Longitudinal Transitions in Intimate Partner Violence among Female Assigned at Birth Sexual and Gender Minority Youth

Abstract

Female-assigned at birth sexual and gender minorities (FAB SGM) are at elevated risk of experiencing intimate partner violence (IPV), yet little longitudinal research has been conducted with this population. In the current study, we attempted to understand how patterns of IPV victimization and perpetration, measured across a wide range of IPV behaviors (psychological, physical, sexual, cyber, and SGM-specific), changed over time for FAB SGM youth. Participants came from a longitudinal cohort study of FAB SGM late adolescents and young adults (FAB400; N = 488) and included anyone who reported a romantic partner at one of the first three waves (six months apart; N = 433). Latent class analysis (LCA) was run at each wave to determine the best-fitting class structure across IPV experiences. These were followed up with latent transition analyses (LTAs) to determine patterns of movement between classes over the course of the three waves. Lastly, we looked at the effects of staying with the same romantic partner on transitions. We found that the same three-class structure replicated across all three waves of the study. That class structure identified no/low, psychological, and high IPV classes at each wave. In the LTA, when transitions occurred for FAB SGM, they were much more likely to transition to a class defined by lower likelihoods of experiencing IPV (i.e., psychological to no/low) than they were to a class defined by higher likelihoods of IPV (i.e., psychological to high). However, we found that FAB SGM youth were less likely to transition to a less severe IPV class if they maintained the same serious romantic partner across waves. This finding, in particular, suggests that IPV is often relationship specific for FAB SGM and that efforts to reduce IPV in SGM communities must consider intervening at the relationship level to effect meaningful change.

Intimate partner violence (IPV) among sexual and gender minority (SGM) youth remains understudied in the broader IPV literature. However, the work that has been done indicates that SGM are at greater risk of experiencing psychological, sexual, and physical IPV in comparison to their cisgender heterosexual counterparts during both adolescence (Dank et al., 2014; Luo et al., 2014; Olsen et al., 2017) and young adulthood (Blosnich & Bossarte, 2012; Porter & Williams, 2011). In contrast to cisgender heterosexual youth, SGM face unique risk factors, such as exposure to anti-SGM stigma (Carvalho et al., 2011; Swann et al., 2019) and less access to resources (Ford et al., 2013), which likely contribute to their differing experiences of IPV. Among SGM, those who are female-assigned at birth (FAB) have reported significantly higher rates of IPV compared to those who are male-assigned at birth (MAB) (Martin-Storey, 2015; Whitton et al., 2019b).

FAB SGM’s elevated risk for IPV is particularly concerning given that they are also more likely to experience more dire consequences of IPV than cisgender, heterosexual women, including injury and post-traumatic stress symptoms (Breiding et al., 2013). There is a clear need for furthering IPV research with FAB SGM, especially because almost no longitudinal research has been done with this population. Work with heterosexual youth, while informative, is limited in what it can tell us about IPV among SGM due to differences in the types of IPV commonly experienced by SGM (Messinger, 2014) and fundamental differences in their romantic relationships created by the pervasiveness of anti-SGM stigma (Mustanski, 2015). In particular, we know little about IPV experiences over time among SGM people, such as the extent to which IPV is stable within individuals over time and across different partners.

The larger body of longitudinal IPV research, conducted with cisgender heterosexual youth, suggests that the stability of IPV experiences over time is greater when youth remain with the same romantic partner (Johnson et al., 2015; Shortt et al., 2012). In a sample of cisgender heterosexual men and women measured across the transition from adolescence to young adulthood, researchers found that IPV experiences were more consistent for those who stayed in the same relationship (Johnson et al., 2015). Similarly, cisgender male youth reported more stability in their perpetration of IPV over the course of 12 years when they had greater relationship continuity (i.e., were with the same partners over longer periods of time) (Shortt et al., 2012). Research with young cisgender heterosexual women has similarly found longer relationships to be associated with a greater likelihood of IPV victimization but that experiencing IPV victimization in one relationship had no association with likelihood of experiencing it in the next (Kennedy et al., 2018). Whether FAB SGM youth similarly report greater IPV stability when they stay in the same partnerships versus make new ones remains to be explored. It could be that the IPV experiences of FAB SGM youth, similar to those of their heterosexual counterparts, differ greatly between different partners (i.e., are partner-specific). Alternately, it is also possible that vulnerability to IPV may operate in a more trait-like manner (i.e., stable across time and across partners) for SGM; the chronic and pervasive societal stigma they face may lead them to form new partnerships with similar propensities toward abuse or limit their ability to shift into healthier relationships.

Only two studies have tracked longitudinal change in IPV with SGM youth and neither considered the effects of relationship continuity. Reuter et al. (2015) measured IPV in a sample of heterosexual and SGM youth at age 15 and again at 17. They found that SGM youth reported higher rates of both IPV perpetration and victimization and more stability in perpetration across their two time points of measurement compared to heterosexual youth. Whitton et al. (2019b) described trajectories of physical and sexual IPV victimization in a sample of SGM youth they followed for five years across six time points spanning the ages of 16–25. They found that 45% of SGM youth experienced physical IPV and 17% experienced sexual IPV at some point during the five-year period. They also found that both forms of IPV victimization remained stable across the transition from adolescence to young adulthood. Though only reported in this single study, the findings of Whitton and colleagues stand in contrast to research with heterosexual samples that have found IPV to peak in late adolescence (Kim et al., 2008; Shortt et al., 2012). Because neither study assessed the effects of changing partners, it is unclear if the observed stability of SGM youth’s IPV experiences over time reflects a trait-like individual-level tendency to form relationships with similar patterns of IPV.

Both previous studies had additional important limitations to understanding change over time in patterns of IPV in SGM youth. Reuter and colleagues only followed their sample across two time points and did not follow up past late adolescence. Whitton and researchers tracked changes in IPV for the longest period of time to date in an SGM youth study, but their measurement of IPV was limited to just physical and sexual IPV, each measured with a single item, and victimization (not perpetration). More thorough measurement across a wider range of IPV domains, is needed; SGM youth have reported other forms of IPV, such as psychological aggression and coercive control, that are integral to understanding the full picture of IPV experiences for this population (Halpern et al., 2004; Whitton et al., 2019a). Further, research with FAB SGM youth, specifically, has highlighted the importance of measuring cyber dating abuse, such as abuse over social media, (Dank et al., 2014) and experiences of IPV that are specific to SGM, like threatening to “out” a partner’s SGM status to others as a form of exerting control (Balsam & Szymanski, 2005; Whitton et al., 2019a). Measuring a wider range of IPV behaviors is necessary to understanding the full spectrum of SGM experiences with IPV and understanding which behaviors co-occur within their romantic relationships. Even the majority of cross-sectional IPV research with SGM has measured only one or two forms of IPV, with the common focus typically being physical IPV (Longobardi & Badenes-Ribera, 2017). No longitudinal study published to date has described IPV experiences over time in SGM that include a wide range of types of IPV (e.g., coercive control, cyber abuse, or SGM-specific IPV experiences).

In the current study, we attempted to understand how patterns of IPV victimization and perpetration, measured across a wider range of IPV behaviors compared to previous longitudinal research with SGM youth, changed over time for FAB SGM youth. Previous cross-sectional research with the FAB SGM cohort has described the baseline rates of physical, sexual, psychological, coercive controlling, cyber-based, and SGM-specific IPV that this cohort has experienced (Whitton et al., 2019a). Their experiences of these IPV indicators were best described by a latent three-class structure in which approximately 50% of FAB SGM fell into a class defined by no IPV or a low likelihood of experiencing most forms of IPV, approximately 35% with a high likelihood of experiencing psychological IPV (including minor and severe psychological aggression and coercive control) but no other IPV types, and approximately 15% who experienced both psychological and physical IPV and were at the highest risk for experiencing every form of IPV measured. Of note, because these classes were derived based on self-report measures of the frequency of certain IPV experiences/incidents, without the surrounding context, they cannot differentiate between behaviors enacted as part of escalated conflict versus intimate terrorism and self-defense, per Johnson’s (2010) IPV typology. Nevertheless, they suggest meaningful common patterns of IPV experienced by FAB SGM, in which coercive control can occur along with other forms of psychological IPV only or, less commonly, along with a host of other IPV experiences including physical aggression and sexual violence.

In the current study, we hypothesized that the class structure previously identified in this cohort that included no/low, psychological only, and high IPV classes would be replicated at two subsequent visits across the following year. We also hypothesized that, in accordance with prior evidence that IPV decreases over the transition to young adulthood in cisgender heterosexual youth, FAB SGM youth would be more likely to transition over time to classes with a lower likelihood of IPV for both IPV perpetration and victimization. Finally, we assessed the effects of remaining in the same partnership over time compared to changing romantic partners between visits on likelihood of transitioning to a different IPV class. We hypothesized that FAB SGM youth with the same partner over time would be less likely to transition to a different class.

Methods

Participants

Participants were from FAB 400, a longitudinal cohort study of the health, development, and intimate relationships of a sample of FAB SGM youth (N = 488). The study consists of two cohorts: (a) a young adult cohort of FAB SGM first recruited in 2007 as part of a SGM youth study called Project Q2 (N = 88; 23–32 years old at the baseline assessment of FAB 400) and (b) a late adolescent cohort recruited in 2016–2017 specifically for FAB 400 (N = 400; 16–20 years old at the baseline assessment). Eligibility criteria for FAB 400 included female sex assigned at birth, ability to speak English, and self-report of same-gender sexual behavior, same-gender attractions, or identity as a sexual or gender minority. Recruitment was accomplished through social media advertisements and directly from venues (i.e., SGM community organizations, health fairs). It also incorporated a snowball sampling approach where participants could refer up to five of their peers into the study and receive $10 for every successfully recruited peer.

All participants completed their first visit in 2016–2017 with follow-up visits every six months. Retention was high for the three waves in the current study (W2: 96.7%, W3: 94.9%). Participants were paid $50 at each visit to answer a series of self-report measures administered via a computer-assisted self-interview. Participants in this study gave written informed consent prior to participation and systems to protect their confidentiality were utilized (i.e., a federal certificate of confidentiality). All study protocols were approved by the Institutional Review Board at Northwestern University including a waiver of parental permission for all participants below the age of 18 under 45 CFR 46, 408(c). The present study was focused on IPV in the context of romantic relationships, so only FAB SGM who reported a serious romantic partner during at least one of the first three waves were included in analyses (analytic N = 433).

The average age of participants in the analytic sample at baseline was 20.23 (SD = 3.80). The largest racial/ethnic group was Black/African American (N = 156, 36.0%), followed by Hispanic/Latinx (N = 110, 25.4%), Caucasian (N = 108, 24.9%), multi-racial (N = 36, 8.3%), Asian/Pacific Islander (N = 18, 4.2%), and participants who selected “other” (N = 5, 1.2%). At the baseline visit, 38.1% identified as bisexual (N = 165), 21.5% as lesbian (N = 93), 17.3% as pansexual (N = 75), 11.8% as queer (N = 51), 3.7% as unsure/questioning (N = 16), 3.0% as gay (N = 13), 1.8% as asexual (N = 8), 1.4% as straight/heterosexual (N = 6), and 1.4% said their orientation was not listed (N = 6). The majority of participants reported their gender identity as female at baseline (N = 327, 75.5%), 6.0% identified as gender queer (N = 26), 5.8% as gender non-conforming (N = 25), 4.4% as male (N = 19), 3.9% as transgender (N = 17), 3.0% as non-binary (N = 13), and 1.4% said their gender identity was not listed (N = 6).

Measures

IPV measurement.

Prior to the administration of the IPV measures included in the current study, participants were asked to identify the first name and last initial of their most serious dating or sexual partner in the previous six months. All subsequent questions about IPV were asked in reference to this partner at each wave. For each question asked about IPV, participants were given the response options “never” (0), “once” (1), “twice” (2), “3–5 times” (3), “6–10 times” (4), “11–20 times” (5), and “more than 20 times” (6). Because we were interested in the occurrence of each form of IPV, participants were coded as “1” if they endorsed any option above 0 on an item in that subscale, and “0” if they did not experience that type of IPV in the previous six months.

Psychological, physical, and sexual IPV.

Experiences of IPV were measured using the Sexual and Gender Minority Conflict Tactics Scale (SGM-CTS2) (Dyar et al., 2019). The SGM-CTS2 was modified from the CTS2 (Straus et al., 1996) to be culturally appropriate for measurement of SGM IPV experiences. The SGM-CTS2 has previously been found to match the CTS2, in factor structure, validity, and reliability, using this sample at baseline (Dyar et al., 2019). The current study used six IPV subscales from the SGM-CTS2: minor psychological, severe psychological, minor physical, severe physical, injury, and sexual. Questions were repeated for both perpetration and victimization (e.g., “I slapped [partner name]” and “[partner name] slapped me.”) and all items were asked in the context of the previous six months.

SGM-specific IPV.

The SGM-Specific IPV Tactics Scale was used to measure SGM-specific experiences of IPV (Dyar et al., 2019). The scale includes IPV experiences that only SGM can experience like outing (“I threatened to out [partner name] to their friends, family, or other people in their life if they didn’t do what I wanted”) and forced displays of affection (“[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.”). Five items were asked for both perpetration and victimization. Previous baseline analysis with this sample demonstrated high validity and internal consistency for this scale (Dyar et al., 2019).

Cyber abuse.

Participants were administered the Cyber Abuse Scale, which consisted of four items for perpetration and four for victimization. It was based on the Cyber Dating Abuse measure (Zweig et al., 2013). The scale measures IPV unique to the online sphere (e.g., “I wrote mean or embarrassing things about [partner name] on social media.”). Previous research found high validity and internal consistency for the scale with SGM youth (Zweig et al., 2013).

Coercive control.

Participants answered the Coercive Control Scale, which drew three items from the 2010 National Intimate Partner and Sexual Violence Survey (Breiding et al., 2013) and five items from the Coercive Behaviors Scale (Frankland & Brown, 2014) with adaptions made to better suit SGM culture (Dyar et al., 2019). The validity and internal consistency for this scale has been described for this sample at baseline (Dyar et al., 2019). The scale measured controlling behaviors within the relationship (e.g., “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”).

Repeat partnerships.

Participants were asked to identify if their most serious partner matched a partner they had reported at a previous visit beginning at wave 2. That information was used to determine if their most serious partner was the same person across waves or someone new. We coded two variables, one identifying if the reported partners were the same for waves 1 and 2, and one identifying if the reported partners were the same for waves 2 and 3 for use within our models. Each variable was coded “1” if the partners matched and “0” if they did not.

Analytic Plan

In previous analyses of baseline FAB 400 data (n = 352; only included participants who reported a serious romantic relationship at that first wave), LCA was used to identify a three-class solution that best described patterns of past 6 month IPV perpetration and IPV victimization across nine IPV indicators: (a) minor psychological, (b) severe psychological, (c) coercive control, (d) minor physical, (e) severe physical, (f) injury, (g) sexual, (h) SGM-specific, and (i) cyber (see: Whitton et al., 2019a). For both perpetration and victimization, the three classes could best be described as a (a) no/low IPV class, (b) psychological IPV only class, and a (c) high IPV class (Figure 1). In that previous study and the current, we did not give focal consideration to coercive control because we did not ask about the surrounding context that would be necessary to determine if coercive control was enacted as part of an escalating conflict versus a pattern of intimate terrorism.

Figure 1.

Latent classes of intimate partner violence (IPV) for waves 1–3. Psych = psychological.

To test our first hypothesis that the same class structure identified at baseline would provide the best solution at waves 2 and 3, we ran separate preliminary LCAs at each of those time points using MPlus. Participants were included in the LCA models if they reported having had a romantic relationship with their most significant dating or sexual partner in the previous six months at that wave (W2 N = 338; W3 N = 338). The best-fitting class solution for each new wave (i.e., W2 and W3) was determined using Bayesian Information Criterion (BIC) and sample size adjusted-Bayesian Information Criterion (Adj-BIC), where lower values indicate better fit, and Lo–Mendell–Rubin (LMR) likelihood ratio tests and parametric bootstrapped likelihood ratio tests (BLRTs), where significant tests indicate a superior fit compared to a model with one less class.

In order to test our second hypothesis that FAB SGM youth would transition to latent classes with a lower likelihood of IPV over time, we followed up the LCA analyses using Asparouhov and Muthén (2014)’s three-step approach for LTA using all three waves of data. We included all participants who reported a romantic partner during at least one wave (N = 433). For waves where a participant did not report a romantic partner, their IPV indicators were coded as missing. MPlus addresses missingness using full information maximum likelihood (FIML), which uses all available data to estimate model parameters. The FIML approach to address missingness has been shown to still perform well even when missingness is not random (Buhi et al., 2008; Newman, 2003). The first step was to determine if the classes identified in the preliminary LCAs were invariant over time by first running an unconstrained model that simultaneously estimated the latent classes for all three waves within one model based on the best-fitting class solution from the preliminary LCA models. We compared this to a constrained model where the classes at each wave were constrained to be equal across waves. The best-fitting of the two models was determined by comparing the BIC and Adj-BIC values. The second step was to obtain most likely class membership and likelihood of misclassification at each wave by running a separate LCA model for each wave where parameters were constrained to be equal with the parameter estimates from the best-fitting of the two models run at the first step (i.e., either the constrained or unconstrained model). The third step was to take the most likely classifications from those separate wave LCA models and the likelihood for misclassification and enter them into a single LTA model that estimated the likelihood of transitioning between classes across the three waves. As part of this step, we compared a stationary model, in which the transition likelihoods from wave 1 to wave 2 were constrained to be equal to the transition likelihoods from wave 2 to wave 3, to a model in which those transitions could differ.

Finally, to answer our last research question about the effects of relationship continuity on transitions, after determining the best model, we tested whether reporting the same partner across waves was associated with differences in likelihood of transitioning from one class to another. Participants had to have completed all three waves to be included in this analysis (n = 409). Of those participants, 47.9% (N = 196) reported the same partner at the first two waves and 49.6% (N = 203) reported the same partner at waves 2 and 3. Having a repeat partner was entered as a predictor of transitions in the LTA model from step 3, in which transitions were allowed to differ across time (i.e., the model without stationary transitions).

Results

Preliminary Latent Class Analyses

Model fit for the initial LCA models is presented in Table 1. Fit statistics for the wave 1 models come from Whitton et al. (2019a) where the three-class solution provided the best fit for both IPV perpetration and IPV victimization. A similar three-class solution provided the best fit for all three waves for both perpetration and victimization based on BIC, Adj-BIC, LMR, and BLRT. The three-class structure provided similar classes at every wave (Figure 1). The first class we called the no/low IPV class because 35%–48% of participants in this class reported minor psychological IPV but did not report that any of the other IPV outcomes had occurred. This was the largest class at each wave (perpetration: W1 53.4%, W2 60.4%, W3 74.0%; victimization: W1 52.5%, W2 56.5%, W3 68.2%). The second class was defined by a higher likelihood of reporting minor psychological (91%–100%), severe psychological (40%–84%), and coercive control (41%–58%) IPV, but a low likelihood of the other six indicators. This was the second largest class at each wave (perpetration: W1 36.0%, W2 32.5%, W3 21.6%; victimization: W1 32.1%, W2 37.0%, W3 27.0%). We called this group the psychological IPV class. The final class reported the highest rates on all nine indicators, and we called it the high IPV class. This was the smallest class at each wave (perpetration: W1 10.5%, W2 7.1%, W3 4.4%; victimization: W1 15.3%, W2 6.5%, W3 4.7%). The classes identified at all three waves were consistent with the classes identified at the first wave by Whitton et al. (2019a). Because the class structure and the interpretation of the classes stayed consistent across all three waves, we included three classes at each wave in all subsequent LTAs.

Table 1.

Model Fit for Latent Class Analysis.

		Perpetration					Victimization
Wave	# Classes	BIC	Adj-BIC	LMR	BLRT	Entropy	BIC	Adj-BIC	LMR	BLRT	Entropy
1	1	2709.08	2680.53	–	–	–	3050.95	3022.39	–	–	–
	2	2359.90	2299.63	<.001	<.001	0.79	2639.11	2578.84	<.001	<.001	0.88
	3	2341.23	2249.23	.05	<.001	0.78	2631.71	2539.71	<.01	<.001	0.73
	4	2378.79	2255.07	.13	.09	0.79	2665.97	2542.24	.70	.18	0.78
2	1	2382.47	2353.92	–	–	–	2689.20	2660.65	–	–	–
	2	2033.25	1972.98	<.001	<.001	0.85	2289.63	2229.36	<.001	<.001	0.86
	3	1999.07	1907.08	<.001	<.001	0.83	2253.79	2161.80	<.01	<.001	0.84
	4	2032.39	1908.67	.57	<.001	0.85	2285.95	2162.24	<.01	.03	0.86
3	1	2134.37	2105.82	–	–	–	2495.50	2466.95	–	–	–
	2	1905.97	1845.70	<.001	<.001	0.83	2141.41	2081.14	<.001	<.001	0.89
	3	1907.51	1815.52	<.01	<.001	0.86	2117.71	2025.71	<.01	<.001	0.84
	4	1942.37	1818.66	.14	.03	0.85	2154.79	2031.07	.25	.06	0.82

Note. BIC = Bayesian Information Criterion; Adj-BIC = adjusted BIC; LMR = Lo–Mendell–Rubin likelihood ratio test; BLRT = parametric bootstrapped likelihood ratio test. p-Values reported for LMR and BLRT. Bold face represents best-fitting model.

Latent Transition Analysis

For IPV perpetration, we compared the constrained and unconstrained models from the first step of Asparouhov and Muthén (2014)’s three-step approach. The superior fit of the constrained (BIC = 5987.36, Adj-BIC = 5882.64) model compared to the unconstrained (BIC = 6268.19, Adj-BIC = 5992.10) model indicated that the classes were invariant and could be interpreted the same across waves. We ran separate LCAs within each wave for step 2 based on the constrained model and incorporated the step 2 classifications into our step 3 LTAs. As the final part of that third step, we compared the LTA model where transitions were stationary to one where they were allowed to vary. The stationary model provided a better fit (BIC = 1570.91, Adj-BIC = 1545.52) compared to the non-stationary model (BIC = 1599.05, Adj-BIC = 1554.62). The results suggest that for IPV perpetration, patterns of movement between classes stayed consistent across the two transitions (i.e., from W1 to W2 and from W2 to W3).

The class transitions for IPV perpetration are presented in Table 2. The no/low IPV class was the most consistent across waves, with 92.7% of those in the no/low IPV class at one wave remaining in that class at the next wave. The psychological IPV and high IPV classes were less stable over time by comparison, with 59.5% of those in the psychological IPV class remaining in that class at the next wave and 47.5% of those in the high IPV class remaining in that class at the next wave. Participants who transitioned to different classes were more likely to transition to a less severe class than to a more severe class. Of those who started in the psychological IPV class, 35.7% transitioned to a less severe class (i.e., no/low IPV), and in the high IPV class, 52.5% transitioned to a less severe class (i.e., no/low or psychological IPV). Only 7.3% of those who started in the no/low IPV class and 4.8% of those who started in the psychological IPV class transitioned to a more severe class.

Table 2.

IPV Perpetration and IPV Victimization Latent Transitions.

		Perpetration			Victimization
		Wave 2			Wave 2
		Class 1: Low IPV	Class 2: Psych IPV	Class 3: High IPV	Class 1: Low IPV	Class 2: Psych IPV	Class 3: High IPV
Wave 1	Class 1: Low IPV	92.7%	6.0%	1.3%	90.2%	9.8%	0.0%
	Class 2: Psych IPV	35.7%	59.5%	4.8%	32.6%	63.8%	3.7%
	Class 3: High IPV	17.4%	35.1%	47.5%	22.7%	28.8%	48.5%
		Wave 3			Wave 3
		Class 1: Low IPV	Class 2: Psych IPV	Class 3: High IPV	Class 1: Low IPV	Class 2: Psych IPV	Class 3: High IPV
Wave 2	Class 1: Low IPV	92.7%	6.0%	1.3%	90.2%	9.8%	0.0%
	Class 2: Psych IPV	35.7%	59.5%	4.8%	32.6%	63.8%	3.7%
	Class 3: High IPV	17.4%	35.1%	47.5%	22.7%	28.8%	48.5%

Results for IPV victimization followed a similar pattern to perpetration. The constrained model (BIC = 6748.85, Adj-BIC = 6644.12) provided a better fit compared to the unconstrained model (BIC = 7025.25, Adj-BIC = 6749.16), which indicated that the IPV victimization classes were invariant across waves. We also found that the model where transitions were treated as stationary (BIC = 1662.62, Adj-BIC = 1637.24) fit better than the model where transitions between waves were allowed to vary (BIC = 1690.35, Adj-BIC = 1645.92), indicating that the pattern of change from W1 to W2 was consistent with the pattern of change from W2 to W3.

The transition pattern for IPV victimization was also similar to perpetration (Table 2). There was high stability for the no/low IPV class (90.2%), but lower stability for the psychological IPV (63.8%) and high IPV (48.5%) classes. Overall, participants were more likely to transition to less severe classes (32.6% in the psychological IPV and 51.5% in the high IPV class at the previous wave) and less likely to transition to more severe classes (9.8% of people who were no/low IPV and 3.7% of people who were in the psychological IPV class at the previous wave).

Repeat Partnerships on Transitions

The effects of predictors on transitions cannot be accurately estimated when transitions are constrained to be stationary (Nylund, 2007), so we added the effects of repeat partnerships to the models where transitions were allowed to vary between waves despite the superior fit of the stationary models. Table 3 shows the most likely final transition classifications for participants broken out by whether the participant reported a new partner or a repeat partner from the previous visit. Results of models testing whether the likelihood of transition differed significantly between those with a new versus repeat partner are presented in Table 4. For IPV perpetration, we found that 55.0% of individuals in the psychological IPV class at W1 who were in a new relationship at W2 transitioned to the no/low IPV class, compared to only 26.1% of those who had the same partner at W1 and W2. We found that this difference in the likelihood of transitioning from the psychological IPV to the no/low IPV class from W1 to W2 was significant, such that people with a repeat partner were significantly less likely to make the transition (OR = 0.22, p < .01). We saw a similar effect from waves 2 to 3; 67.3% of individuals in the psychological IPV class at W2 who were in a new relationship at W3 transitioned to the no/low IPV class compared to 24.1% of those with a repeat partner. The difference between individuals with a repeat partner versus a new partner was significant across this transition (OR = 0.21, p <.05). Reporting the same partner at multiple waves was not significantly associated with any of the other transitions for IPV perpetration.

Table 3.

Most Likely Transition Classifications by Repeat Partnership.

		IPV Perpetration
		New Partner			Repeat Partner
		Wave 2			Wave 2
		Low IPV	Psych IPV	High IPV	Low IPV	Psych IPV	High IPV
		N (%)	N (%)	N (%)	N (%)	N (%)	N (%)
Wave 1	Low IPV	132 (93.6)	9 (6.4)	0 (0.0)	90 (86.5)	12 (11.5)	2 (1.9)
	Psych IPV	33 (55.0)	25 (41.7)	2 (3.3)	18 (26.1)	47 (68.1)	4 (5.8)
	High IPV	0 (0.0)	5 (41.7)	7 (58.3)	3 (13.0)	9 (39.1)	11 (47.8)
		Wave 3			Wave 3
		Low IPV	Psych IPV	High IPV	Low IPV	Psych IPV	High IPV
Wave 2	Low IPV	134 (93.7)	7 (4.9)	2 (1.4)	132 (99.2)	0 (0.0)	1 (0.8)
	Psych IPV	33 (67.3)	14 (28.6)	2 (4.1)	14 (24.1)	42 (72.4)	2 (3.4)
	High IPV	10 (71.4)	1 (7.1)	3 (21.4)	0 (0.0)	8 (66.7)	4 (33.3)
		IPV Victimization
		New Partner			Repeat Partner
		Wave 2			Wave 2
		Low IPV	Psych IPV	High IPV	Low IPV	Psych IPV	High IPV
Wave 1	Low IPV	128 (95.5)	6 (4.5)	0 (0.0)	81 (76.4)	25 (23.6)	0 (0.0)
	Psych IPV	29 (49.2)	29 (49.2)	1 (1.7)	17 (27.9)	40 (65.6)	4 (6.6)
	High IPV	9 (45.0)	8 (40.0)	3 (15.0)	2 (6.9)	10 (34.5)	17 (58.6)
		Wave 3			Wave 3
		Low IPV	Psych IPV	High IPV	Low IPV	Psych IPV	High IPV
Wave 2	Low IPV	141 (100.0)	0 (0.0)	0 (0.0)	125 (100.0)	0 (0.0)	0 (0.0)
	Psych IPV	33 (60.0)	16 (29.1)	6 (10.9)	10 (15.9)	53 (84.1)	0 (0.0)
	High IPV	6 (60.0)	3 (30.0)	1 (10.0)	1 (6.7)	2 (13.3)	12 (80.0)

Table 4.

Effect of Repeat Partnerships on IPV Class Transitions.

	IPV Perpetration		IPV Victimization
Transitions	Repeat Partner (W1–W2)	Repeat Partner (W2–W3)	Repeat Partner (W1–W2)	Repeat Partner (W2–W3)
	Odds Ratios		Odds Ratios
Low IPV (W1) to
Psych IPV (W2)	1.09	–	4.18	–
High IPV (W2)	–1	–	–1	–
Psych IPV (W1) to
Low IPV (W2)	0.22**	–	0.37	–
High IPV (W2)	1.04	–	2.52	–
High IPV (W1) to
Low IPV (W2)	–1	–	0.06*	–
Psych IPV (W2)	0.79	–	0.26	–
Low IPV (W2) to
Psych IPV (W3)	–	0.29	–	0.67
High IPV (W3)	–	0.19	–	–1
Psych IPV (W2) to
Low IPV (W3)	–	0.21*	–	0.12**
High IPV (W3)	–	0.46	–	–1
High IPV (W2) to
Low IPV (W3)	–	0.02	–	0.03*
Psych IPV (W3)	–	3.12	–	0.11

Note. All transitions are in comparison to staying in the same class.

¹Indicates parameter could not be calculated due to empty cells. W = wave.

^*p < .05.

^**p < .01.

For transitions between the IPV victimization classes, 45.0% of individuals in the high IPV class at W1 who reported a new partner at W2 transitioned to the no/low IPV class compared to 6.9% of those with the same partner (OR = 0.06, p < .05). From the second to third waves, 60.0% of individuals in the Psychological IPV class at W2 with a new partner at W3 transitioned to the no/low IPV class compared to 15.9% of individuals with the same partner (OR = 0.12, p < .01). For participants in the high IPV class at W2, 60.0% of those with a new partner at W3 transitioned to the no/low IPV class compared to 6.7% of those with a repeat partner (OR = 0.03, p < .05). Having a repeat partner at multiple waves was not significantly associated with any other transitions between the IPV victimization classes. However, the large but non-significant odds ratios for the transition from no/low to psychological IPV (OR = 4.18) and from psychological to high IPV (OR = 2.52) classes for FAB SGM with repeat partners may indicate that there were meaningful transition differences that we were underpowered to significantly detect or could be a product of measurement error.

Discussion

The purpose of the present study was to explore patterns of IPV experiences among FAB SGM youth over time. Expanding beyond the very limited previous longitudinal research on SGM IPV, we used multi-item measures assessing a wide array of IPV experiences to assess for change versus stability in the patterns of IPV experiences that tend to co-occur (i.e., latent classes of IPV) across three waves of data spanning 18 months. Then, we tested whether these patterns of IPV experiences changed within individuals over time and across partners. The first key finding was that the three-class structure of IPV experiences previously identified at baseline (Whitton et al., 2019a), replicated across the second and third waves of the study. Specifically, participants’ experiences of IPV tended to fall into three classes: no/low IPV, psychological IPV only, and high IPV (experiencing multiple forms of IPV, including psychological, physical, and sexual) at each wave across the year of measurement. This suggests that over time, there is consistency in the patterns of IPV that FAB SGM youth commonly experience. Further, the same pattern was present for experiences of both IPV perpetration and victimization, likely speaking to the dyadic nature of IPV. Because the majority of IPV reported by FAB SGM is bi-directional (i.e., perpetrated by both partners; Whitton et al., 2019a), it makes sense that patterns observed in types of IPV that co-occur are similar for perpetration and victimization. However, because the present study addressed the occurrence of different IPV types, not their severity, so finding that IPV was bi-directional does not necessarily mean that it was equal in chronicity or severity between partners.

Modeling transitions between the three IPV classes across the three waves of measurement revealed that the majority of participants remained in the same class across waves. This general stability in IPV is consistent with previous research indicating physical and sexual IPV victimization remained stable over time in SGM youth (Whitton et al., 2019b), and extends those findings to indicate stability over time in a wider range of IPV types and in perpetration, not just victimization. For many young FAB SGM, particular patterns of IPV experiences appear to persist over time, at least over the 18 months assessed in this study. However, the degree of temporal stability differed between the latent classes of IPV. For both perpetration and victimization, the no/low IPV class was the most stable over time; over 90% remained in this class at the following wave. The more severe IPV classes were somewhat more volatile, with around 60% of those in the psychological IPV class and 48% of those in the high IPV class remaining in the same class over time.

While the majority of FAB SGM remained in the same class from wave to wave, a significant minority moved at each transition point, especially those within the riskier classes. In accordance with our hypothesis, when transitions occurred for our youth, they were much more likely to transition to a class defined by lower likelihoods of experiencing IPV than they were to a class defined by higher likelihoods of IPV. This pattern of findings may indicate that FAB SGM either learn to navigate out of higher IPV relationships or develop healthier conflict resolution skills within their relationships over time. It also may imply a trend toward a decrease over time in IPV experiences for FAB SGM across the transition to young adulthood, as has been observed in cisgender heterosexual youth (Kim et al., 2008; Shortt et al., 2012). However, unlike the previous studies, which focused on change in IPV experiences by age, the current study assessed change across visits for a sample spread out in age. That limits what we can meaningfully say about developmental change in IPV and highlights the importance of additional longitudinal research on the IPV experiences of FAB SGM.

Particularly of note, youth in the psychological IPV class transitioned to the high IPV class less than 5% of the time compared to over 30% that transitioned to the no/low IPV class between each wave for both victimization and perpetration. These findings suggest that for most FAB SGM youth, experiences of psychological IPV do not act as a precursor to escalation toward more serious IPV experiences. However, one of the strengths of taking a latent transition approach to observe change over time in IPV is that we can observe that, even though transitioning to a less severe IPV class was more common, there were a minority of FAB SGM who transitioned to more severe classes between waves. For these participants, there appeared to be a clear pattern to their movement in which those in the no/low IPV class would transition to the psychological IPV class prior to transitioning to the high IPV class. For instance, of FAB SGM in the no/low IPV class for victimization, 9.8% transitioned to the psychological IPV class but none transitioned directly to high IPV. This could indicate a unique prevention opportunity. If psychological IPV is a precursor for a subset of SGM to experiencing multiple forms of IPV within a relationship (as defined by the high IPV class), then targeting SGM who have perpetrated or been the victim of psychological IPV might be an effective strategy for reducing the prevalence of physical and sexual IPV. Conversely, resources may be saved by shifting them away from SGM who have not experienced any IPV, including psychological aggression, knowing that they are the most unlikely to experience physical or sexual IPV in the subsequent six-month period.

Our final hypothesis was that FAB SGM who retained the same romantic partner would show greater stability in their IPV experiences than those who changed partners. Consistent with this hypothesis and past research with cisgender heterosexual youth (Johnson et al., 2015; Kennedy et al., 2018), we found that FAB SGM youth were less likely to transition to a less severe IPV class if they maintained the same serious romantic partner across visits. Specifically, if they reported a consistent partner across waves, participants were less likely to transition to the no/low IPV class from the psychological IPV victimization class, the psychological perpetration class, or the high IPV victimization class. The higher consistency of IPV experiences seen in the same romantic relationships than across different relationships suggests that IPV among FAB SGM youth may be more influenced by relationship dynamics within partnerships than by static individual factors. This is consistent with research with heterosexual samples that has found dyadic processes between two specific partners to be central to how patterns of IPV develop (Capaldi & Langhinrichsen-Rohling, 2012; Langhinrichsen-Rohling & Capaldi, 2012). Together with the striking similarities between IPV perpetration and victimization in the latent classes and transitions, and the bi-directional nature of most IPV reported by FAB SGM youth (Whitton et al., 2019a), these findings suggest that interventions aimed at reducing situational couple violence at dyadic level (e.g., interventions aimed at couples) or assisting youth in extricating themselves from relationships where they have perpetrated and/or been the victim of IPV may be the most fruitful in reducing IPV rates for FAB SGM. As such, they are supportive of recent efforts to develop relationship education programs to teach healthy conflict management skills to SGM couples (Newcomb et al., 2017; Whitton et al., 2016, 2017) and healthy relationship skills to SGM youth (Mustanski et al., 2015), which may ultimately help reduce IPV experiences in these groups.

It is also noteworthy that we did not see significant evidence that remaining with the same partner was associated with escalation of IPV over time. This suggests that relationships defined by no IPV or only minor psychological IPV are likely to stay that way and relationships where psychological IPV is present are not more likely to escalate to more serious forms of IPV, at least within the period of time observed in this study. However, we observed large odds ratios for some transitions that did not reach significance, which may imply that we were underpowered to identify some of these patterns. In particular, in the transition from the first to second waves for IPV victimization, we observed that FAB SGM with a repeat partner were four times more likely to transition from the no/low IPV to the psychological IPV class and were two and half times more likely to transition from the psychological IPV class to the high IPV class than those with new partners. If these odds ratios reflect meaningful differences, they would imply that staying with the same partner is associated with a higher risk of escalating IPV victimization experiences. Future analyses with a larger sample of FAB SGM youth, with longer longitudinal follow-up, should be conducted to determine if these are meaningful differences between youth who maintain the same romantic partner over time compared to youth who change partners, or if the large, non-significant odds ratios found in this study are a product of measurement error. Similarly, the small size of the high IPV class, which grew smaller at each wave, may have meant we were underpowered to detect some transitions in and out of that class.

There were further important limitations to the current study, in addition to the small size of the high IPV class. Despite measuring both victimization and perpetration of IPV, we could not achieve convergence for latent class models that included both at the same time. This was likely due to the strong correlations between victimization and perpetration of each IPV type; across the three waves, 64.4% of participants who experienced a given type of IPV reported being both a victim and a perpetrator of it. Replicating these analyses with larger population-based samples, even if they have to sacrifice the depth of measurement in the current study, may provide the power necessary to assess for patterns of change unique to bi-directional IPV experiences, in contrast to those perpetrated only by one partner on the other. In addition, we used the CTS-2, which has been criticized for overestimating the parity in IPV between partners compared to other measures of partner violence (Hamby, 2014). Future research is needed that captures the context of control in which the IPV incidents occur, allowing for differentiation between bi-directional IPV that actually represents intimate terrorism by one partner and self-defense by the other from situational couple violence that involves aggression by both partners (e.g., Badenes-Ribera et al., 2016). Finally, the current study encompassed only three data points measured over the course of a yearlong period (i.e., an 18-month reporting window). Together with the wide age range of our sample, this meant that we were unable to explore developmental trajectories of change in IPV for FAB SGM. As more waves of data for the FAB 400 project are collected over the course of additional years, we will begin to explore these questions.

Now that we have begun to understand the latent IPV classes that exist for FAB SGM and their consistency over time, we should also expand the predictors we use to improve our understanding of when and why transitions between classes occur. This should include demographic differences and aspects of relationship functioning that could explain why transitions happen, but that the present study was underpowered to explore within a latent transition framework. In the existing cisgender heterosexual youth literature, relationship skills, perceptions of infidelity, relationship stability, and other measures of relationship functioning have been associated with the occurrence of IPV (Giordano et al., 2015; Halpern-Meekin et al., 2013; Johnson et al., 2015). Researchers should explore if the relationship functioning of FAB SGM youth similarly influence their experiences of IPV over time. Researchers should also focus on risk factors unique to SGM youth such as experiences of discrimination based on their SGM status and feelings of internalized stigma that could add additional stress to relationship dynamics. Identifying the factors that predict how patterns of IPV change for SGM can inform how prevention and intervention efforts are targeted and the design of their content. Finally, researchers should replicate the present findings with MAB SGM youth or samples that include both FAB and MAB SGM, to determine if MAB SGM experience similar IPV classes and transitions as their FAB counterparts.

The dearth of existing literature exploring longitudinal change in how SGM youth experience IPV has left significant gaps in our understanding of violence within SGM intimate relationships. This study sought to help remedy this by documenting common patterns of IPV experiences for FAB SGM across a large battery of IPV measures and the ways those experiences change over time. We identified encouraging patterns that suggest declines in both IPV perpetration and victimization over time for FAB SGM youth. We also found that these declines were more pronounced for FAB SGM who changed romantic partners between visits. This finding, in particular, suggests that IPV is often relationship specific for FAB SGM and that efforts to reduce IPV in SGM communities must consider the relationship dyad to affect meaningful change. This study begins to answer what longitudinal change in IPV looks like for FAB SGM, but many important questions still persist, and the continuation of this work remains crucial.

Biography

Gregory Swann, MS, MA, is a senior data analyst for Institute of Sexual and Gender Minority Health and Wellbeing at Northwestern University. His previous research work has been in quantitative and molecular behavioral genetics. More recently his work has focused on developmental change and health disparities in LGBTQ populations as they transition from adolescence into adulthood.

Christina Dyar, PhD, is a research assistant professor at the Institute for Sexual and Gender Minority Health and Wellbeing at Northwestern University. Dr Dyar’s research broadly focuses on understanding and reducing health disparities affecting sexual and gender minority populations, especially sexual minority women and bisexual individuals. She is particularly interested in understanding mechanisms through which different types of stress (e.g., discrimination, internalized stigma).

Michael E. Newcomb, PhD, is an assistant professor in the Department of Medical Social Sciences at Northwestern University and the associate director for Scientific Development of the Northwestern Institute for Sexual and Gender Minority Health and Wellbeing (ISGMH). Dr Newcomb’s research broadly focuses on health disparities in LGBT youth, particularly in the areas of HIV/AIDS, alcohol and drug use, and mental health problems. His work emphasizes the interpersonal contexts that influence health outcomes, including romantic relationships and families.

Sarah W. Whitton, PhD, is an associate professor in the Department of Psychology at the University of Cincinnati. Dr Whitton’s research aims to better understand modern couples and families and to help them build and maintain the types of strong, stable relationships that promote health and well-being. She is particularly interested in understudied and marginalized groups, including sexual and gender minorities. Dr Whitton conducts research to identify factors that promote strong relationships in the face of adversity.