Development and Initial Validation of Three Culturally-Sensitive Measures of Intimate Partner Violence for Sexual and Gender Minority Populations

Abstract

Measures of intimate partner violence (IPV) have largely been developed and validated in heterosexual, cisgender samples, with little attention to whether these measures are culturally appropriate for sexual and gender minority (SGM) populations. However, rates of IPV are 2-3 times higher among SGM than heterosexual populations, highlighting the importance of culturally appropriate measures of IPV for SGM populations. In this paper, after reviewing key problems with the use of existing IPV measures with SGM samples, we describe the development of a toolkit of new and adapted measures of IPV for use with SGM assigned female at birth (SGM-AFAB) populations, including an adapted version of the Conflict Tactics Scale Revised, an adapted measure of coercive control, and the newly developed SGM-Specific IPV Tactics Measure. Using data from a sample of 352 SGM-AFAB individuals, we then test the psychometric properties of these three measures, including their factor structures, internal reliability, and convergent/divergent validity. Results provide initial evidence of the reliability and validity of each measure. Together, these three measures comprise a culturally appropriate and psychometrically validated measurement toolkit for studying a broad range of IPV tactics among SGM-AFAB that will help build a foundation for more in-depth research into IPV in SGM populations.

Research examining intimate partner violence (IPV; physical, sexual, or psychological abuse by a romantic/sexual partner) has focused almost exclusively on relationships in which partners are heterosexual and cisgender (i.e., individuals whose current gender identity is the same as their sex assigned at birth) (see Edwards, Sylaska, & Neal, 2015). However, a number of probability sampled studies have shown that IPV is 2-3 times more prevalent in relationships in which at least one partner is a sexual or gender minority (SGM) than in relationships between heterosexual cisgender men and women (e.g., McLaughlin, Hatzenbuehler, Xuan, & Conron, 2012; Walters, Chen, & Breiding, 2013). Consequently, recent studies have begun to examine IPV among SGM individuals, but most of this research has used measures of IPV that were designed for and validated on heterosexual, cisgender samples (e.g., Conflict Tactics Scale 2; (CTS2Straus, Hamby, Boney-McCoy, & Sugarman, 1996). This is problematic as these measures may not be culturally appropriate for use with SGM populations and do not capture the unique types of IPV they may experience (i.e., SGM-specific IPV; (Moradi, Mohr, Worthington, & Fassinger, 2009).

We sought to address these concerns by creating a measures toolkit that can assess a range of types of IPV (psychological, physical, sexual, SGM-specific IPV; coercive control; and injury by partner) in a culturally appropriate manner with SGM assigned female at birth (SGM-AFAB). We did so by adapting existing measures of IPV, developing a new measure, and psychometrically testing these new and adapted measures among SGM-AFAB. In the next section, we review two aspects of existing measures that make their use with SGM problematic: their lack of cultural sensitivity due to heteronormative and cisnormative content and their failure to capture SGM-specific IPV. We then briefly review constructs linked with IPV in heterosexual samples to inform the validation of the new and adapted measures from the current study.

Heteronormativity and Cisnormativity in IPV Measures

As Follingstad and Rodgers (2013) note, most IPV measures were designed to understand IPV perpetrated by a cisgender man against his cisgender female partner. Consequently, these measures often contain language and content that is not appropriate for SGM participants due to cisnormativity and heteronormativity– the often subconscious beliefs that all people are cisgender and heterosexual. Therefore, these measures often require adaptations to make them culturally appropriate for use with SGM populations.

One particularly problematic example of this lack of cultural sensitivity is the heterocentric assumptions about sex within items used to assess sexual IPV.. For example, a CTS2 item asks how often “I made my partner have sex without a condom.” The use of the term “condom” is based on a heterocentric definition of sex as penile-vaginal intercourse, and the item fails to ask about other forms of protection against sexually transmitted infections more commonly used by female same-sex couples (e.g., dental dams, gloves; see McNair, 2005 for review). Consequently, this item can not only be alienating to SGM, but also miss instances of sexual IPV in which a partner is forced to engage in unsafe sex. As another example, the CTS2 includes separate items to assess acts of sexual IPV that involve “sex” versus “oral or anal sex.” This differentiation is heterocentric as it implies that penile-vaginal penetration is the norm (i.e., “real sex”) and that oral and anal sex are not “real sex.” For many SGM individuals, the word “sex” refers to oral and/or anal sex and other sexual activities, making this item potentially invalidating or unclear and contributing to inaccurate reporting in SGM samples (Nichols, 2004). Additionally, most existing IPV measures, including the CTS2, use gendered pronouns, which makes them inappropriate for use with individuals with non-binary gender identities (e.g., genderqueer) who may prefer gender neutral pronouns (e.g., they/them).

To address these concerns, some researchers have adapted existing measures to make them more appropriate for SGM samples (e.g., Balsam & Szymanski, 2005). However, these studies rarely provide details about how items were adapted or test the psychometric validity of the adapted measures. Others have created and psychometrically tested novel measures of IPV for SGM populations (e.g., McClennen, Summers, & Daley, 2002). Because they were not adapted from existing measures, these measures do not allow for comparison of IPV rates between SGM and heterosexual populations. Further, researchers have used measures of IPV that have not been adapted for SGM populations without psychometrically evaluating these measures in SGM populations (e.g., Craft & Serovich, 2005). There is a need for measures of IPV that have been adapted to be culturally appropriate for SGM populations, have been psychometrically evaluated in SGM populations, and allow for comparisons with non-SGM populations.

SGM-Specific IPV

In addition to experiencing the same forms of IPV that cisgender, heterosexual individuals can experience, SGM individuals can also experience unique forms of IPV as a result of their stigmatized identities (e.g., Balsam & Szymanski, 2005; Edwards et al., 2015). These unique forms of IPV reflect the leveraging of the stigmatization of SGM individuals to cause psychological harm or to control an intimate partner (Balsam & Szymanski, 2005). Several SGM-specific IPV tactics have been identified, with the most frequently studied type being threats to disclose a partner’s SGM identity without their consent (i.e., “outing”; (e.g., Balsam & Szymanski, 2005; McClennen et al., 2002). Individuals may also shame their partner for being SGM by using anti-SGM language (Woulfe & Goodman, 2018) or leverage the stigmatization of SGM individuals to verbally abuse and control their partner or justify their abuse (Woulfe & Goodman, 2018). Alternatively, individuals may shame their partner for not being “queer enough” or question their partner’s SGM identity (e.g., accusing a partner of being straight; (Balsam & Szymanski, 2005; Woulfe & Goodman, 2018). Individuals may also control or hurt their SGM partner by isolating them from the SGM community (Woulfe & Goodman, 2018). Unsurprisingly, IPV measures developed to assess IPV in relationships between heterosexual cisgender men and women do not capture these unique forms of IPV.

In recent years, a few studies have assessed some of these SGM-specific IPV tactics (e.g., Balsam & Szymanski, 2005; Woulfe & Goodman, 2018). However, the measures used in these studies are limited. First, most studies have used idiosyncratic measures that assess one or two SGM-specific IPV tactics. Consequently, there is little consistency across studies. Second, these measures have generally not been psychometrically validated. In one recent exception, Woulfe and Goodman (2018) developed and psychometrically evaluated the Identity Abuse Scale, which assesses several SGM-specific IPV tactics (Woulfe & Goodman, 2018). While the Identity Abuse Scale demonstrated psychometric validity and is an important advance in the SGM-specific IPV literature, it is limited in that it assesses victimization but not perpetration. As IPV is frequently bi-directional – in both heterosexual, cisgender and SGM samples (Edwards et al., 2015) – it is important to capture both IPV victimization and perpetration.

Validating Measures of IPV for SGM Samples

Our review of the literature indicated that comprehensive and culturally appropriate assessment of IPV would require some new and some adapted measures of psychological IPV, physical IPV, sexual IPV, injury by partner, coercive control, and SGM-specific IPV victimization and perpetration. Following adaptation and development, these measures would need to be psychometrically evaluated in SGM populations. As part of this evaluation, adapted and novel measures would be expected to produce associations between IPV and other measures that are similar to associations from the existing literature. This would provide evidence of the new measures’ construct validity (i.e., convergent and divergent validity). For example, types of IPV would be expected to be associated with one another, such that individuals who experienced one type of IPV (e.g., psychological IPV) would be more likely to experience other types of IPV (Lucente, Fals-Stewart, Richards, & Goscha, 2001; Mechanic, Weaver, & Resick, 2008; Scheer, Woulfe, & Goodman, 2018). Additionally, adapted and novel measures of IPV would be expected to demonstrate convergent validity via moderate to strong positive associations with variables linked to IPV in the existing literature: antisocial behavior (Andrews, Foster, Capaldi, & Hops, 2000); jealousy (Kerr & Capaldi, 2011); cohabitation with partner and more serious relationship status (Kaukinen, 2014); and ineffective couple communication (i.e., invalidation, escalation, and withdrawal; (Marshall, Jones, & Feinberg, 2011). Similarly, they would be expected to demonstrate divergent validity via weak associations with less conceptually related variables, like relationship quality (Linder, Crick, & Collins, 2002) and relationship negotiation (constructive conflict resolution strategies; (Lucente et al., 2001; Straus et al., 1996).

Current Study

In the current study, we aimed to create a comprehensive set of culturally appropriate measures capturing a broad range of types of IPV among SGM. Measure adaptations (described below) were guided by the goal of making them culturally sensitive and thus appropriate for use with an SGM sample. Following measure development, we performed psychometric evaluation, examining the internal reliability, factor structure, and content validity of these measures among a sample of SGM-AFAB individuals. We examined construct validity by testing associations among IPV subscales and between IPV subscales and constructs associated with IPV among heterosexuals. Based on existing research (see summary above), we expected the following:

1. Moderate to strong positive associations among all types of IPV.

2. Relationship negotiation would be moderately positively correlated with psychological IPV but weakly correlated with all other types of IPV.

3. Ineffective couple communication patterns, antisocial behavior, and participant/partner jealousy would demonstrate moderate positive associations with all types of IPV.

4. Relationship quality would have weak negative associations with all types of IPV.

5. Cohabitation with partner and more serious relationship status would be associated with increased risk for all types of IPV.

Methods

Procedure: Measure Adaptation and Development

A review of the literature indicated that a comprehensive IPV measurement toolbox would require assessment of psychological IPV, physical IPV, sexual IPV, injury by partner, coercive control, and SGM-specific IPV. We elected to adapt the CTS2 because it captures several important types of IPV (psychological, physical, sexual, injury) and an adapted version would allow for comparisons between heterosexual and SGM populations.

The CTS2 does not assess coercive control, which is considered to be an important component of IPV (Hamberger, Larsen, & Lehrner, 2017; Stark, 2013). Therefore, we developed a measure of coercive control. After a thorough review of the literature, we selected two commonly used measures that had been previously used with SGM samples and had limited content overlap with psychological IPV: Coercive Behaviors Measure and the National Intimate Partner and Sexual Violence Survey (NISVS; (Black et al., 2011; Frankland & Brown, 2014). Given that neither of these measures covered the range of coercive controlling behaviors identified in the literature (intimidation, surveillance, social isolation, threats, dominance, and financial control tactics; (Hamberger et al., 2017; Stark, 2013), we selected items from both to create a brief measure that more fully captured the range of coercive controlling behaviors.

Given the limitations of measures available when data collection began in 2016, we also elected to develop a novel measure of SGM-specific IPV. Based on existing literature, we defined SGM-specific IPV as actions that leverage the stigmatization of SGM individuals to cause psychological harm or to control an intimate partner (Balsam & Szymanski, 2005) and identified two types of SGM-specific IPV to include in the measure: outing and social isolation.

An iterative process was used to develop and adapt all three of these measures. First, the research team examined the existing literature to inform measure adaptation and development. Second, based on this review, we generated an initial set of items. Third, feedback was sought from several sources within the target population, including study team members and an SGM youth advisory council. Fourth, feedback was used to refine measures. Steps three and four were repeated until members of the target population and research team deemed the measures appropriate. For all measures, we changed “his/her” to “their” to remove gender binary language and piped in their partner’s name in place of “my partner.” Other specific adaptations are described in the measures section.

Participants and Procedures: Psychometric Testing

Participants were drawn from FAB400 (Whitton, Dyar, Newcomb, & Mustanski, 2019), an ongoing cohort study of 488 young SGM-AFAB. FAB400 focuses on SGM-AFAB’s health, development, and intimate relationships. FAB400 employs a merged cohort accelerated longitudinal design (Galbraith, Bowden, & Mandler, 2017), and includes two cohorts: (1) a late adolescent cohort recruited for FAB400 in 2016-2017 (N = 400; 16-20 years old at baseline), and (2) a young adult cohort comprised of SGM-AFAB participants from Project Q2 (Mustanski, Garofalo, & Emerson, 2010), a longitudinal study of SGM youth that began in 2007 (N = 88; 23-32 years old at the FAB400 baseline assessment). Inclusion criteria for FAB400 and Project Q2 were identical, requiring participants to be 16-20 years old when they enrolled, speak English, and either identify with an SGM label, report same-gender attractions, or report same-gender sexual behavior. To enroll in FAB400, participants were also required to be female-assigned at birth. Each cohort was recruited using an incentivized snowball sampling approach.

In 2016-2017, all 488 participants completed the FAB400 baseline assessment. Participants were paid $50 for the assessment, which included a battery of self-report measures. The study protocol was approved by the Institutional Review Board (IRB) at Northwestern University with a waiver of parental permission for participants under 18 years of age under 45 CFR 46, 408(c). Participants provided written informed consent, and mechanisms to safeguard participant confidentiality were used (i.e., a federal certificate of confidentiality).

Participants were asked to provide the first name and last initial of their three most recent sexual and/or romantic partners in the past six months and indicate which of these three partners was most significant (i.e., “… the person that you spent the most time with, were most serious about, or who had the biggest effect on you.”). The analytic sample included 352 participants who had a romantic relationship with their most significant partner. The majority of the analytic sample were part of the adolescent cohort (78.1%) and the average age was 20.4 (SD = 3.9). Notably, this sample is diverse in race/ethnicity (37.8% non-Hispanic African American, 26.4% Latinx, 23.3% non-Hispanic European American, 12.5% other racial/ethnic identity), gender identity (76.4% cisgender women, 7.7% transgender or male identified, 16.0% non-binary/genderqueer identified), and sexual orientation identity (37.5% bisexual; 26.7% lesbian, 17.6% pansexual, 10.2% queer, 7.9% identified with other labels).

Measures

Measures of IPV

SGM Conflict Tactics Scale 2 is an adapted form of the CTS2 (Straus et al., 1996)¹ used to assess victimization and perpetration of psychological, physical, injury, and sexual IPV as well as conflict negotiation. Several adaptations were made to the CTS2 (see Tables 1-2 for items). We changed “his/her” to “their” to remove gender binary language. The CTS2 has three items about sexual IPV that includes oral/anal sex and three parallel items about “sex.” To avoid this heteronormativity, the three items referring to “oral/anal” sex were removed, those referring to “sex” were retained, and an inclusive definition of sex was provided. The item, “I made my partner have sex without a condom,” was adapted to “I refused to use the safe sex methods that [partner name] requested to use (e.g., a condom, dental dam, etc.)” to better reflect the variety of safe sex methods that may be appropriate for SGM-AFAB. Given evidence that drug/alcohol-incapacitated sexual assault is one of the most prevalent forms of sexual assault - describing up to 70% of sexual assaults (Krebs, Lindquist, Warner, Fisher, & Martin, 2009; Lawyer, Resnick, Bakanic, Burkett, & Kilpatrick, 2010), we added the item “My partner had sex with me when I was unable to consent because I was so drunk, high, or passed out” and a parallel perpetration item to broaden the range of sexual IPV tactics captured to include this form of sexual assault.

Table 1.

CFA for Adapted CTS2 – Victimization

Factor	Item #	Item	Factor Loading	SE	R2
Psychological IPV	6	[Partner name] swore at me.	.70	.04	.49
	34	[Partner name] shouted or yelled at me.	.78	.03	.61
	48	[Partner name] stomped out of the room or house or yard during a disagreement.	.75	.04	.57
	62	[Partner name] did something to spite me.	.60	.07	.36
	24	[Partner name] called me names, insulted them, or treated me disrespectfully in front of others.	.90	.03	.80
	28	[Partner name] destroyed something belonging to me.	.88	.06	.77
	60	[Partner name] accused me of being a lousy partner.	.71	.05	.51
	64	[Partner name] threatened to hit or throw something at me.	.99	.03	.99
Physical IPV	8	[Partner name] threw something at me that could hurt.	.89	.04	.78
	10	[Partner name] twisted my arm or hair.	.93	.03	.87
	18	[Partner name] pushed or shoved me.	.93	.02	.86
	44	[Partner name] grabbed me.	.87	.04	.76
	52	[Partner name] slapped me.	.91	.04	.83
	20	[Partner name] used a knife or gun on me.	.70	.11	.50
	26	[Partner name] punched or hit me with something that could hurt.	.96	.02	.92
	32	[Partner name] choked me.	.85	.05	.73
	36	[Partner name] slammed me against a wall.	.95	.03	.89
	42	[Partner name] beat me up.	.98	.02	.97
	58	[Partner name] burned or scalded me on purpose.	-	-	-
	68	[Partner name] kicked me.	.87	.07	.75
Injury	11	I had a sprain, bruise, or small cut because of a fight with [partner name].	.94	.05	.89
	21	I passed out from being hit on the head by [partner name] in a fight.	-	-	-
	29	I went to a doctor because of a fight with [partner name].	.88	.04	.78
	39	I needed to see a doctor because of a fight with [partner name], but I didn't.	.92	.03	.85
	53	I had a broken bone from a fight with [partner name].	-	-	-
	65	I felt physical pain the next day because of a fight we had.	.99	.02	.99
Sexual	16	[Partner name] refused to use the safe sex methods that I requested to use.	.55	.12	.30
	46	[Partner name] used force (like hitting, holding down, or using a weapon) to make me have sex.	.99	.09	.99
	50	[Partner name] insisted on having sex when I did not want to (but did not use physical force).	.73	.09	.53
	70	[Partner name] used threats to make me have sex.	-	-	-
	73	[Partner name] had sex with me when I was unable to consent because I was so high, drunk, or passed out.	-	-	-
Negotiation	2	[Partner name] showed care for me even though we disagreed.	.55	.08	.31
	4	[Partner name] explained their side of a disagreement to me.	.50	.07	.25
	14	[Partner name] showed respect for my feelings about an issue.	.37	.06	.14
	38	[Partner name] was sure we could work out a problem.	.69	.05	.48
	56	[Partner name] suggested a compromise to a disagreement.	.59	.04	.35
	72	[Partner name] agreed to try a solution I suggested.	.76	.05	.58

IPV terms are treated as dichotomous, probit regression coefficients presented for IPV items. Negotiation items were treated as continuous, standardized regression coefficients are presented for negotiation. All factor loadings were significant at p < .001. "-" indicates items that were dropped from CFAs due to sparse reporting. R-squared represents the proportion of the continuous latent response variable underlying each dichotomous item that is explained by the factor. "-" indicates items that were dropped from CFAs due to sparse reporting. For types of IPV with minor and severe subscales on the CTS2, items in the severe subscale are denoted by italics. CFA = confirmatory factor analysis; CTS2 = Conflict Tactics Scale 2; IPV = intimate partner violence.

Table 2.

CFA for Adapted CTS2 – Perpetration

Factor	Item #	Item	Factor Loading	SE	R2
Psychological IPV	5	I swore at [partner name].	.68	.05	.46
	33	I shouted or yelled at [partner name].	.82	.03	.68
	47	I stomped out of the room or house or yard during a disagreement.	.68	.95	.46
	61	I did something to spite [partner name].	.47	.07	.22
	23	I called [partner name] names, insulted them, or treated [partner name] disrespectfully in front of others.	.85	.05	.73
	27	I destroyed something belonging to [partner name].	.86	.06	.74
	59	I accused [partner name] of being a lousy partner.	.57	.06	.33
	63	I threatened to hit or throw something at [partner name].	.95	.03	.91
Physical IPV	7	I threw something at [partner name] that could hurt.	.75	.06	.56
	9	I twisted [partner name]s arm or hair.	.68	.08	.46
	17	I pushed or shoved [partner name].	.91	.03	.83
	43	I grabbed [partner name].	.84	.04	.71
	51	I slapped [partner name].	.79	.06	.63
	19	I used a knife or gun on [partner name].	.74	.14	.55
	25	I punched or hit [partner name] with something that could hurt.	.95	.04	.91
	31	I choked [partner name].	.78	.07	.60
	35	I slammed [partner name] against a wall.	.81	.07	.66
	41	I beat up [partner name].	.96	.04	.92
	57	I burned or scalded [partner name] on purpose.	-	-	-
	67	I kicked [partner name].	.78	.09	.61
Injury	12	[Partner name] had a sprain, bruise, or small cut because of a fight with me.	.95	.05	.89
	22	[Partner name] passed out from being hit on the head in a fight with me.	.74	.16	.55
	30	[Partner name] went to a doctor because of a fight with me.	.86	.08	.74
	40	[Partner name] needed to see a doctor because of a fight with me, but they didn't.	-	-	-
	54	[Partner name] had a broken bone from a fight with me.	-	-	-
	66	[Partner name] felt physical pain the next day because of a fight we had.	.93	.05	.86
Sexual	15	I refused to use the safe sex methods that [partner name] requested to use (e.g., a condom, dental dam, etc.).	.56	.17	.32
	45	I used force (like hitting, holding down, or using a weapon) to make [partner name] have sex.	-	-	-
	49	I insisted on having sex when [partner name] did not want to (but did not use physical force).	.60	.16	.36
	69	I used threats to make [partner name] have sex.	-	-	-
	74	I had sex with [partner name] when they were unable to consent because they were so high, drunk, or passed out.	-	-	-
Negotiation	1	I showed [partner name] I cared even though we disagreed.	.50	.07	.25
	3	I explained my side of a disagreement to [partner name].	.42	.07	.17
	13	I showed respect for [partner name]'s feelings about an issue.	.57	.08	.32
	37	I said I was sure we could work out a problem.	.70	.05	.49
	55	I suggested a compromise to a disagreement.	.54	.06	.30
	71	I agreed to try a solution [partner name] suggested.	.77	.06	.60

IPV tems are treated as dichotomous, probit regression coefficients presented for IPV items. Negotiation items were treated as continuous, standardized regression coefficients are presented for negotiation. All factor loadings were significant at p < .001. "-" indicates items that were dropped from CFAs due to sparse reporting. For types of IPV with minor and severe subscales on the CTS2, items in the severe subscale are denoted by italics. CFA = confirmatory factor analysis; CTS2 = Conflict Tactics Scale 2; IPV = intimate partner violence.

The final adapted measure included 74 items. On each item, participants indicated how frequently the IPV tactic described occurred: 0 (never), 1 (once), 2 (twice), 3 (3-5 times), 4 (6-10 times), 5 (11-20 times), 6 (more than 20 times), and 7 (not in the past 6 months, but it did happen before). For current analyses, 7 was treated as 0. Subscale scores were calculated by taking the mean of subscale items. The same response scale and methods of scoring was used for the SGM-Specific IPV Tactics Scale and the Coercive Control Measure, described next, for consistency.

Coercive Control Measure was created by combining items assessing victimization from two measures: 5 items from the Coercive Behaviors Scale (Frankland & Brown, 2014) and 3 items from the 2010 National Intimate Partner and Sexual Violence Survey (Black et al., 2011). We created 8 parallel items assessing perpetration, for a total of 16 items (see Table 3 for items; see procedure section for more information on measure development).

Table 3.

CFAs for SGM-Specific and Coercive Control IPV Measures

Model	Item #	Item	Factor Loading	SE	R2
SGM-Specific - V	1	[Partner name] threatened to out me to my friends, family, or other people in my life if I didn't do what they wanted.	.89	.15	.79
	3	[Partner name] repeatedly told me how alone I would be if I left them, because other people don't know about or aren't accepting of my sexuality.	.84	.09	.70
	5	[Partner name] threatened to turn people in the queer community against me, or spread rumors about me in this community.	.92	.10	.85
	7	[Partner name] outed me --to friends or family, at school, work, or in other social circles.	.68	.13	.46
	9	[Partner name] forced or coerced me into public displays of affection (e.g., hand holding, kissing, etc.) that they knew I wasn't comfortable with.	.60	.12	.36
SGM-Specific – P	2	I threatened to out [partner name] to their friends, family, or other people in their life if [partner name] didn't do what I wanted.	.88	.10	.78
	4	I repeatedly told [partner name] how alone they would be if they left me, because other people don't know about or aren't accepting of their sexuality.	.99	.07	.99
	6	I threatened to turn people in the queer community against them or spread rumors about them in this community.	-	-	-
	8	I outed [partner name] --to friends or family, at school, work, or in other social circles.	.87	.11	.75
	10	I forced or coerced [partner name] into public displays of affection (e.g., hand holding, kissing, etc.) that I knew they weren't comfortable with.	.69	.15	.47
Coercive Control - V	1	[Partner name] monitored my time and made me account for my whereabouts.	.85	.05	.72
	3	[Partner name] controlled or limited my access to money and other financial resources.	.77	.09	.59
	5	[Partner name] made me feel afraid through looks, actions or gestures.	.75	.07	.56
	7	[Partner name] limited my use of the phone or computer.	.81	.06	.66
	9	[Partner name] made it difficult for me to see friends or family.	.89	.05	.78
	11	[Partner name] threatened to hurt themself or commit suicide when they were upset with me.	.84	.06	.71
	13	[Partner name] threated to hurt someone I love.	.78	.12	.61
	15	[Partner name] made decisions for me that should have been mine to make, such as the clothes I wear, things I eat, or friends I have.	.86	.05	.74
Coercive Control - P	2	I monitored [partner name]'s time and made them account for their whereabouts.	.74	.06	.54
	4	I controlled or limited [partner name]'s access to money and other financial resources.	.79	.10	.62
	6	I made [partner name] feel afraid through looks, actions or gestures.	.93	.05	.86
	8	I limited [partner name]'s use of the phone or computer.	.84	.97	.70
	10	I made it difficult for [partner name] to see friends or family.	.91	.06	.82
	12	I threatened to hurt myself or commit suicide when I was upset with [partner name].	.85	.05	.72
	14	I threated to hurt someone [partner name] loves.	.98	.05	.95
	16	I made decisions for [partner name] that should have been theirs to make, such as the clothes they wear, things they eat, or friends they have.	.87	.04	.76

Probit regression coefficients presented. All factor loadings were significant at p < .001. Actual item text is only presented for the victimization subscales. "-" indicates items that were dropped from CFAs due to sparse reporting

The SGM-specific IPV Tactics Scale was created to capture the unique types of IPV that may be experienced by SGM-AFAB. For each of five SGM-specific IPV behaviors, there was an item assessing victimization and a parallel item assessing perpetration. See Table 3 for items. See procedure section for more information on measure development

Convergent/Discriminant Validity Measures

Ineffective couple communication was measured using the Danger Signs Scale (Stanley, Markman, & Whitton, 2002), a four item measure of common ineffective couple communication patterns during disagreements (α = .76; e.g., “My partner criticizes or belittles my opinions, feelings, or desires”). Responses are provided on a scale of 1 (almost never) to 3 (frequently).

Relationship quality was measured using the 7-item Relationship Assessment Scale (α = .82; e.g., “How good is your relationship compared to most?”) (Vaughn & Matyastik Baier, 1999). Responses are provided on a scale of 1 to 5, and endpoint labels match each question, with the higher endpoint indicating higher relationship quality. For example, the response scale for the example item ranges from 1 (not good) to 5 (very good).

Relationship status was measured using the item “What is/was the status of this relationship with [partner name]?” Those who indicated they were “casually dating but not serious” were coded as casually dating (n = 66), “serious relationship (e.g., girlfriend/ boyfriend) – someone you dated for a while and felt/feel very close to” as serious relationship (n = 251), and responses of “engaged to be married,” “domestic partnership/civil union,” “married,” “a lifelong committed relationship but not married or in a domestic partnership/civil union” were coded as long-term committed relationships (n = 35).

Cohabitation was assessed using the item “Do/did you live with [partner name]?” with the response options no (0) and yes (1).

Antisocial behavior was assessed using an adapted version of the General Delinquency Scale (Elliott, Ageton, & Huizinga, 1982). We adapted the measure by adding minor forms of antisocial behavior from Achenbach Adult Self-Report, shortening it to 6 items (α = .70), and removing items with very low endorsement in prior research (Kim, Conger, Elder Jr, & Lorenz, 2003). Items: “steal money or take something that did not belong to you,” “act very mean to others,” “purposely damage or destroy property that did not belong to you,” “threaten to hurt someone,” “get into a physical fight,” and “physically attack someone.” Participants indicated how often they engaged in each behavior in the past year on a scale of 0 (never) to 4 (6+ times).

Jealousy was assessed using a 6 item scale developed for the current study, which includes two 3-item subscales: participant jealousy (α = .69) and partner jealousy (α = .86). Participant jealousy items include: “I am suspicious that [partner name] is secretly in a romantic and/or sexual relationship with someone else,” “I am jealous when [partner name] is around people they may be attracted to,” and “I have accused [partner name] of cheating.” Parallel items assess partner jealousy. Responses range from 1 (strongly disagree) to 7 (strongly agree).

Results

Less than .02% of data were missing and were handled using full information maximum likelihood. All analyses were conducted in Mplus Version 7.

Internal Reliability

To test internal reliability, we calculated Cronbach’s alpha (α) and McDonald’s omega (ω) for each subscale of the SGM-CTS2 and for SGM-Specific IPV and coercive control. Cronbach’s α for all subscales were acceptable (see Table 4), except those for sexual IPV, which were reduced by low inter-item correlations with refusal to use requested safe-sex methods, a pattern present in the CTS2 (Straus et al., 1996). As α is biased based on the number of items in a measure and when items are skewed (Sijtsma, 2009), we also reported ω, a less biased measure of internal consistency (Dunn, Baguley, & Brunsden, 2014). All ω values were acceptable.

Table 4.

Correlations among and with Intimate Partner Violence

							Internal Reliability Coefficients
	Psychological	Physical	Injury	Sexual	Coercive Control	SGM- Specific IPV	Cronbach's Alpha	McDonald's Omega
Psychological	-	.68**	.46**	.40**	.54**	.30**	.82	.92
Physical	.59**	-	.73**	.52**	.47**	.41*	.88	.97
Injury	.43**	.78**	-	.55*	.39**	.55*	.83	.96
Sexual	.32**	.56**	.62*	-	.42**	.48*	.48	.76
Coercive Control	.47**	.40*	.44*	.46*	-	.38*	.78	.94
SGM-Specific
Internal Reliability
Coefficients	.25*	.46*	.65*	.67*	.43*	-	.63	.89
Cronbach's Alpha	.80	.84	.71	.54	.73	.85	-	-
McDonald's Omega	.91	.96	.91	.85	.92	.96	.92	-
					Relationship Status
	Negotiation	Couple Conflict	Relationship Quality	Cohabitation	Serious Relationship	Committed Relationship	Antisocial Behavior	Jealousy: Self	Jealousy: Partner
Victimization
Psychological	.41**	.69**	−.23**	.54**	.17**	.26**	.39**	.50**	.51**
Physical	.18**	.40**	−.12*	.47**	.08	.13	.38**	.33**	.41**
Injury	.10*	.29**	−.09*	.11	−.04	.02	.31**	.22**	.24**
Sexual	.15**	.31**	−.16**	.03	−.01	.14	.24**	.30**	.30**
Coercive Control	.20**	.48**	−.21**	.08	.13*	.15*	.38**	.30**	.47**
SGM-Specific	.07	.29**	−.31**	−.06*	−.06	−.07	.17*	.17**	.20**
Perpetration
Psychological	.33**	.61**	−.18*	.15*	.16*	.21*	.44**	.47**	.49**
Physical	14*	.35**	−.09	.09	.07	.09	.39**	.32**	.37**
Injury	.01	.26**	−.14*	.10	−.03	−.04	.26**	.23**	.23**
Sexual	.003	.26**	−.09*	.06	−.04	.05	.22*	.29**	.24**
Coercive Control	.08	.40**	−.13*	.05	.10	.16*	.39**	.35**	.35**
SGM-Specific	.01	.16**	−.10*	−.04	−.10	−.09*	.21*	.09*	.12*

Top panel: correlations among types of IPV; correlations above the diagonal are victimization, below the diagonal are perpetration. Bottom panel: correlations between types of IPV and convergent/divergent validity variables.

^*p < .05;

^**p < .001.

Factorial Validity

Next, we conducted confirmatory factor analyses to test factorial validity, the extent to which items load on latent variables in an expected pattern. While we adapted the CTS2 to make it appropriate for use with SGM populations, these changes were not expected to alter the factor structure of the CTS2. Thus, we expected the SGM-CTS2 to demonstrate the same five-factor structure as the CTS2 (psychological IPV, physical IPV, sexual IPV, injury, and relationship negotiation). As the new/adapted measures of coercive control and SGM-specific IPV tactics were each expected to capture a single type of IPV, we tested unidirectional factor structures separately for victimization and perpetration for each measure.

Factorial Validity: SGM-CTS2.

As many of the more severe items on the SGM-CTS2 had low endorsement (< 10%), we dichotomized all items assessing IPV or injury by partner. Nearly all participants indicated negotiating with their partner in the past 6 months, so we treated negotiation items as continuous. The confirmatory factor analyses (CFAs) were conducted using diagonally weighted least squares estimation (WLSMV) and a probit link for dichotomized items. Probit factor loadings represent the variance shared by the factor and the latent continuous response variables underlying each dichotomous observed item (Agresti, 2003). Probit analyses also provide an R² for each indicator variable, which represents the proportion of variance in the latent continuous response variable explained by the model (Agresti, 2003). Using established guidelines (Comrey & Lee, 1992), we considered items for which the factor explained 20-29% of the variance in the continuous latent variable underlying the observed dichotomous item to be acceptable, 30-49% good, and 50% or more excellent. We examined the fit of the factor structure outlined by Straus et al. (1996), using the comparative fit index (CFI), Tucker-Lewis index (TLI), and root mean squared error of approximation (RMSEA). Good model fit was indicated by CFI and TLI values ≥ .95 and RMSEA values < .06 (Hu & Bentler, 1999). Chi-squared goodness of fit values are reported, but this test is overpowered in moderately sized samples.

We tested CFA models for victimization and perpetration separately. Items endorsed by fewer than 3 participants were excluded (6 removed from CFA for perpetration; 4 removed from CFA of victimization) as sparse data is problematic in CFAs with binary indicators (e.g., Flora & Curran, 2004; Wirth & Edwards, 2007).² In both models, we tested the 5 factor structure outlined by Straus and colleagues (1996), which includes factors representing each subscale. Factors were allowed to correlate with one another and each item was only allowed to load on the factor representing the subscale to which it belonged (i.e., no cross-loadings).

Model fit indicated that the 5 factor model fit well for victimization (χ²[485] = 628.95, p < .001; RMSEA = .03, CFI = .98, TLI = .98) and perpetration items (χ²[424] = 619.39, p < .001; RMSEA = .04, CFI = .94, TLI = .93) and factor loadings were acceptable to excellent. This indicates that the SGM-CTS2 has the same factor structure as the original CTS2. Of note, we also conducted a factor analysis with robust maximum likelihood estimation using continuous variables for IPV items endorsed by at least 5% of the sample (14 perpetration items; 13 victimization items). Results, not presented here, followed the same pattern as the probit CFA.

Factorial Validity: Coercive Control and SGM-Specific IPV.

As in the CFAs of the SGM-CTS2, we examined victimization and perpetration in separate models and used dichotomous items, WLSMV estimation, and a probit link. Results are presented in Table 3. One item was excluded from the CFA of the SGM-Specific IPV Tactics perpetration subscale due to sparseness.³ All four CFAs were characterized by a single factor and good to excellent factor loadings and model fit, with non-significant chi-squared goodness of fit (ps > .20), RMSEA of .03 to < .001, CFI values of .98 to 1.00, and TLI values of .96 to 1.02.

Construct Validity

Next, we examined associations among types of IPV and between IPV and other variables to test convergent/divergent validity. To allow for comparison of correlations between types of IPV in this and existing studies, we used continuous subscale scores for IPV and Pearson correlations. Given non-normal distributions, we used robust maximum likelihood estimation. We examined the direction and size of associations among variables and tested specific hypotheses about the relative strengths of associations among types of IPV using dependent correlation comparisons. These hypotheses simultaneously assess convergent and divergent validity of each type of IPV by testing whether one type of IPV was more strongly associated with types of IPV that are theoretically or causally linked (e.g., physical IPV and injury) than with types of IPV that are not theorized to be closely associated (e.g., psychological IPV and injury). A significant dependent correlation comparison indicates that a given variable (e.g., psychological IPV) is more strongly associated with one variable (e.g., coercive control) than another (e.g., injury). P values for dependent correlation comparisons are reported below.

Associations among types of IPV.

Correlations among types of IPV (shown in Table 4) were moderate to high (.25 to .78), indicating that, as expected, all types of IPV were positively associated, demonstrating convergent validity. Next, we conducted dependent correlation comparisons. Based on existing research, we hypothesized that physical IPV would be most strongly associated with injury (Lucente et al., 2001; Straus et al., 1996). In the current sample, this was partially supported as physical IPV was more strongly associated with injury than any other IPV type (ps < .001), with the exception of psychological IPV (p = .15). We also expected that coercive control would be most strongly associated with psychological IPV based on prior research and conceptual overlap between psychological IPV and coercive control (Mechanic et al., 2008). Coercive control victimization was more strongly correlated with psychological IPV victimization than with other types of IPV victimization (ps < .05). However, the association between coercive control perpetration and psychological IPV perpetration was not stronger than those between coercive control perpetration and other types of IPV (ps > .10).

Associations between IPV and other constructs.

Next, we examined divergent and convergent validity via associations between types of IPV (subscale scores) and other constructs (Table 4). Consistent with hypotheses, negotiation was moderately correlated with psychological IPV (demonstrating convergent validity) but weakly correlated with all other types of IPV (demonstrating divergent validity). Ineffective couple communication, antisocial behavior, and jealousy demonstrated moderate to strong correlations with all types of IPV, providing evidence of convergent validity. Relationship quality showed weak to moderate, negative correlations with all types of IPV providing evidence of divergent validity. Cohabitating with one’s partner was moderately associated with more psychological and physical IPV victimization and weakly associated with psychological IPV perpetration but, contrary to hypotheses, was not significantly associated with other types of IPV. Being in a more serious relationship was weakly to moderately associated with psychological IPV and coercive control victimization and perpetration but was not significantly associated with other types of IPV.

Next, we examined differences in the strengths of associations between types of IPV and related constructs. Based on existing research, we expected negotiation to be most strongly correlated with psychological IPV (Lucente et al., 2001; Straus et al., 1996), and results supported this hypothesis (ps < .001). Ineffective couple communication and jealousy were more strongly associated with psychological IPV than with other types of IPV, which was expected given overlap in the content of psychological IPV and ineffective couple communication and previous research linking jealousy and psychological IPV (Kerr & Capaldi, 2011). We expected antisocial behavior would be more strongly associated with IPV perpetration than victimization and this was the case for psychological IPV (p < .05) but not for other forms of IPV.

Discussion

Existing measures of IPV, largely developed for heterosexual and cisgender populations (see Edwards et al., 2015), are not appropriate for use with SGM populations (Moradi et al., 2009). They contain content that communicates assumptions that respondents are heterosexual and cisgender, which can alienate SGMs and reduce data quality, and neglect important SGM-specific IPV experiences. When IPV measures have been adapted or created for SGM populations, they have rarely been psychometrically evaluated, leading to concerns about the validity and reliability of these measures. To address these limitations, we created a toolkit of measures that can be used (individually or jointly) to assess a broad range of IPV tactics among SGM populations and psychometrically evaluated these measures.

The new measurement toolkit demonstrated internal reliability and factorial validity. The SGM-CTS2 demonstrated the same factor structure as the CTS2, and the factor loadings for the SGM-CTS2 were similar in magnitude to those obtained in psychometric studies of the CTS2 (e.g., Lucente et al., 2001). These findings indicate that the adaptations made to the CTS2 did not affect the way individual items or subscales function. Therefore, the SGM-CTS2 assesses the same types of IPV as the CTS2 but has the important benefit of being appropriate for use in SGM populations. The Coercive Control Measure and SGM-Specific IPV Tactics Scale each demonstrated unidimensional factor structures and items loaded well on the factors providing support for the internal reliability and factorial validity of these measures.

Study findings were also largely supportive of the validity of the new measurement toolkit. First, the pattern of correlations among the measures of different IPV types is consistent with the existing literature on IPV in heterosexual individuals, providing evidence of convergent validity. Within the SGM-CTS2, there were moderate to strong correlations among the IPV perpetration and among the IPV victimization subscales, which were similar in magnitude to those observed in the original CTS2 (Straus et al., 1996; Lucente et al., 2001). Also consistent with prior research (Straus et al., 1996; Lucente et al., 2001), the strongest correlation was between physical IPV and injury. The new coercive control and SGM-specific IPV scales also demonstrated moderate to strong correlations with other types of IPV. Given the psychological nature of coercive control, we expected it to be more strongly associated with psychological IPV than other types of IPV. This was the case for victimization; however, coercive control perpetration was moderately correlated with all other types of perpetration. This pattern of findings may reflect differences in participants’ willingness to report their own, versus their partner’s, coercive controlling behaviors. Overall, the observed associations match those seen in prior research and the types of IPV with the strongest conceptual links were most strongly correlated, providing evidence of convergent validity. The moderate to strong range of correlations among types of IPV (r = .25-.78; 6.2-60.8% shared variance across type of IPV) also indicates that the types of IPV assessed in this toolkit are distinct from one another.

The pattern of associations between the new IPV scales and other constructs was also supportive of the new measures’ validity. Types of IPV that were theoretically or conceptually associated with one another demonstrated associations in the expected directions, evidence of the new measures’ convergent validity. Negotiation and psychological IPV (which includes minor behaviors such as yelling) both indicate verbal engagement in conflict resolution, and consistent with this conceptual overlap and prior literature (Lucente et al., 2001; Straus et al., 1996), couple conflict negotiation was moderately associated with psychological IPV. Consistent with existing research, there were moderate to strong correlations between all types of IPV and risk factors for IPV identified in heterosexual populations, including ineffective couple communication (Marshall et al., 2011), antisocial behavior (Andrews et al., 2000), and jealousy (Kerr & Capaldi, 2011). The largest associations for ineffective couple communication and jealousy were with psychological IPV, consistent with the psychological nature of each of these constructs.

Results also provided evidence of the divergent/discriminant validity of these new measures of IPV by demonstrating weak associations between IPV and constructs that are theoretically or conceptually unrelated to IPV. Consistent with couple negotiation being theoretically unrelated to IPV (Straus et al., 1996), couple negotiation was only weakly associated with physical IPV, injury, sexual IPV, coercive control, and SGM-specific IPV. Weak associations between low relationship quality and all types of IPV were also present, which is consistent with the literature on IPV (Kerr & Capaldi, 2011; Linder et al., 2002). Together, these findings provide strong evidence of the validity of our IPV measurement toolkit.

Contrary to hypotheses, antisocial behavior and jealousy were not more strongly associated with perpetration than victimization. This is likely due to the high degree of mutual IPV in this sample, with 37-86% of relationships demonstrating mutuality in most types of IPV (Whitton et al., 2019). Additionally, participant- and perceived partner- jealousy were highly correlated (r = .67), limiting their ability to show distinct associations with other variables.

Also contrary to expectations, relationship status and cohabitation were not associated with all types of IPV. In much of the existing literature on IPV, cohabitation and relationship status are examined simultaneously, with dating non-cohabitating relationships compared to dating cohabitating and married cohabitating (see Capaldi, Knoble, Shortt, & Kin, 2012). In the current study, we examined cohabitation and relationship status separately because the number of married relationships was too small to examine as a separate group. Research indicates that dating cohabitation is associated with higher risk of IPV than married cohabitation or dating (see Capaldi et al., 2012), but the limited literature examining cohabitation separately indicates that cohabitation is associated with higher risk of IPV (see Kaukinen, 2014). Similarly, research examining associations between IPV and relationship status is limited and results are mixed with some studies finding that IPV is more prevalent in more committed relationships than more casual relationships and others failing to find associations between IPV and relationship status (see Kaukinen, 2014). Further research is needed to clarify the association between IPV, cohabitation, and relationship status. Given these somewhat mixed literatures, the lack of consistent associations between IPV, cohabitation, and relationship status does not appear to represent evidence against the convergent validity of the newly developed measures.

Results should be considered in light of study limitations. First, though we demonstrated initial evidence of the scales’ reliability and validity, further psychometric evaluation would be valuable (e.g., test-retest reliability, invariance of the measures between SGMs and heterosexuals). Second, the sample included only SGMs assigned female at birth in adolescence or young adulthood. Future research should evaluate these measures in SGMs assigned male at birth and older adults and examine whether the SGM-CTS2 and coercive control measure are culturally appropriate for heterosexual-identified individuals with same-sex attractions, same-sex sexual or romantic partners, or gender minority partners.

Despite these limitations, the measurement toolkit addresses many concerns about the measurement of IPV in SGM samples, including lack of cultural sensitivity, lack of attention to the unique forms of IPV experienced by SGM, and lack of validation in SGM populations. The SGM-CTS2 retains the strong psychometric properties of the CTS2, a commonly used measure of IPV, but is culturally sensitive for use with SGM samples. It will be a valuable tool for appropriately measuring IPV in SGM populations and making meaningful comparisons to rates of IPV in cisgender, heterosexual populations. The new coercive control and SGM-specific IPV scales fill a critical gap in the field, providing validated measures for these types of IPV that are common among SGM (Balsam & Szymanski, 2005; Walters et al., 2013), with significant health impacts (Black et al., 2011). Together, these measures represent a comprehensive toolkit that assesses a range of IPV tactics, is culturally sensitive, and has demonstrated psychometric validity in an SGM sample. As such, we hope it will be used by researchers as well as practitioners to sensitively and comprehensively screen for IPV in this vulnerable population.