: assigned male trans boy; CI: confidence interval; GMS: gender measurement strategy.
When changing the reference group from cisgender girls to cisgender boys, all GMS subgroups had significantly higher odds of considering suicide in the prior 12 months by a conversion factor of 1.76 (see Supplementary Table). This was most notable for GMS 3.1 where those GDY with both binary and nonbinary identities had more than 5 times greater odds of considering suicide than cisgender boys (OR, 5.01; 95% CI, 2.04-12.27).
DISCUSSION
We assessed how varying approaches for analyzing the 2-step gender question related to our outcome of interest, suicidal ideation, in the prior 12 months. Specifically, we wanted to determine whether assigned sex, the binary or nonbinary nature of gender identity, or reference group choice impacted interpretation of our findings. We used a version of the 2-step gender identity question that allowed for identification of youth with incongruent sex assignment and gender identity. This allowed us to be more inclusive in designating youth as gender diverse than has been done previously using single-item questions.
GMS strategy 1, which includes all GDY in a single category, builds on prior studies8 demonstrating significantly higher mental health disparities among GDY when compared with both their cisgender girl and cisgender boy peers. While comparing all GDY obscures important differences demonstrated in our subsequent analyses, it continues to be used as a strategy for datasets with small sample sizes. We argue that, whenever possible if sample size allows, researchers should further explore differences within this broad category to better reflect the lived experiences of GDY.
Analyzing gender identity based on sex assigned at birth is challenging, as this system of classification can feel invalidating to many gender-diverse individuals. We included this GMS in our analysis, as it continues to be used in research and sought to understand differences within gender-diverse groups as opposed to grouping all gender minorities into a single category.25,26 In our regression models, those who were assigned female at birth had significantly higher rates of suicidal ideation compared with cisgender girls, suggesting that sex assigned at birth was associated with our outcome of interest. This was true regardless of how AFTG and AMTB participants were designated. The AFTG and AMTB group, who expressed both a gender-diverse identity (“trans girl” or “trans boy,” respectively) and an aligned sex assigned at birth (“girl” or “boy,” respectively), may have misunderstood the meaning of the terms “trans girl” or “trans boy,” but prior studies, including those with youth in our own region,33 suggest that young people understand this language. That led us to suspect that this small portion of youth may have rejected selecting their sex assigned at birth and instead selected that which best aligned with their affirmed gender identity. This may have been compounded by item placement in the distributed survey, as the items in the 2-step gender identity question were not sequential and instead were separated by several pages of unrelated items.
Analyzing gender identity based on the binary or nonbinary nature may be perceived as a more affirming reflection of identity for gender-diverse individuals and is a newer practice.28,29 Small sample sizes often limit the ability to separate out individuals with nonbinary identities, but because of the relatively large subset of GDY in our sample who expressed a gender-diverse identity, we were able to explore this further. We developed categorization strategies by introducing “exclusive” and “inclusive” terms. For GMS 3.1, we categorized gender-diverse participants into exclusively binary or nonbinary groups and then used a third category for those who expressed both binary and nonbinary identities. Surprisingly, our data suggest that the group of youth who endorse both binary and nonbinary identities had the highest odds of suicidal ideation in all of the models. We suspect that youth whose identities fit less neatly into categories like binary and nonbinary may have more gender fluidity as well as different or compounded minority stressors than other gender-diverse peers, particularly when narratives typically center static gender-diverse experiences. This is in line with findings in sexual minority research that suggest youth who identify as “mostly heterosexual,” also not fitting neatly into a category, experienced increased health disparities compared with youth who identify clearly as nonheterosexual.34 This finding, if replicated with other health outcomes, suggests that future studies would benefit from conducting analyses that allow us to better understand the unique experiences and subsequent disparities faced by GDY who identify as both binary and nonbinary. In GMS 3.2 and 3.3, we included all youth who identified with a binary identity as “binary inclusive” (3.2) or all who identified with a nonbinary identity as “nonbinary inclusive” (3.3) and found that in both models (as well as in GMS 3.1) the group with binary identities had significantly higher odds of suicidal ideation when compared with cisgender girls. We feel that these findings illustrate the need to consider examining the experiences of binary, nonbinary, and both binary and nonbinary individuals given the potential for differences in health disparities experienced between these groups.
Changing the reference group in our models from cisgender girls to cisgender boys resulted in all GMSs having significantly higher odds of suicidal ideation compared with cisgender boys. This was unsurprising, given prior literature demonstrating that cisgender girls have higher rates of suicidal ideation than cisgender boys.8 We initially chose to use cisgender girls as the reference group for the models, as we desired a more conservative estimate of potential differences in our GMS groups. Technically, there is no difference between these 2 models (Table 4; Supplementary Table), as all results obtained from one model can be mathematically converted to results from the other by a constant factor. Despite this model equivalence, our interpretation of the ORs will differ, particularly for those less versed in this proportionality. For example, reporting that youth with both binary and nonbinary gender identities (GMS 3.1) have 5 times higher odds of suicidal ideation than cisgender boys may carry a different weight than saying that this group has 2.85 times higher odds of suicidal ideation than cisgender girls, though both are accurate reflections of these data. Alternatively, researchers could use the overall average as a reference category and subsequently decentralize cisgender experience, which would be more appropriate in studies not seeking to focus on this comparison.
Overall, when considering the potential impact of assigned sex, binary and nonbinary identities, and reference group on our interpretation of our regression models, assigned sex and binary and nonbinary identities demonstrated differences in level of disparity, while reference group impacted only our interpretation of the magnitude of these disparities. These findings may not be consistent for other outcomes, and likely are not consistent for all mental health disparities, but researchers should consider these methodological approaches when examining health disparities for GDY and be both thoughtful and transparent in their analytic approach for categorizing gender identity. To do otherwise risks overlooking more nuanced findings by categorizing young people into broad groups that do not fully reflect their lived experiences. There are also ethical considerations in collecting data with the intention of collapsing and subsequently erasing diversity in the analysis. This has been explored in the context of gender identity, and prior authors have recommended pursuing both accuracy and sensitivity in expressing the diversity of this population so as to avoid imposing assumed or White or Western views of gender identity on the national or international population.35 Additionally, there has been a push for inclusion of gender-diverse stakeholders to guide development, conduct, and interpretation of research in this growing field.36
The ethical and methodological implications of this become more pronounced when considering ongoing integration of SOGI/GSSO forms and fields into EHRs and use of EHR data collected via 2-step gender identity. Informatics research needs to keep pace with the integration of these vitally important data with the goal of optimizing health care and outcomes for this population experiencing marginalization. These findings suggest that the presence of both gender identity and sex assignment at birth allows for a rich analysis and discussion of the role of gender identity in health and health disparities and supports tailored, personalized medicine that appreciates the diversity within this population.
This study is limited in that we used data from a single urban school district, and thus our results may not be broadly generalizable. Additionally, there may be regional and socioeconomic factors contributing to our results that we are unable to account for. We also had a relatively large number of GDY in the sample but not so many that we could develop GMS to be fully inclusive of all gender diversity reflected through our use of the 2-step gender identity question. Our 2-step gender identity question is also limited in that youth may feel their gender identity was not represented on the list of options, as indicated in that most who selected “another identity” did not select other options to describe their gender identity. Notably, this is a limitation of the 2-step gender identity question more broadly.23 Some may have felt that inclusion of “trans girl” and “trans boy” in addition to “girl” and “boy” was othering, though this allowed for exploration of those who indicated unexpectedly congruent gender identities (eg, AFTG/AMTB). Also, our questions were not in series but instead separated by several other questions on the distributed survey, which may have impacted responses. We recommend that future studies using this methodology incorporate the questions sequentially to avoid this limitation. Future studies should also consider the impact of intersectionality of race and ethnicity as well as sexual orientation for gender-diverse people, as such lived experiences are critical in our understanding of health disparities37 but were considered beyond the scope of this study. An additional limitation is our use of a school-based survey to explore GMS analysis and its relevance for clinical informatics research. The growth of research with transgender or gender-diverse individuals, particularly through use of SOGI or GSSO data, requires consideration of overlaps in these related research domains as well as reevaluation of our analytic framework for data collected within clinical care including EHRs.
CONCLUSION
The field of clinical informatics has a unique opportunity to intentionally incorporate inclusive items like the 2-step gender identity question into EHRs to capture diverse gender identities. Ongoing attention to the diversity among GDY, including in clinical research, will support efforts to tailor and personalize clinical care and improve health outcomes that disproportionately impact GDY.
Supplementary Material
ACKNOWLEDGMENTS
We thank the Allegheny County Health Department for their role in data collection and for the use of these data. We are grateful to Pittsburgh Public School Board Leadership for their collaboration.
FUNDING
This research was supported by the National Institutes of Health through the National Institute of Alcohol Abuse and Alcoholism (K01AA027564 [to RWSC]) as well as the National Center for Advancing Translational Science of the NIH (TL1TRR1858; PI Kraemer, for KMK). The content is solely the responsibility of the authors and does not necessarily represent the official views of the National Institutes of Health. The National Institutes of Health was not involved in the study design the writing of the protocol, or the decision to submit for publication. This research was also supported by the Seattle Children’s’ Research Institute Career Development Award. The data collection was supported by the Heinz Endowments and the Grable Foundation.
AUTHOR CONTRIBUTIONS
KMK conceptualized and designed the study, conducted the analysis, drafted the initial manuscript, and revised the manuscript. GMS, SDR, AK, KS, EM, and RWSC, as well as TP, provided substantial contributions to conception, design, analysis, and interpretation of the data and revised the article critically. All authors approved the final manuscript as submitted and agree to be accountable for all aspects of the work.
SUPPLEMENTARY MATERIAL
Supplementary material is available at Journal of the American Medical Informatics Association online
DATA AVAILABILITY STATEMENT
The data underlying this article cannot be shared publicly due to concern for the privacy of individuals that participated in the study. The data will be shared on reasonable request to the corresponding author.
COMPETING INTERESTS STATEMENT
The authors have no competing interests to disclose.
References
- 1. Bauer GR, Braimoh J, Scheim AI, Dharma C. Transgender-inclusive measures of sex/gender for population surveys: Mixed-methods evaluation and recommendations. PLoS One 2017; 12 (5): e0178043. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 2. Lindqvist A, Bäck EA, Bäck H, Sendén MG. Measuring gender in surveys. Paper presented at the Gender Diversity in Survey Research Workshop; June 11-12, 2018; Gothenburg, Sweden.
- 3. Lindqvist A, Sendén MG, Renström EA. What is gender, anyway: a review of the options for operationalising gender. Psychol Sexual 2020. Feb 18 [E-pub ahead of print]. [Google Scholar]
- 4. Puckett JA, Brown NC, Dunn T, Mustanski B, Newcomb ME. Perspectives from transgender and gender diverse people on how to ask about gender. LGBT Health 2020; 7 (6): 305–11. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 5. Conron KJ, Scout, Austin SB. Everyone has a right to, like, check their box:” Findings on a measure of gender identity from a cognitive testing study with adolescents. J LGBT Health Res 2008; 4 (1): 1–9. [DOI] [PubMed] [Google Scholar]
- 6. Conron KJ, Landers SJ, Reisner SL, Sell RL. Sex and gender in the US health surveillance system: a call to action. Am J Public Health 2014; 104 (6): 970–6. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 7. Frohard‐Dourlent H, Dobson S, Clark BA, Doull M, Saewyc EM. I would have preferred more options: accounting for non‐binary youth in health research. Nurs Inq 2017; 24 (1): e12150. [DOI] [PubMed] [Google Scholar]
- 8. Johns MM, Lowry R, Andrzejewski J, et al. Transgender identity and experiences of violence victimization, substance use, suicide risk, and sexual risk behaviors among high school students—19 states and large urban school districts, 2017. MMWR Morb Mortal Wkly Rep 2019; 68 (3): 67–71. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 9. Bosse JD, Leblanc RG, Jackman K, Bjarnadottir RI. Benefits of implementing and improving collection of sexual orientation and gender identity data in electronic health records. Comput Inform Nurs 2018; 36 (6): 267–74. [DOI] [PubMed] [Google Scholar]
- 10. Lau F, Antonio M, Davison K, Queen R, Devor A. A rapid review of gender, sex, and sexual orientation documentation in electronic health records. J Am Med Inform Assoc 2020; 27 (11): 1774–83. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 11. Reisner SL, Conron KJ, Baker K, et al.; GenIUSS Group. “Counting” transgender and gender-nonconforming adults in health research: recommendations from the Gender Identity in US Surveillance Group. Transgend Stud Q 2015; 2 (1): 34–57. [Google Scholar]
- 12. Tate CC, Ledbetter JN, Youssef CP. A two-question method for assessing gender categories in the social and medical sciences. J Sex Res 2013; 50 (8): 767–76. [DOI] [PubMed] [Google Scholar]
- 13. Rubin JD, Atwood S, Olson KR. Studying gender diversity. Trends Cogn Sci 2020; 24 (3): 163–5. [DOI] [PubMed] [Google Scholar]
- 14. Noel JK, Lutz TM. Measuring sex, gender, and sexual orientation in national disease surveillance systems: a pilot study. J Sex Res 2020; 57 (8): 987–96. [DOI] [PubMed] [Google Scholar]
- 15.Park A. Reachable: data collection methods for sexual orientation and gender identity. https://williamsinstitute.law.ucla.edu/publications/data-collection-sogi/. Accessed August 23, 2021.
- 16. Lagos D, Compton DL. Evaluating the use of a two-step gender identity measure in the 2018 General Social Survey. Demographics 2021; 58 (2): 763–72. [DOI] [PMC free article] [PubMed]
- 17. Harrison-Quintana J, Grant J, Herman JL. A gender not listed here: Genderqueers, gender rebels, and otherwise in the National Transgender Discrimination Survey. LGBTQ Publ Policy J Harvard Kennedy School 2012; 2 (1): 13. [Google Scholar]
- 18.Gender Identity in US Surveillance (GenIUSS) Group. Gender-Related Measures Overview. Los Angeles, CA: The Williams Institute at the University of California Los Angeles School of Law; 2013. [Google Scholar]
- 19. Deutsch MB, Green J, Keatley J, et al. Electronic medical records and the transgender patient: recommendations from the World Professional Association for Transgender Health EMR Working Group. J Am Med Inform Assoc 2013; 20 (4): 700–3. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 20. Gower AL, Rider GN, Brown C, et al. Supporting transgender and gender diverse youth: Protection against emotional distress and substance use. Am J Prev Med 2018; 55 (6): 787–94. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 21. Jadwin-Cakmak L, Reisner SL, Hughto JMW, et al. HIV prevention and HIV care among transgender and gender diverse youth: design and implementation of a multisite mixed-methods study protocol in the US. BMC Public Health 2019; 19 (1): 1531. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 22. Witcomb GL, Claes L, Bouman WP, Nixon E, Motmans J, Arcelus J. Experiences and psychological wellbeing outcomes associated with bullying in treatment-seeking transgender and gender-diverse youth. LGBT Health 2019; 6 (5): 216–26. [DOI] [PubMed] [Google Scholar]
- 23. Westbrook L, Saperstein A. New categories are not enough: rethinking the measurement of sex and gender in social surveys. Gender Soc 2015; 29 (4): 534–60. [Google Scholar]
- 24. Baker KE, Streed CG, Durso LE. Ensuring that LGBTQI+ people count - collecting data on sexual orientation, gender identity, and intersex status. N Engl J Med 2021; 384 (13): 1184–6. [DOI] [PubMed] [Google Scholar]
- 25. Clements-Nolle K, Marx R, Guzman R, Katz M. HIV prevalence, risk behaviors, health care use, and mental health status of transgender persons: implications for public health intervention. Am J Public Health 2001; 91 (6): 915–21. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 26. Kenagy GP. Transgender health: findings from two needs assessment studies in Philadelphia. Health Soc Work 2005; 30 (1): 19–26. [DOI] [PubMed] [Google Scholar]
- 27. Burgwal A, Gvianishvili N, Hård V, et al. Health disparities between binary and non binary trans people: a community-driven survey. Int J Transgend 2019; 20 (2–3): 218–29. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 28. Todd K, Peitzmeier SM, Kattari SK, Miller-Perusse M, Sharma A, Stephenson R. Demographic and behavioral profiles of nonbinary and binary transgender youth. Transgend Health 2019; 4 (1): 254–61. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 29. Poquiz JL, Coyne CA, Garofalo R, Chen D. Comparison of gender minority stress and resilience among transmasculine, transfeminine, and nonbinary adolescents and young adults. J Adolesc Health 2021; 68 (3): 615–8. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 30. Brener ND, Collins JL, Kann L, Warren CW, Williams BI. Reliability of the youth risk behavior survey questionnaire. Am J Epidemiol 1995; 141 (6): 575–80. [DOI] [PubMed] [Google Scholar]
- 31. Cramer RJ, Kaniuka AR, Yada FN, et al. An analysis of suicidal thoughts and behaviors among transgender and gender diverse adults. Soc Psychiatry Psychiatr Epidemiol 2021. Jun 9 [E-pub ahead of print]. [DOI] [PubMed] [Google Scholar]
- 32. Steinhausen H-C, Winkler Metzke CW. The impact of suicidal ideation in preadolescence, adolescence, and young adulthood on psychosocial functioning and psychopathology in young adulthood. Acta Psychiatr Scand 2004; 110 (6): 438–45. [DOI] [PubMed] [Google Scholar]
- 33. Sequeira GM, Kidd K, Coulter RW, et al. Transgender youths' perspectives on telehealth for delivery of gender-affirming care. J Adolesc Health 2020; 66 (2): S31–2. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 34. Corliss HL, Austin SB, Roberts AL, Molnar BE. Sexual risk in “mostly heterosexual” young women: Influence of social support and caregiver mental health. J Women's Health 2009; 18 (12): 2005–10. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 35. Adams N, Pearce R, Veale J, et al. Guidance and ethical considerations for undertaking transgender health research and institutional review boards adjudicating this research. Transgend Health 2017; 2 (1): 165–75. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 36. Bouman WP. Transgender and gender diverse people’s involvement in transgender health research. Int J Transgend 2018; 19 (4): 357–8. [Google Scholar]
- 37. Bauer GR, Churchill SM, Mahendran M, Walwyn C, Lizotte D, Villa-Rueda AA. Intersectionality in quantitative research: a systematic review of its emergence and applications of theory and methods. SSM Popul Health 2021; 14: 100798. [DOI] [PMC free article] [PubMed] [Google Scholar]
Associated Data
This section collects any data citations, data availability statements, or supplementary materials included in this article.
Supplementary Materials
Data Availability Statement
The data underlying this article cannot be shared publicly due to concern for the privacy of individuals that participated in the study. The data will be shared on reasonable request to the corresponding author.