The transgender-specific intimate partner violence scale for research and practice: Validation in a sample of transgender women

Abstract

Study/research objective.

To develop and validate a brief intimate partner violence (IPV) scale that screens for controlling behaviors and psychological abuse tactics directed toward transgender individuals.

Rationale.

Transgender individuals are at elevated risk of physical and sexual IPV compared to cisgender individuals. IPV often takes on unique dimensions against transgender individuals, such as when an abusive partner threatens to “out” the transgender person, or use other tactics that weaponize transphobia within the relationship. Standard IPV screeners do not assess this type of transgender-specific IPV (T-IPV).

Methods.

Between March 2018 and October 2019, a T-IPV scale was tested in two samples (in-person and online) of transfeminine adults (i.e. assigned a male sex at birth and identify with femininity) from the eastern and southern U.S. Exploratory factor analysis (EFA) was conducted with the in-person sample (N=661) to assess construct validity. Confirmatory factor analysis (CFA) was then used in an independent online sample (N=481). Using the combined sample (N=1,137), convergent validity was assessed using correlations with other forms of victimization. Multivariable regression models were fit to estimate the relationship between T-IPV and health outcomes.

Results.

Factor analyses yielded an 8-item unidimensional scale with moderate to good fit. Nearly half the sample (48.7%) experienced at least one scale item. Internal consistency reliability was strong (KR-20=0.827). Significant correlations with other forms of victimization indicated convergent validity. Lifetime T-IPV was significantly associated with psychological distress (adjusted prevalence ratio [aPR]=1.32, 95% CI=1.13, 1.53), PTSD (aPR=1.50, 95%CI=1.31, 1.72), alcohol abuse (aPR=1.21, 95%CI=1.01, 1.44), and drug use disorder (aPR=1.30, 95%CI=1.06, 2.59).

Conclusions.

This T-IPV scale is a reliable and unidimensional measure with strong construct validity. T-IPV is independently associated with mental health burden and substance use. Service providers working with transgender clients should screen for T-IPV to avoid missing cases of IPV, and refer to violence response services.

Introduction

Intimate partner violence (IPV) involves physical, sexual, or psychological harm perpetrated against an individual by a current or former partner or spouse (Black et al., 2011). A recent systematic review found that transgender individuals – individuals who identify as a gender different than the sex they were assigned at birth – are at 2.2 times the risk of physical IPV, and 2.5 times the risk of sexual IPV, as compared to cisgender (i.e., non-transgender) individuals (Peitzmeier et al., 2020b). Transgender individuals’ heightened vulnerability to IPV is likely because abusive partners are able to leverage societal transphobia as a tool of power and control within the relationship. For instance, qualitative evidence with transgender individuals finds that abusers may control their partner’s actions or blackmail them into unwanted sex by threatening to “out” them as transgender to their family, their coworkers, or simply to passers-by when out in public (Guadalupe-Diaz, 2013). Knowing that this type of victimization can lead to loss of social support or employment, as well as threats to physical safety (Hughto et al., 2015), transgender individuals may be forced to comply with an abuser’s wishes. Abusers may also undermine the partner’s sense of self-worth and confidence by critiquing their every action as not the way a “real” man or woman would behave or act. These trans-specific tactics have been linked to increased mental health burden for survivors (Peitzmeier et al., 2019). In other work with lesbian, gay, bisexual, transgender, or queer (LGBTQ) individuals more broadly, LGBTQ individuals who experienced IPV targeting their sexual and/or gender identity are also more likely to seek housing, mental health services, medical care, and support services than those who experience psychological abuse not specific to their identity, supporting the idea that identity-related IPV can be particularly damaging and require a higher level of support (Scheer and Baams, 2019).

Transgender women, especially transgender women of color, may face particular risks for IPV because of harmful societal stereotypes that specifically position trans women as hypersexual and stigmatize cisgender men who have relationships with transgender women (Gamarel et al., 2020). There is evidence to suggest that coercive control, violence, and even homicide perpetrated by cis male partners of transgender women is often to conceal the stigmatized relationship from other people. Further, transgender women feel pressure to engage in relationships with some men that they would not otherwise choose to partner with, for fear that these men may turn violent if rejected, or alternatively out of an internalized belief that they will not find other partners (Gamarel et al., 2020). Higher levels of economic precarity, discrimination in employment, and family rejection among trans women as compared to cisgender women (James et al., 2016) also increase dependency on abusive partners for economic or social support.

Common scales and screeners for IPV such as the Revised Conflict Tactics Scale (CTS-2) were originally developed and validated in samples of predominantly heterosexual cisgender women (Straus et al., 1996). These scales are considered behavioral measures in that they ask about specific abusive behaviors perpetrated against the individual such as being hit, punched, or slapped. Behavioral measures are advantageous because they do not require an individual to label their experience as abuse in order to report experiencing IPV. Because many survivors minimize abuse and do not label their experiences as abusive per se, behavioral measures are widely used in IPV research and service provision (Basile et al., 2007).

Yet, existing behavioral IPV measures developed with cisgender women are problematic when applied to transgender populations. Research finds that many abusers preferentially use trans-specific abuse tactics, as their power is reinforced by the weight of societal transphobia (Guadalupe-Diaz, 2013). Existing IPV measures fail to screen for these most commonly used abuse tactics that are unique to the experiences of transgender people; thus, they are likely insufficiently sensitive as IPV screening tools for transgender populations. Low sensitivity in practice or clinical settings can result in transgender survivors being incorrectly “screened out,” deprioritized for services, denied services, or never offered services for IPV in the first place.

Researchers have made efforts to address this gap by developing behavioral scales that include transgender-specific tactics. The US Trans Survey (USTS; N=27,715 transgender individuals) found that 27% of participants had experienced some kind of transgender-specific IPV in their lifetime as measured by three questions: their partner had prevented them from accessing hormones (3%), told them that they were not a “real” woman or man (25%), or threatened to “out” them as transgender (11%) (James et al., 2016). Garthe and colleagues included two trans-specific items in a larger IPV scale administered to young transgender women, which focused on lifetime coercive control of gender presentation (18%) and belittling a partner due to their transgender identity (22%) (Garthe et al., 2018). Peitzmeier and colleagues (2019) developed a 4-item transgender-specific IPV (T-IPV) scale and piloted it in a sample of 150 transmasculine individuals (i.e., individuals assigned a female sex at birth who identify their gender on a spectrum of masculinity). Their measure had good convergent and divergent validity, but only moderate reliability (as well as incomplete domain coverage) given the small number of items (Peitzmeier et al., 2019). Woulfe and Goodman (2018) developed a 7-item scale of “identity abuse” meant to be used by LGBTQ individuals broadly, and tested it in a sample of 734 LGBTQ individuals, including 142 transgender people. Woulfe and Goodman found that 40.1% reported any identity abuse in adulthood, with individual item frequencies ranging from 6.3% (“the person threatened to tell my employer, family, or others about my sexual orientation or gender identity”) to 28.3% (“the person questioned whether my sexual orientation or gender identity was ‘real’”) (Woulfe and Goodman, 2018). Further, transgender participants in their sample were more likely than cisgender participants to experience identity abuse, consistent with the hypothesis that transgender individuals are particularly vulnerable to identity-specific IPV. Similarly, Dyar and colleagues developed a 5-item IPV scale for LGBTQ individuals that focused on outing, social isolation, and pressuring someone to stay in a relationship by telling them no one else would ever love them due to their LGBTQ identity (Dyar et al., 2019). However, the items in both of these scales were not truly transgender-specific as they were developed to be relevant across the range of identities and experiences in the LGBTQ community. As such, these LGBTQ-focused measures lack domain coverage and content validity for transgender individuals by not being able to include common abusive behaviors directed against transgender people specifically, such as partners sabotaging gender transition or belittling gender expression.

The current study builds on this existing work by developing and validating a transgender-specific IPV scale that screens for abuse tactics directed against transgender people and piloting a potential 10-item scale with transfeminine adults (i.e., individuals assigned a male sex at birth who identify their gender along the feminine gender spectrum, including but not limited to transgender women). We chose to continue focusing on transgender-specific IPV, rather than LGBTQ-related IPV broadly as in Woulfe and Goodman’s or Dyar et al’s work, because this allows for the inclusion of items that are not relevant to cisgender LGBQ individuals, such as hiding and/or sabotaging hormones by a partner. Further, qualitative evidence suggests that abusers may use abuse tactics related to gender identity even more commonly than tactics related to sexual orientation, because of the centrality of gender to identity and because transgender identities are even more stigmatized than sexual minority identities (Guadalupe-Diaz, 2013). As stated by Woulfe and Goodman, “At its heart, [identity abuse] involves the exploitation of a target’s vulnerabilities by attacking their most oppressed identities. It makes sense then, that the greater the stigma attached to an identity, the greater the risk of [identity abuse]” (Woulfe and Goodman, 2018). Naming these transgender-specific tactics is therefore critical for IPV screening in transgender populations.

Methods

Study overview and data collection

Participants were enrolled as part of the American Cohort To Study HIV Acquisition Among Transgender Women (known to participants and the community as the LITE study) (Wirtz et al., 2019). LITE is a multi-site prospective cohort examining HIV acquisition and other health outcomes among transfeminine adults in the Southern and Eastern US across 24-months. Participants contributing data to this analysis completed a baseline assessment between March 22, 2018 and October 22, 2019 (n=1137). Adult (ages 18 years and older) transfeminine individuals of any HIV status were recruited via convenience sampling methods, including but not limited to: peer referrals, social media, dating apps, and clinic referrals. Transgender identity was assessed in screening using a two-step measure (Reisner et al., 2014). All participants were provided $50 to compensate for time completing baseline study procedures.

In-Person Sample:

Participants were enrolled in a facility-based study visit at baseline and completed a socio-behavioral survey (self-administered or interviewer-administered according to participant preference and results of literacy screener) and laboratory-confirmed HIV and STI testing (Wirtz et al., 2019). Participants were recruited in six eastern and southern U.S. cities (Atlanta, Baltimore, Boston, Miami, New York City, and Washington, DC).

Online Sample:

An online cohort of transfeminine adults was recruited in January 2019. The cohort was recruited primarily through ads on Reddit, Facebook, and Google AdWords from over 50 medium-to-large cities throughout the southern and eastern US. Online cohort participants completed the survey via phone app or secure website. Participants had to provide email and cell phone number and use two-factor authentication to aid in identification of duplicate participation or fraudulent (e.g. bot/spam) participants.

This study was approved by the Johns Hopkins Medicine Institutional Review Board. All partner sites ceded review to the single IRB of record.

Measures

T-IPV.

Preliminary work to develop a T-IPV scale was completed by members of the study team in 2015–2016 (Peitzmeier et al., 2019). Four items were developed based on a review of the literature, including reports from LGBTQ service organizations (Quinn, 2011, Munson and Cook-Daniels, 2003) and scholarly work (Guadalupe-Diaz, 2013, Greenberg, 2012), as well as review by transgender individuals who were members of the study team. We tested these four items in a study of 150 transmasculine individuals in Boston, MA. Basic psychometric validation found moderate reliability (KR-20=0.56) and good convergent and divergent validity, as well as associations with negative mental health outcomes including symptoms of posttraumatic stress disorder (PTSD), depression, and psychological distress. Participants were invited to provide feedback on the measure.

Based on these preliminary results, we set out to refine the wording of existing items, expand the scale to capture additional domains of T-IPV reported in the literature, and test it in a larger sample of transfeminine spectrum adults, permitting more sophisticated psychometric testing and evaluation. This process allowed us to address 1) incomplete domain coverage in the original 4-item scale, and 2) inadequate sampling, i.e. the mismatch between sample in which original scale was validated (transmasculine individuals) and the desired target population (both transmasculine and transfeminine individuals). To address incomplete domain coverage and expand the scale, we drew on suggestions from participants, reviewed literature that had been published since the original study, and solicited suggestions from other researchers. Based on these inputs, we expanded the measure to capture additional abusive behaviors that were previously not included in the original scale. We developed a set of 10 items, including refined versions of the original 4 items, after conducting collaborative discussions among members of the study team (items found in Table 1). T-IPV was assessed using both past 12 month and lifetime reference periods.

Table 1.

Prevalence of each transgender-specific (T-IPV) item and overall T-IPV, past-year and lifetime among transfeminine adults in the Eastern and Southern US. Factor loadings derived from exploratory factor analysis (N=661 with complete data).

		Past-year % (n/N)	Lifetime % (n/N)	Item-Total Correlation	Factor Loading (EFA)	Factor loading (CFA)
	Retained items
Item 1	Did a partner hide, damage, or destroy your hormones, clothing, or other items (makeup, binders, prosthetics) related to gender transition?	4.4 (30/680)	10.6 (72/680)	0.481	0.747	0.818
Item 2	Did a partner force or pressure you into not pursuing aspects of gender transition that you wanted, such as name change, hormones, or surgery?	8.0 (54/674)	20.7 (140/675)	0.513	0.683	0.823
Item 3	Did a partner force or pressure you into wearing certain clothes, hairstyles, makeup, or other aspects of gender presentation that did not match your gender identity?	6.1 (41/676)	13.5 (91/676)	0.473	0.662	0.801
Item 4	Did a partner “out” you or threaten to “out” you, i.e. tell someone else you are transgender against your wishes, in order to shame you or make you feel unsafe?	6.1 (41/677)	12.5 (85/679)	0.516	0.781	0.813
Item 5	Did a partner ever tell you that no one else would want to be with someone like you because you are transgender?	10.3 (69/673)	22.1 (149/675)	0.601	0.854	0.922
Item 6	Did a partner ever tell you that you are not a “real” man or “real” woman?	16.9 (114/674)	31.7 (214/676)	0.639	0.902	0.938
Item 7	Did a partner ever intentionally call you by the wrong name or pronoun to hurt you?	12.0 (81/677)	22.2 (151/679)	0.578	0.833	0.880
Item 8	Did a partner tell you to not “act” trans, or to act more feminine or more masculine?	11.3 (77/679)	20.9 (142/680)	0.607	0.810	0.863
	Any T-IPV	25.0 (164/679)	48.7 (322/661)
	Kuder-Richardson Coefficient (K-20)	--	0.827
	Proportion of Variance Explained	--	0.620
	Dropped Items
Item 9	Did a partner force or pressure you into pursuing aspects of gender transition that you did not want, such as name change, hormones, or surgery?a	2.8 (19/679)	4.9 (33/679)	--	--	--
Item 10	Did a partner pressure you to have a certain kind of sex with them or with others by telling you that a “real” man or woman would do it?b	6.2 (42/676)	12.0 (81/677)	--	--	--

^aDropped based on low item-total correlation (0.248) and low factor loadings compared to other items (0.544) in EFA

^bDropped based on low item-total correlation (0.446) and relatively low factor loadings compared to other items (0..767), as well as conceptual concerns that this item taps into other constructs beyond T-IPV, such as sexual IPV.

The 10-item question set was introduced with the following instructions: “Relationships can have good and bad moments. Some questions may be difficult to answer. Your answers will be kept confidential. In your lifetime, did any of your partners do any of the following things to you? By partner, we mean a person you were dating, in a sexual relationship with, or married to.” Answer choices for each question included “No” or “Yes.” For participants who said “yes” to a question, a follow up question was asked, “Has this happened in the last 12 months?” Respondents were considered to have experienced T-IPV if they said “Yes” to any of the items. For convergent validity analyses only, T-IPV was treated as a score ranging from 0 to 8 with one point for each item reported.

Demographic variables and other potential confounders.

Age was assessed with the question, “What is your age, as of today?” and allowed a write-in response. Income was assessed with the measure, “In the last 30 days , what was your household income before taxes? Household income refers to the total amount of money from all people living in the household,” with answer options “$0–$499,” “$500–$999,” “$1,000–$1,999,” $2,000–$3.999,” and “$4000 or more.” Reported income was reclassified in the analysis to its relation to the federal poverty line. Education was assessed with the measure, “Which category best describes your educational background?” and 10 answer options ranged from “Did not complete 8^th grade” to “Completed graduate school.” These were dichotomized into “high school diploma or less” and “any post-secondary education.” Race/ethnicity was assessed with “Which of the following describes your race? Please select all that apply” with answer choices White, Black or African American, Asian, American Indian/Alaskan Native, Hawaiian Native/Pacific Islander, Latina/Hispanic, Other.

Gender identity was assessed with the question, “How do you self-identify in terms of gender? Please select the response that best describes your gender,” with the options “Female or woman,” “Trans female, trans woman, male-to-female (MTF)”, “Transfeminine, trans femme,” “Non-binary, agender, gender fluid, gender variant, gender non-conforming,” “Woman of trans experience,” “Person of trans experience,” “Two-spirit,” or “Other identity.” Hormone use was assessed with the question, “In the last 3 months, have you taken hormones for your gender identity or gender transition?”

Sex work was assessed with the question, “In your lifetime, have you ever had sex with someone so that they would give you money, drugs, alcohol, food, a place to sleep or other material goods? By sex, we mean oral, anal and/or vaginal sex.” Food insecurity was assessed with the question, “How often do you run out of food or money to purchase food at the end of the month?” Family support was measured using a 5-point Likert scale response of agreement with the statement: “My family is accepting and supportive of my gender identity.”

Gender of regular partners was assessed with “Of your regular partners in the last 3 months, what gender(s) were your regular partners? (check all that apply).” Relationship status was described with the question “What would be the best way to describe your current relationship status?”

Although there is limited literature on T-IPV, potential confounders were selected based on a review of the literature on known correlates of other types of IPV. Younger individuals are more likely to experience recent IPV, though older individuals are more likely to report lifetime IPV by virtue of having a longer time-at-risk (Breiding et al., 2014). IPV varies by race/ethnicity, with Black, Latinx, American Indian, and multiracial individuals at highest risk (Breiding et al., 2014). It is well-established that individuals who engage in sex work experience higher prevalence of many types of violence, including IPV, due to stigma against sex workers (Rouhani et al., 2020, James et al., 2016). Financial precarity, measured with income and food insecurity, induces vulnerability for IPV (Breiding et al., 2014) by making individuals more dependent on partners, even abusive ones, to meet basic survival needs. Social isolation, including lack of family support, also reduces the ability to leave an abusive partnership. All of these variables are also known to be associated with mental health and substance use outcomes, often because structural vulnerabilities can cause negative mental health outcomes and coping strategies. Gender of intimate partners was included as cisgender men are more likely than cisgender women to perpetrate certain kinds of IPV, including sexual IPV and severe physical IPV (Breiding et al., 2014). Very little is known about rates of perpetration among transgender individuals (Peitzmeier et al., 2020b). While less is known about how gender identity or stage of medical transition (which we measured with hormone use) within trans communities is correlated with IPV (Peitzmeier et al., 2020b), we hypothesized that gender and gender presentation may be associated with IPV vulnerability.

Violence victimization variables for convergent validity.

Physical IPV was measured with a series of 6 questions that were adapted from the WHO multi-country study on women’s health and domestic violence (Garcia-Moreno et al., 2006). The question set began, “Has anyone ever done the following to you? We are not referring to rough play/BDSM [bondage/discipline/sadism/masochism] that you and your partner have agreed to in advance,” followed by 6 binary questions: 1) slapped you or thrown something at you that could hurt; 2) pushed or shoved you; 3) hit you with a fist or something else that could hurt you; 4) kicked you, dragged you, or beaten you up; 5) choked or burned you on purpose; and 6) threatened to use or actually used a knife, gun, or some other weapon against you. This was followed by the question “Who did this to you? Please say yes or no to the following” with the first two options being “current partner” and “ex-partner.” We considered someone to have experienced physical IPV if they said yes to any of the 6 items, and said yes to one of those two partner types as perpetrators. Kuder-Richardson Formula 20 (KR-20) in the sample was 0.93.

Sexual IPV was measured with a series of four questions. The question set began with the same introductory wording, followed by 4 binary items: 1) Physically forced you to have sexual intercourse or do something sexual when you did not want to; 2) Forced you to do something sexual that you found degrading or humiliating; 3) Had sex or did something sexual you did not want to because you were afraid of what they might do; 4) Had sexual intercourse or did something sexual when you did not want to because they told you it was their right. This was followed by the same perpetrator options. KR-20 in the sample was 0.88.

Non-partner physical or sexual violence was measured by using the same violence items above, but selecting a perpetrator option that included: sex work client (date), family member, employer/coworker, someone in the transgender community, someone in your neighborhood, police/law enforcement, stranger, or other. KR-20 in the sample was 0.87.

We created two variables for physical IPV, sexual IPV, and non-partner violence: a total scale score calculated by summing the number of items endorsed (for convergent validity correlation analyses) and a dichotomous exposure score indicating whether participants reported experiencing any item (for use in bivariate associations and multivariable models).

Adverse childhood experiences (ACEs) were assessed with eight questions asking about household circumstances prior to age 18 including living with someone 1) with mental illness, 2) who was an alcoholic, 3) who used drugs, or 4) who went to jail, in addition to 5) having divorced parents or a single-parent household (yes/no). Other ACEs included: 6) observing domestic violence; 7) physical abuse by parent or other adult in the home; 8) being insulted or sworn at by parent or other adult in the home.(Bynum et al., 2010) Answer choices were never (0), once (1), or more than once (2). Sexual abuse items were excluded due to mandatory reporting requirements at some sites. Responses were summed across the 8 items to calculate an ACEs score (range: 0–11).

Religiosity variable for divergent validity.

Religiosity was measured with the question “How important is religion to you?” with a four-point scale and treated as a binary variable of “not at all important” versus “a little important,” “important,” or “very important.”

Mental health and substance use outcome variables.

Serious distress was measured using the Kessler-6, with a cutoff of greater than or equal to 13 (Kessler et al., 2002). Cronbach’s alpha was 0.88. PTSD was measured using the four-item PC-PTSD, with a cutoff greater than or equal to 3 indicating PTSD symptomatology (Prins et al., 2003). KR-20 was 0.86. Alcohol use disorder (AUD) was measured with AUDIT-C with a score greater than or equal to 4 (Bush et al., 1998). Cronbach’s alpha was 0.73. Illicit drug use disorder (DUD) was assessed using a modified DAST-10 with score greater than or equal to 3 (moderate level of problems related to drug use with further investigation suggested) (Yudko et al., 2007). KR-20 was 0.86.

Statistical Analysis

Internal consistency reliability.

Internal consistency reliability was assessed using KR-20 (Kuder and Richardson, 1937). KR-20 above 0.80 was considered strong reliability.

Exploratory factor analysis (EFA).

EFA was performed on the sample of participants recruited in-person (n=661). Item-total correlations were calculated for each lifetime exposure to each item using the ‘psych’ package from R and an initial EFA was run after calculating the Kaiser-Meyer-Olkin measure of sampling adequacy and Bartlett’s sphericity test. Items with low item-total correlation and low factor loadings in EFA one-factor model were dropped from the scale. Participants with a missing response to any of the remaining items were then dropped (n=29) and final item-total correlations were calculated and a final exploratory factor analysis (EFA) was performed using the factanal function in R and a tetrachoric covariance matrix. The EFA included a one-factor model along with a two- and three-factor solution with oblique and orthogonal rotations. Number of factors was assessed using scree plots and reviewing eigenvalues. The lifetime prevalence of each transgender-specific (T-IPV) item was calculated with a “yes” or “no” response to each specific item; prevalence of any lifetime T-IPV was calculated for those who said “yes” to any item.

Confirmatory factor analysis (CFA).

Based on the scree plot from the EFA and examination of factor loadings in one-, two-, and three-factor models, a one-factor model using the eight remaining items was tested using CFA using the lavaan package in R, a tetrachoric covariance matrix, and diagonally weighted least squares estimator. The CFA was fit to an independently obtained sample of online participants (n=481). Best practices in scale validation include using a separate, independent sample for the EFA and CFA (Kline, 2015). Standardized factor loadings and robust measures of model fit (Standardized Root Mean Square Residual (SRMR), comparative fit index (CFI), Tucker-Lewis Index (TLI), model Chi-square, and Root Mean Square Error of Approximation (RMSEA)) are reported.

Convergent and divergent validity.

The in-person and online cohort participants were combined for use in subsequent analyses (n=1,137) after confirming T-IPV had a similar factor structure in both through EFA and CFA. For convergent validity analyses only, T-IPV was treated as a score ranging from 0 to 8 with one point for each item reported. Using a score instead of a dichotomized variable to test convergent validity is consistent with validation of other violence scales (Straus et al., 1996, Woulfe and Goodman, 2018, Scheer et al., 2019). Convergent validity was examined by measuring the correlation between lifetime T-IPV and other variables using Pearson’s r for continuous variables (physical IPV, sexual IPV, non-partner violence, ACEs). For divergent validity, polychoric correlation was used to assess the correlation between lifetime T-IPV (ever/never) with dichotomous variables (religiosity). These variables were selected a priori based on conceptual understanding of the IPV literature and what variables likely should and should not be correlated with T-IPV.

Associations with demographics and health outcomes.

Bivariate analyses (t-tests and chi-square tests) examined the association between lifetime T-IPV and demographic variables, other lifetime IPV, mental health variables, and substance use disorder.

The associations between lifetime T-IPV (ever vs. never) and four health outcomes were assessed using Poisson regression with robust standard errors. We focused on mental health and substance use outcomes given that these are well-established IPV-related health outcomes, particularly for psychological IPV (Carbone-López et al., 2006, Lagdon et al., 2014). All models were adjusted for age, race/ethnicity, educational attainment, food insecurity, gender identity, hormone use, and lifetime sex work. The covariates were chosen from the list of demographic variables examined in bivariate analyses. Income was dropped due to high percentage of missing data (14.8%) and a few additional variables were dropped to help with model convergence (e.g., site, family support); all dropped variables were not significant at p<0.10 in the bivariate models. Final regression models were assessed for multicollinearity by examining variance inflation factor (VIF) between variables. All VIFs aside from the VIFs for the categorical variable of age were between 1 and 1.5 in all models; VIFs for age were between 2 and 4 (no concern for multicollinearity). We did not include physical and sexual IPV as covariates in the main multivariable models. T-IPV often serves to make individuals more vulnerable to physical and sexual IPV, so including physical and sexual IPV in the models may be adjusting for a mediator rather than a confounder, underestimating the total effect of T-IPV on health outcomes. However, a sensitivity analysis was conducted where we added physical IPV and sexual IPV to the models as two additional covariates.

All regression analyses were conducted with SAS 9.4 (SAS Institute Inc., Cary, NC) and all other analyses were conducted with R version 3.6.0 (https://www.r-project.org). The effective sample size for each part of the analysis was allowed to vary depending on how many participants had missing data for particular variables (i.e. pairwise deletion). Two percent (2.3%) of participants had data missing from modelled covariates and were dropped from the models; missingness in outcome variables ranged from 0.88% (psychological distress) to 3.34% (alcohol use disorder). We conducted a sensitivity analysis with multiply imputed data and found results of the models to be essentially unchanged (see Web Appendix).

Results

Lifetime prevalence of each of the 10 T-IPV items ranged from 4.9% to 31.7%. (Table 1).

Exploratory factor analysis

The Kaiser-Meyer-Olkin (KMO) test for adequacy (0.87) and Bartlett’s test of sphericity (p<0.001) indicated sample adequacy for EFA. Two items in the initial EFA had low item-total correlation and/or relatively low factor loadings, and were dropped from the scale (Table 1). The remaining 8 items had a KR-20 of 0.827, indicating good reliability, and had factor loadings ranging from 0.494 to 0.734 (moderate to strong). Proportion of the variance explained was 0.620. The scree plot indicated a one-factor solution was best, based on the knee in the curve and the eigenvalues for factors after the first (Figure 1). The prevalence of any experience of T-IPV, indicated by a reported experience of any of the 8 items, was 25.0% in the last 12 months while lifetime prevalence was 48.7%.

Fig. 1.

Scree plot from exploratory factor analysis.

Note. Scale items are standardized to have a standard deviation of 1 and the latent variable is constrained to have a standard deviation of 1; standardized factor loadings were able to range from −1 to 1. FA = factor analysis.

Confirmatory factor analysis

The fit of a one-factor model using CFA in a separate sample of online participants was moderate to good. SRMR was 0.049, CFI was 0.994, the TLI was 0.992, the model chi-square was p=0.001 (χ²=45.522, df=20), and the RMSEA was 0.052 (90% CI = 0.032, 0.072). Standardized factor loadings were between 0.801 and 0.938 (Table 1).

Sensitivity analyses using past-year instead of lifetime T-IPV found the same one-factor solution and similar factor loadings and reliability. Fit indices in CFA were slightly lower but similar (data not shown).

Sample demographics and characteristics of the combined sample included in validity analyses

The median age group was 25–29 (Table 3). The sample was half (52.9%) white, 18.3% Latinx, and 13.7% Black. About one-quarter (27.5%) had a high school diploma or less education, one-third (36.5%) had incomes below the federal poverty line, and one-third (36.7%) had ever engaged in sex work. The majority (79.9%) had a binary gender identity such as woman or trans woman and recently used hormones (79.5%). Gender of sexual partners was diverse, with many reporting a recent cis male (48.6%), cis female (31.1%), or trans female (16.8%) partner. Roughly one-quarter reported a lifetime history of physical IPV (25.2%) or sexual IPV (23.2%), and most reported non-partner physical or sexual violence (60.1%). Mental health and substance abuse burden was significant with many screening positive for psychological distress (38.3%), PTSD (46.6%), alcohol use disorder (32.8%), and drug use disorder (27.8%).

Table 3.

Characteristics of transfeminine adults in the Eastern and Southern United States and prevalence of any lifetime T-IPV by demographics, other lifetime IPV, relationship characteristics, mental health, and substance use.

	Prevalence of characteristic in the total Sample (N=1137)	Prevalence of Any Lifetime T-IPV among group	p-value
	n (%)	n (%)
Demographics
Site
Boston, MA	140 (12.3)	64 (45.7)	0.458
New York, NY	196 (17.2)	98 (50.0)
Baltimore, MD	70 (6.2)	36 (51.4)
Washington, DC	111 (9.8)	53 (47.7)
Miami, FL	83 (7.3)	50 (60.2)
Atlanta, GA	59 (5.2)	26 (44.1)
Online	478 (42.0)	242 (50.6)
Age Group in Years
18–24	361 (31.8)	134 (37.1)	<0.001
25–29	281 (24.8)	139 (49.5)
30–39	289 (25.5)	162 (56.1)
40–49	109 (9.6)	75 (68.8)
50+	94 (8.3)	57 (60.6)
Race / Ethnicity
Asian or Hawaiian Native/Pacific Islander	27 (2.4)	8 (29.6)	0.211
Black	156 (13.7)	78 (50.0)
White	602 (52.9)	288 (47.8)
Latinx	208 (18.3)	108 (51.9)
Multiracial	116 (10.2)	72 (62.1)
Other or unknown	28 (2.5)	15 (53.6)
Educational Attainment
High school diploma or less	312 (27.5)	152 (48.7)	0.668
Any post-secondary education (including technical/ vocational)	822 (72.5)	414 (50.4)
Household Income
100% federal poverty level	354 (36.5)	177 (50.0)	0.087
200%	202 (20.8)	119 (58.9)
300% or more	413 (42.6)	208 (50.4)
Gender Identity
Female or Woman	318 (28.0)	151 (47.5)	0.228
Transwoman	590 (51.9)	292 (49.5)
Nonbinary	62 (5.5)	31 (50.0)
Another Gender Identity	166 (14.6)	95 (57.2)
Sex work (ever)
No	712 (63.3)	290 (40.7)	<0.001
Yes	413 (36.7)	273 (66.1)
Hormone use (past 3 months)
No	232 (20.5)	122 (52.6)	0.454
Yes	900 (79.5)	446 (49.6)
Food insecurity
Do not/seldom/rarely	679 (59.9)	298 (43.9)	<0.001
Sometimes/most of the time	454 (40.1)	269 (59.3)
Family support
Agree or strongly agree	516 (46.3)	242 (46.9)	0.063
Strongly disagree, disagree, or neither agree nor disagree	598 (53.7)	315 (52.7)
Relationship characteristics
Relationship status
Single, not in a relationship	476 (42.0)	234 (49.2)	0.004
Casually dating	182 (16.1)	102 (56.0)
In a committed relationship	305 (26.9)	130 (42.6)
Legally married or in a civil union	103 (9.1)	63 (61.2)
Other relationship type	66 (5.8)	37 (56.1)
Regular partner gender past 3 months
Cisgender man
No	302 (51.4)	150 (49.7)	0.310
Yes	286 (48.6)	155 (54.2)
Cisgender woman
No	405 (68.9)	199 (49.1)	0.059
Yes	183 (31.1)	106 (57.9)
Transgender woman
No	489 (83.2)	265 (54.2)	0.017
Yes	99 (16.8)	40 (40.4)
Transgender man
No	542 (92.5)	281 (51.8)	0.937
Yes	44 (7.5)	22 (50.0)
Genderqueer AFAB
No	523 (89.1)	274 (52.4)	0.483
Yes	64 (10.9)	30 (46.9)
Genderqueer AMAB
No	556 (94.7)	293 (52.7)	0.093
Yes	31 (5.3)	11 (35.5)
Place where most recent regular partner was met
Online dating apps	186 (32.2)	110 (59.1)	0.203
Public spaces	50 (8.7)	28 (56.0)
Work or school	111 (19.2)	51 (45.9)
Bar or club	28 (4.8)	15 (53.6)
Party	25 (4.3)	10 (40.0)
Hotel	7 (1.2)	5 (71.4)
Other	84 (14.5)	40 (47.6)
Other online	87 (15.1)	41 (47.1)
No regular partner last 3 months
No	589 (52.0)	305 (51.8)	0.259
Yes	543 (48.0)	262 (48.3)
Number of sexual partners (last 3 months)	4.37 (overall mean)	5.40 (mean among TIPV) vs 3.31 (mean without TIPV)	0.002
Other violence
Violence History (lifetime)
Physical IPV
No	840 (74.8)	346 (41.2)	<0.001
Yes	283 (25.2)	217 (76.7)
Sexual IPV
No	859 (76.8)	365 (42.5)	<0.001
Yes	260 (23.2)	195 (75.0)
Nonpartner physical or sexual violence
No	450 (39.9)	184 (40.9)	<0.001
Yes	678 (60.1)	383 (56.5)
Mental Health
Serious Psychological Distress
No	695 (61.7)	323 (46.5)	0.003
Yes	432 (38.3)	241 (55.8)
PTSD
No	598 (53.4)	239 (40.0)	<0.001
Yes	522 (46.6)	320 (61.3)
Substance Use
Alcohol Use Disorder (AUDIT-C)
No	753 (67.2)	362 (48.1)	0.041
Yes	367 (32.8)	201 (54.8)
Illicit drug use (DAST-10)
No	793 (72.2)	369 (46.5)	<0.001
Yes	306 (27.8)	182 (59.5)

IPV=intimate partner violence. AFAB=assigned female at birth. AMAB=assigned male at birth. PTSD=Post-traumatic stress disorder. AUDIT-C=Alcohol Use Disorders Identification Test – Consumption Questions. DAST-10=Drug Abuse Screening Test.

Convergent and divergent validity

The lifetime T-IPV scale demonstrated strong convergent validity in the combined sample. T-IPV was highly correlated (all p-values <0.001) with physical and sexual IPV, adverse childhood experiences, and non-partner physical or sexual violence (Table 2). Correlation coefficients were of moderate magnitude, indicating both convergent validity with and distinctness from these other types of violence (Table 2). T-IPV was still a distinct construct compared to physical and sexual IPV and occurred without physical or sexual IPV in 40.0% of all IPV cases (Figure 2). Divergent validity analyses found that, as expected, T-IPV was not associated with religiosity (r=−0.081, p>0.05).

Table 2.

Associations with lifetime T-IPV among transgender women as evidence of convergent and divergent validity.

		Correlation with lifetime T-IPV
1	Physical IPV	0.368***
2	Sexual IPV	0.410***
3	Adverse childhood experiences	0.239***
4	Non-partner physical or sexual violence	0.347***
5	Religiosity	−0.081

^*p < 0.05;

^**p < 0.01;

^***p < 0.001.

Fig. 2.

Venn diagram showing co-occurrence of lifetime physical, sexual, and transgender-specific IPV.

Note. Area of circles is proportional to the number of individuals represented in each group. Denominators for percentages include all participants who experienced any physical, sexual, or trans-specific IPV in their lifetime (not inclusive of participants who experienced no IPV), n=658.

Bivariate associations with lifetime T-IPV

Lifetime T-IPV was significantly associated with age, with T-IPV most prevalent in the 40–49 age group (68.8%) and least frequent in the 18–24 age group (37.1%) (Table 3). T-IPV was more prevalent among those who had ever engaged in sex work (66.1%) versus those who had not (40.7%), and among those who experience food insecurity sometimes or most of the time (59.3%) versus those who do not, seldom, or rarely experience food insecurity (43.9%). Lifetime T-IPV was more prevalent among those who reported currently being in a legal marriage or civil union (61.2%), casually dating (56.0%), or in another relationship type (56.1%) versus those who reported being single (49.2%) or in a committed relationship (42.6%). Among those who had a regular partner in the last 3 months, T-IPV was also less common among those who reported one of those partners was a transgender woman (40.4%) versus those who did not (54.2%).

Lifetime T-IPV was more prevalent among those who experienced physical IPV (76.7%) versus those who did not (41.2%), among those who experienced sexual IPV (75.0%) versus those who did not (42.5%), and among those who experienced non-partner physical or sexual violence (56.5%) versus those who did not (40.9%). T-IPV was also more prevalent among those with serious psychological distress (55.8%) versus those without (46.5%), those with PTSD (61.3%) versus those without (40.0%), those with AUD (54.8%) versus those without (48.1%), and those with DUD (59.5%) versus those without (46.5%). T-IPV was also associated with a greater number of sexual partners in the past 3 months (5.40 partners on average versus 4.37 partners on average in those with or without T-IPV).

Adjusted associations of lifetime T-IPV with health outcomes

After adjusting for potential demographic and other confounders, lifetime T-IPV was significantly associated with 32% increased risk of serious psychological distress (adjusted prevalence ratio (APR)=1.32, 95% CI=1.13, 1.53), 50% increased risk of PTSD (APR=1.50, 95%CI=1.31,1.72), 21% increased risk of AUD (APR=1.21, 95%CI =1.01, 1.44), and 30% increased risk of DUD (APR=1.30, 95%CI=1.06, 1.59) (Table 4).

Table 4.

Multivariable logistic regression models: adjusted associations of lifetime T-IPV exposure and mental health and substance use outcomes among transfeminine adults.

	Mental Health		Substance Use
	Serious Psychological Distress (K-6)	Posttraumatic Stress Disorder (PTSD)	Alcohol Use Disorder (AUDIT-C)	Drug Use Disorder/Drug Abuse (DAST-10)
	aPR (95% CI)	aPR (95% CI)	aPR (95% CI)	aPR (95% CI)
Exposure
Any Lifetime T-IPV	1.32 (1.13, 1.53)	1.50 (1.31, 1.72)	1.21 (1.01, 1.44)	1.30 (1.06, 1.59)
Covariates
Age in Years
18–24	2.31 (1.53, 3.48)	1.2 (0.95, 1.52)	1.22 (0.85, 1.76)	1.88 (1.18, 2.98)
25–29	1.93 (1.27, 2.92)	1.12 (0.87, 1.42)	1.38 (0.96, 1.98)	1.77 (1.11, 2.84)
30–39	1.67 (1.10, 2.54)	1.02 (0.8, 1.31)	1.03 (0.71, 1.50)	1.56 (0.97, 2.50)
40–49	1.23 (0.75, 2.03)	0.91 (0.67, 1.23)	1.20 (0.79, 1.82)	1.27 (0.73, 2.20)
50+	1.00 (ref)	1.00 (ref)	1.00 (ref)	1.00 (ref)
Race / Ethnicity
Asian or Hawaiian Native/Pacific Islander	1.11 (0.73, 1.69)	1.05 (0.70, 1.59)	1.09 (0.66, 1.80)	0.65 (0.34, 1.26)
Black	0.48 (0.36, 0.64)	0.78 (0.63, 0.96)	0.70 (0.51, 0.95)	0.65 (0.48, 0.87)
Latinx	0.60 (0.48, 0.76)	0.81 (0.67, 0.98)	0.90 (0.71, 1.13)	0.71 (0.53, 0.94)
Multi-racial	0.77 (0.61, 0.98)	1.19 (1.00, 1.41)	0.65 (0.46, 0.91)	0.64 (0.45, 0.90)
Other or unknown	0.65 (0.38, 1.12)	1.04 (0.72, 1.49)	0.69 (0.36, 1.32)	1.43 (0.89, 2.31)
White	1.00 (ref)	1.00 (ref)	1.00 (ref)	1.00 (ref)
Educational attainment
High school diploma or less	1.00 (ref)	1.00 (ref)	1.00 (ref)	1.00 (ref)
Any post-secondary education (including technical/ vocational)	0.89 (0.75, 1.06)	1.02 (0.88, 1.18)	1.22 (0.97, 1.52)	1.23 (0.97, 1.55)
Food insecurity
Do not/seldom/rarely	1.00 (ref)	1.00 (ref)	1.00 (ref)	1.00 (ref)
Sometimes/most of the time	1.63 (1.40, 1.90)	1.47 (1.29, 1.68)	0.98 (0.82, 1.18)	1.36 (1.12, 1.67)
Gender Identity
Female or Woman	1.00 (ref)	1.00 (ref)	1.00 (ref)	1.00 (ref)
Transwoman	0.98 (0.82, 1.16)	0.91 (0.79, 1.05)	0.95 (0.78, 1.17)	1.13 (0.89, 1.43)
Nonbinary	1.19 (0.85, 1.65)	0.87 (0.64, 1.18)	1.05 (0.72, 1.54)	1.34 (0.85, 2.13)
Another Gender Identity	1.00 (0.79, 1.25)	1.08 (0.90, 1.28)	1.22 (0.95, 1.57)	1.52 (1.15, 2.01)
Hormone use
No	1.00 (ref)	1.00 (ref)	1.00 (ref)	1.00 (ref)
Yes	1.09 (0.90, 1.32)	0.97 (0.84, 1.13)	0.88 (0.72, 1.08)	1.11 (0.86, 1.42)
Sex work (lifetime)
No	1.00 (ref)	1.00 (ref)	1.00 (ref)	1.00 (ref)
Yes	0.94 (0.73, 1.21)	1.15 (0.96, 1.39)	1.02 (0.80, 1.31)	1.74 (1.35, 2.22)

aPR = Adjusted Prevalence Ratio. 95% CI = 95% Confidence Interval. Models do not contain income because missing data.

^*statistically significant at p<0.05.

Sensitivity analyses adding physical IPV and sexual IPV to the models slightly attenuated the association between T-IPV and each of the outcomes, but associations remained significant or marginally significant: serious psychological distress (APR=1.29, 95%CI=1.10, 1.51), PTSD (APR=1.39, 95%CI=1.20, 1.61), AUD (APR=1.18, 95%CI=0.98, 1.42), and DUD (APR=1.20, 95% CI=0.97, 1.48) (data not shown but available on request).

Discussion

This study demonstrates that the T-IPV scale is a reliable and unidimensional measure of partner controlling tactics that are specific to the experience of transgender individuals. The measure has strong construct validity, substantially building on a previous iteration of the scale that was tested in transmasculine individuals by including more items to increase content validity and reliability, and undergoing rigorous psychometric validation in a large sample of transfeminine adults. Having yielded a valid and reliable scale, this study found that nearly half of the sample had experienced T-IPV at some point in their lives.

T-IPV was significantly associated with poor mental health and substance use, even after adjusting for other types of IPV experienced. Clinicians and service providers working with transgender clients should screen for T-IPV in addition to screening for physical and sexual IPV to avoid missing cases of IPV, as 40% of IPV survivors had experienced T-IPV but not physical or sexual IPV. Anecdotally, some service providers report that transgender individuals may be deprioritized for shelters or other services because screening tools used during intake do not capture trans-specific dynamics of abuse. As a result, service providers can get an incomplete picture of the scope of the violence experienced. This brief screening tool can be used by providers to identify T-IPV in transgender clients and more accurately triage severity of abuse and allocate resources to transgender clients. With regards to clinical settings, cisgender women most commonly first disclose to healthcare providers when seeking formal help for IPV (World Health (Organization, 2013). in part because routine screening for IPV in cisgender women is recommended by the US Preventive Services Task Force. The existence of a brief screener validated for transgender people should empower clinicians to screen for T-IPV in transgender populations as well so that transgender survivors can be connected to resources and interventions. Research shows that when clinicians are able to make a warm handoff to trans-competent mental health provider or anti-violence organization, transgender patients who screen positive for IPV complete the referral at high rates (Hartwick et al., 2021). Clinicians without strong local anti-violence resources should familiarize themselves with national LGBTQ-led anti-violence organizations (e.g. FORGE, the Northwest Network, The Network/La Red) that offer trans-focused hotlines, safety planning resources, and educational materials about intimate partner violence nationally. However, ultimately, local organizations need to be strengthened to handle local needs for transgender survivors such as housing or accompaniment to legal proceedings or sexual assault forensic exams. This process can be best accomplished through greater funding and training for providers, as well as passing legal protections against discrimination by providers.

The transfeminine participants in this study reported a high burden of T-IPV. Indeed, lifetime and past-year prevalence was higher than what has been reported using other scales to measure T-IPV in other samples of transgender individuals. While this higher prevalence is unsurprising given that this scale has eight items compared to other scales with four (Peitzmeier et al., 2019) or three (James et al., 2016) items, even the individual item prevalence was higher in our current sample. Compared to our previous work administering a four-item version of the scale with a sample of transmasculine individuals, lifetime prevalence was higher in this sample for three of the four items found in both versions of the scale (Peitzmeier et al., 2019). Lifetime prevalence of the three items shared between the USTS and our current scale was also higher in our sample for all three items (James et al., 2016). While the prevalence cannot be directly compared given changes to item wording, the item suggesting that no one else would partner with them because they are transgender was endorsed by 13.4% of transmasculine respondents in our previous work, but 22.1% in the current sample. Blackmail via outing was endorsed by 7.4% in our previous work, but 12.5% in the current sample, similar to the 11% of respondents in the USTS who reported experiencing this behavior (James et al., 2016). Hiding or sabotaging hormones or other items needed for transition was endorsed by 2.7% in our previous work, but 10.6% in the current sample. This also exceeds the prevalence of those reporting their partner would not let them have their hormones in the USTS (3%) (James et al., 2016). In our sample, 31.7% reported being told they were not a “real” man or woman, compared to 25% of USTS respondents. In the current scale, we ask two questions about coercive control of transition (20.7%) and coercive control of gender presentation (13.5%); in our previous work we had a single item that combined these two concepts and was endorsed by 33.6%. Differences in item wording and major differences in sampling strategy may account for much of these differences in prevalence.Yet, this sample of transfeminine adults clearly has a markedly high prevalence of T-IPV, raising questions about whether transfeminine individuals may be particular at risk of T-IPV as compared to transmasculine spectrum individuals. Measuring T-IPV in samples of comparably recruited transmasculine and transfeminine individuals would help clarify whether transfeminine individuals may be particularly at risk for this unique type of IPV, but clearly T-IPV is a common experience for both transmasculine and transfeminine individuals. More resources are needed to address IPV across the trans community. We also found no association between the participant’s specific gender identity within the transfeminine spectrum and T-IPV experience, and no differences in T-IPV experience in those who recently used hormones to transition, and those who did not. Findings therefore suggests that transfeminine individuals are at high risk of T-IPV regardless of their specific gender identity or transition status, but more research is needed on how gender identity, expression, and transition may impact T-IPV risk within transfeminine communities.

Within this sample, specific subgroups were found to be differentially at-risk for experiencing T-IPV, including older transfeminine adults, those with a history of sex work, and those experiencing food insecurity. Since our analysis examined lifetime T-IPV, it follows that transfeminine individuals would be more likely to report a lifetime experience of T-IPV as they age. Our findings are consistent with other studies identifying sex work as a risk factor for other types of IPV in transgender populations (James et al., 2016, Goldenberg et al., 2018, Logie et al., 2017). Concerning food insecurity, while no prior research has linked food insecurity to IPV among transgender individuals, our findings are consistent with prior research showing that IPV can restrict access to essential resources for transgender people (Hughto et al., 2015). The relationship between IPV and food insecurity is particularly concerning, given that food insecurity can increase dependence on partners to meet basic needs and promote staying in unhealthy or abusive relationships out of necessity. Finally, to our knowledge, this is one of the first studies to examine distribution of any type of IPV with partnership type and status and partner gender in a sample of transgender individuals (Peitzmeier et al., 2020a). Those within marriages or civil unions, casually dating, or in some “other” relationship type had a higher lifetime prevalence of T-IPV than those who reported being single or in a committed relationship, and those who reported recently partnering with a transgender women were less likely to report lifetime T-IPV. Increased research on partner and partnership characteristics, with a focus on collecting partnership-level data, is critical for informing interventions with perpetrators or potential perpetrators of IPV against transgender individuals.

While many subgroups were found to be at increased risk for T-IPV, no statistically significant differences in prevalence of lifetime T-IPV by race/ethnicity were found, even in this large and diverse sample with nearly half of participants identifying as people of color. This finding is unexpected given higher rates of many types of violence toward trans women of color, yet consistent with Woulfe and Goodman’s null findings around race/ethnicity and identity abuse (Woulfe and Goodman, 2018). Woulfe and Goodman hypothesized that race-neutral items on their identity abuse scale, much like the items on our T-IPV scale, may not capture the specificities of abuse for LGBTQ people of color. Race-specific examples might include being sexually objectified by a partner as both a transgender woman and a Latina, Black, or Asian woman, or being told by a partner that their gender identity is invalid because they “can’t be both” transgender and a person of color. Notably, the USTS did find elevated rates of trans-specific IPV in American Indian transgender women and multiracial transgender women, though no elevated rates of trans-specific IPV were reported in any other non-white group (James et al., 2016). We lacked the sample size in our study to report on American Indian participants specifically, though we did find higher rates of T-IPV among multiracial participants than white participants (62.1% versus 47.8%) that did not reach significance at p<0.05. More research using an intersectional lens (Bowleg, 2008) should be conducted on the IPV experiences of transgender people of color, to further understand how T-IPV may manifest in the context of multiple marginalized identities.

These eight trans-specific IPV items are linked by their strong focus on gender: they measure behaviors that invalidate gender identity, disrupt or police gender expression, or leverage societal transphobia as a threat to physical or emotional safety. We have strong evidence from factor analyses that these eight items share a single underlying construct. This may be surprising, given that some items in the T-IPV scale that relate to the coercive control of gender expression appear to be more like items that have traditionally been thought of as controlling behaviors, while others (such as intentional misgendering as an insult) appear to be more like traditional forms of psychological IPV. Indeed, research with cisgender heterosexual women often makes a distinction between psychological IPV (e.g., insults, humiliation, intimidation, and threats) and controlling behaviors (e.g., isolation from friends and family, surveillance or restriction of movement, etc.) (Heise et al., 2019). Despite including items that conceptually represent both controlling behaviors and psychological forms of IPV, the results of this analysis suggest that these acts of T-IPV may be motivated by the same fundamental impulse on the part of perpetrators and/or experienced similarly by survivors, yielding a unidimensional construct.

Limitations

This study had several limitations. Robust scale development typically begins with a potential item set 5 to 10 times larger than the final goal size; testing over 100 items would not be unusual (Borjesson et al., 2003). Because the focus of the baseline LITE study survey was not just on IPV, length constraints prohibited this. Potential items were drawn from literature review and consultation with experts in the field of transgender health and IPV. Ideally, a scale would be developed entirely de novo through exhaustive qualitative work eliciting a large pool of potential items considered by transgender respondents as trans-specific intimate partner violence, which would then be greatly reduced via EFA and CFA. This work is currently underway (Stephenson et al., 2020). Future research should validate the scale in a large sample of transgender people, including transmasculine and transfeminine individuals, that will additionally permit testing gender differences in the psychometric performance of the scale. This cross-sectional study was unable to prove causality between T-IPV and the outcomes of interest, and may have incompletely captured bidirectional relationships between T-IPV and potential confounders included in multivariable models. Additional research is needed to understand differences in the prevalence and correlates of T-IPV in different subgroups of transfeminine individuals, particularly by age group and by race and ethnicity, as well as prevalence and correlates of T-IPV in transmasculine spectrum individuals.

Despite these limitations, this analysis contributes to the literature on experiences of intimate partner violence among transgender people by developing and validating a scale to capture the transgender-specific tactics used by intimate partners to enact violence against transgender people. Thus, these findings introduce a valuable tool for health and service providers who serve transgender patients and clients.

Conclusion

This eight-item scale can be used to assess a unique and impactful form of psychological and controlling forms of T-IPV experienced among transgender populations, which is correlated with but distinct from physical and sexual IPV. The measure can be used to screen for abuse tactics directed against transgender survivors in clinical and research settings, filling a key gap in existing tools to accurately assess IPV against transgender people. The high prevalence of lifetime T-IPV found in this sample (49%) demonstrates that transfeminine individuals experience endemic abuse through the leveraging of societal transphobia as a tool of power and control within relationships. These experiences of T-IPV are associated with negative health outcomes and urgently necessitate public health intervention.

Highlights.

• Transgender people are at elevated risk of intimate partner violence (IPV).

• IPV can take on trans-specific dimensions, e.g. the abuser threatening to out a partner.

• We developed an 8-item scale to screen for trans-specific IPV (T-IPV).

• Our scale shows strong reliability and construct validity.

• Our scale can be used to screen and refer trans individuals for IPV services.