A national epidemiologic profile of physical intimate partner violence, adverse childhood experiences, and supportive childhood relationships: Group differences in predicted trends and associations

Abstract

Introduction

Adverse childhood experiences (ACEs) are common in the US and associated with multiple health sequelae. Physical intimate partner violence (IPV) is a type of revictimization that some adults with ACEs may be more prone to. Positive and supportive childhood environments may buffer the effects of ACEs, but little is known about the differential associations between physical IPV and ACEs and supportive childhood environments. We sought to illustrate racial/ethnic and gender differences in the adjusted predicted probability of physical IPV across multiple ACE and supportive childhood scores.

Methods

We used multivariate linear regression to model the predicted probability of experiencing physical IPV across ACE (physical, psychological, sexual, household environment, mother’s abuse) and supportive childhood scores in a national sample (National Epidemiologic Survey on Alcohol and Related Conditions-III, 2012–2013, n=35,614). Data analyses were conducted in 2019.

Results

American Indian/Alaska Native (AI/AN) women demonstrated the highest proportion of experiencing physical IPV (21%). AI/AN men had the highest mean physical ACE score (1.6/4), while AI/AN women had the highest mean scores for all other ACE typologies. ACE scores were positively associated with predicted physical IPV among women, and among AI/AN women in particular. Supportive childhood scores were negatively associated with predicted physical IPV primarily among women.

Conclusion

Physical IPV and ACEs are exceedingly high among AI/AN women. A better understanding of differential associations between childhood experiences and IPV is needed to more effectively tailor childhood and family-based health promotion strategies among multiple diverse communities.

Introduction

Physical violence perpetrated by an intimate partner is a public health and human rights problem in the United States. The vast majority of those exposed to physical intimate partner violence (IPV) are women [1, 2]. The prevalence of multiple IPV typologies (physical violence, sexual violence, threat of physical or sexual abuse, psychological or emotional abuse, and stalking) varies widely by race/ethnicity and gender. According to the Bureau of Justice Statistics (BJS), American Indian/Alaska Native (AI/AN) women reported the highest rate of any IPV (18.2 per 1,000), followed by Black women (8.2), White women (6.3), Latina women (6.0), and AI/AN men (5.0) in 2003 [3]. A more recent BJS report demonstrated similar demographic trends, but failed to include AI/AN as a standalone group [4]. Analyses of racial/ethnic differences using the National Intimate Partner and Sexual Violence Survey are in line with the BJS reports [2]. Considering only physical IPV, AI/AN women demonstrated the highest prevalence of lifetime physical violence (51.7%) perpetrated by an intimate partner, while the lowest proportion was observed among Asian/Pacific Islander men (15.3%) [2]. While prevalence differences in IPV are clear, reports on differential associations between risks and outcomes by race/ethnicity and gender are scant.

Adverse childhood experiences (ACEs) include many types of abuse, neglect, and trauma among people under 18 years of age [5], some of which are positively associated with subsequent IPV in adulthood [6, 7]. Surveys of ACEs vary in terms of the qualitative types of adverse experiences assessed, but conventional categories include psychological, physical, and sexual abuse, witnessing violence against a child’s mother, or living with household members with substance use or mental health diagnoses [8]. In an assessment of the 2016 National Survey on Children’s Health, Bethell and Davis [9] found that ACEs are relatively common in the US; 46% of children experienced at least one of nine negative experiences captured by the survey, and 22% reported more than one negative experience. Reports of two or more ACEs varied by race/ethnicity, such that Black children demonstrated the highest proportion of two or more ACEs (34%), followed by Latino/Latina (22%), White (19%), and Asian (6%). AI/AN were not identified as a distinct group. Several studies of ACEs and non-IPV outcomes in adulthood have disaggregated results by race/ethnicity, including investigations into the effect of ACEs on excessive alcohol use [10], strained social relationships [11], and mental health and depression [12]. In each of these studies, ACEs were predictive of negative health outcomes primarily among racial/ethnic minority groups. In addition, gender differences in the relationships between ACEs and other non-IPV factors are evident in the literature, including parental bonding [13], alcohol dependence [14], and mental health problems [15]. To our knowledge, no IPV studies have disaggregated associations between ACEs and physical violence by race/ethnicity and gender. Given the wide variation in the prevalence of ACEs and IPV between racial/ethnic and gender groups, as well as the associations between ACEs and multiple negative health outcomes most common among racial/ethnic minority groups, an investigation into racial/ethnic differences in the association between ACEs and IPV is warranted.

While conceptualizations of ACEs and IPV might naturally frame ACEs as exposures and IPV as an outcome, important frameworks exist that help contextualize both. First, child abuse is a life-course social determinant of health that indirectly affects health in adulthood through its influence on multiple health risk behaviors [16]. Second, it is important to emphasize IPV as a type of revictimization that some ACE survivors may be more prone to [17, 18]. In this sense, there may be moderating pathways between ACEs and IPV, including other experiences of violence or positive relationships in adulthood. In addition, some health disparity characteristics among adults, including low-SES and rural geography, may confound the relationships between ACEs and IPV.

Investigations of positive childhood experiences are less common than those of ACEs, but recent research has argued for a focus on safe, supportive, and nurturing childhood relationships (SSNRs) in addition to ACEs, child abuse, and neglect as predictors of outcomes in adulthood [19]. The National Center for Injury Prevention and Control defines safety as “the extent to which a child is free from fear and secure from physical or psychological harm within their social and physical environment”, and defines nurturing as “the extent to which children’s physical, emotional, and developmental needs are sensitively and consistently met” [19, p.6]. An estimated 1 in 15 children in the US experience consistently low levels of SSNR. Compared to ACEs, there is less convention around defining positive childhood typologies, and questions addressing the cumulative effects of multiple SSNRs are understudied [20]. For example, supportive relationships at school and between peers were positively associated with overall functioning and a low risk of substance abuse as an adult [21]. Black and Schutte [22] studied positive and loving relationships between children and either parent, and found an association with more stable and trusting romantic relationships in adulthood. Definitions of SSNR vary in the public health literature, and few reports disaggregate the correlates of SSNRs by race/ethnicity or gender. The effects of SSNRs and ACEs are generally in opposite directions, but evidence is mixed regarding a possible interaction effect of SSNRs on the association between ACEs and negative outcomes. While SSNRs are broadly associated with positive health and social outcomes, Thornberry and colleagues [23] found that the strength of the effect of ACEs may outweigh the effect of SSNRs among children who experience both. Relatedly, Herenkohl and colleagues [24] found no evidence that the effect of child abuse on adult mental health varied across levels of SSNR. In contrast, findings from the Kauai Longitudinal Study identified that among children who experienced ACEs, those who developed close bonds with at least one competent, emotionally stable, nurturing adult were much more likely to demonstrate high levels of resilience in adulthood, compared to children without those close bonds [25, 26]. Several other reports have identified paths through which positive childhood experiences buffer the effects of negative experiences on health in adulthood [27–29]. To our knowledge, no investigations have explored the extent to which IPV experiences may vary by racial/ethnic and gender groups while parsing out the competing influence of ACEs and SSNR.

While there is broad consensus that some ACEs are associated with IPV, less is known about the associations between multiple ACE typologies on physical IPV or differential associations by race/ethnicity and gender. Similarly, SSNRs are generally associated with positive health outcomes, but evidence of racial/ethnic and gender differences in the relationship between SSNRs and physical IPV is also lacking. Finally, no reports have demonstrated comparisons in IPV probabilities between racial/ethnic and gender groups while also accounting for the potential confounding of multiple health disparities characteristics. Thus, using nationally representative data, we sought to create a comprehensive epidemiologic profile of group differences in the relationship between physical IPV and the competing factors of multiple ACE typologies and SSNRs. Our hypothesis of differential associations was guided by an examination of the extant literature, which has demonstrated racial/ethnic differences in risk for both ACEs and IPV, as well as relationships between ACEs and other non-IPV factors for which racial/ethnic minority groups are at a particularly high risk of exposure.

Methods

Data source and sample

We used the National Epidemiologic Survey of Alcohol and Related Conditions-III (NESARC-III) (2012–2013) to identify a sample of Black, Latino/Latina, AI/AN, Asian/Native Hawaiian/Pacific Islander (Asian/NH/PI), and White men and women with complete data for all study variables (n=35,614). All analyses were conducted in 2019. Access to NESARC-III was granted to the study team by the National Institute on Alcohol Abuse and Alcoholism and approved by the local university’s Institutional Review Board. The complex survey design of NESARC-III has been described extensively elsewhere [30]. In short, NESARC-III is nationally representative, cross-sectional data, and includes many variables describing a wide range of social and environmental determinants of health. The survey also includes values for population weights and strata to reflect its complex sampling methods. Black, Latino/Latina, and Asian/NH/PI adults were sampled at a higher rate than the remainder of the population to ensure reliable estimates of these groups. Additionally, there were special considerations with outreach to the AI/AN population. A recruitment strategy was developed for notifying AI/AN organizations and tribes about NESARC-III participation, primarily to address survey subject matter and the process for non-compulsory saliva collection [30].

Measures

Race/ethnicity, gender, and intimate partner violence

We used all available NESARC-III race/ethnicity and gender variables to indicate White, Latino(a), Black, Asian/NH/PI, and AI/AN, as well as men and women. NESARC-III survey participants were asked if they had ever personally experienced a stressful or traumatic event. Those who responded affirmatively to this question were then asked to identify up to four stressful or traumatic events from a list. From this list, we defined any experience with physical IPV as whether a respondent reported ever being “beaten up by a spouse or romantic partner”. Modern best practices in measuring IPV are more comprehensive than our IPV indicator [31–33]. Thus, our definition of IPV is intended as a proxy indicator for experiences of physical interpersonal violence perpetrated by a partner. Importantly, there is no indication of the respondent’s age at the time of IPV exposure, meaning it is possible that a respondent could be answering this question based on adolescent partner trauma. In one scenario, cases of IPV could possibly precede experiences of ACEs, resulting in limited inference about the causal path between ACEs and IPV. However, because the prevalence of reported IPV was lowest among our youngest age group of 18–29 years (17%) and increased with age (x²=91.05, p<0.0001), it is reasonable to expect that the majority of the sample reported on IPV experienced in adulthood.

Adverse childhood experiences

NESARC-III ACE and SSNR survey items are listed in Supplemental Table 1. We categorized 24 ACE variables into five groups identified by Felitti and colleagues [8] in the Adverse Childhood Experiences Study (physical abuse, psychological abuse, sexual abuse, witnessing abusive treatment of a mother, and abusive/dysfunctional household environment). Response options for questions addressing physical abuse, psychological abuse, sexual abuse, and witnessing abuse against a mother included never, almost never, sometimes, fairly often, and very often. Household environment questions had binary response options of yes or no. We scored each variable as 0 if the respondent reported no experience (never or no), or 1 if the respondent reported any experience (almost never, sometimes, fairly, very often, or yes). The scores for the five ACE categories were derived from the sum of the variables in each category, resulting in scores for physical abuse (0–4), psychological abuse (0–6), sexual abuse (0–4), abuse towards a mother (0–4), and household environment (0–6).

Safe, supportive, and nurturing childhood family relationships

We used five SSNR questions that closely resemble items grouped as emotional indicators in the Childhood Trauma Questionnaire-Short From screening instrument developed by Bernstein and colleagues [34]. Response options for SSNR questions included never true, rarely true, sometimes true, often true, and very often true. Responses of never true and rarely true were recorded as 0, while any other level of positive endorsement (sometimes true, often true, and very often true) was recorded as 1. Family SSNR scores ranged from 0–5.

For both our ACEs and SSNR scores, we chose an analytic strategy of dichotomizing each item then creating a scale score, as opposed to treating each item as continuous. The qualitative nature of these Likert scale response options does not translate well into a continuous framework. The differences between each subsequent response option are not numeric. This makes the argument for treating the difference between ‘sometimes’ and ‘fairly often’ as numerically equivalent to the distance between ‘fairly often’ and ‘often’ difficult to defend. In addition, our scoring strategy of any versus no reported ACEs is easily interpretable by a wide range of audiences.

Covariates

We considered additional covariates based on a priori understanding of factors in adulthood that likely diminish or exacerbate disparities related to IPV, including age group (18–29, 30–39, 40–49, 50–59, or 60+ years) educational attainment (less than high school, high school completion, some college, or college completion), self-reported health (excellent/very good, good, or fair/poor), any experience with a major depressive episode throughout the lifetime (yes or no), household income (less than $20,000, $20,000–40,000, $40,000–60,000, or more than $60,000), and urbanicity (residence in urban or rural geographies).

Analysis

We used SAS (v9.4) for all analyses [35]. We used WEIGHT and STRATUM statements within SURVEY procedures to account for the parent study’s complex sampling design. First, we generated descriptive statistics by calculating within-group proportions of ever experiencing IPV each race/ethnicity by gender. We also calculated within-group mean scores for each of our five ACE typologies, mean scores for our SSNR variable, and frequency and percentage distributions for educational attainment, household income, and urbanicity.

We calculated within-group adjusted predicted IPV probabilities for each ACE category and SSNR. First, we generated group IPV means for all ACE category and SSNR scores by race/ethnicity and gender. Second, in a total of six models representing each of the five ACE scores and SSNR, we used linear regression to model each score predicting the IPV mean, stratified by race/ethnicity and gender. All ACE models were adjusted for SSNR score, and the SSNR model was adjusted for five individual ACE scores. All predictive models were also fully adjusted for age, education, self-reported health, major depression, income, and urbanicity. For the SSNR model, we assessed the possibility of multicolinearity between the five ACE scores using a threshold of r<0.80. All models also included educational attainment, household income, and urbanicity as covariates. Finally, we plotted the predicted IPV probability by race/ethnicity and gender across all possible ACE and SSNR scores.

Results

The group proportions of those who had ever experienced physical IPV varied widely by race/ethnicity and gender (Table 1). Nearly 21% of AI/AN women reported physical IPV, followed by Black women (10.4%) and Latina and White women (9.5%). Men reported low rates of physical IPV, ranging from 1.0% among Asian/NH/PI men to 2.4% among Black men. The highest mean ACE score was among AI/AN men reporting physical childhood abuse (1.6 out of 4). AI/AN women demonstrated the highest mean score for every other non-physical ACE typology, as well as the lowest SSNR score.

Table 1:

Group proportions ever experiencing physical IPV and ACEs and SSNR mean scores – NESARC-III (2012–2013, n=35,614)

		Women (n=20,018)
Variables		Black		Latina		Asian/NH/PI		AI/AN		White		ANOVA SS	p
		n=4,480 (12.4%)		n=3,864 (14.3%)		n=929 (5.7%)		n=294 (1.8%)		n=10,451 (65.8%)
ACEs
ACE typologies (mean score, 95% CLM)
Physical (0–4)		1.04 (0.99–1.09)		1.00 (0.94–1.05)		0.76 (0.66–0.86)		1.46 (1.26–1.67)		1.04 (1.01–1.08)		133.01	<0.0001
Psychological (0–6)		0.88 (0.83–0.93)		0.99 (0.94–1.05)		0.79 (0.70–0.89)		1.50 (1.25–1.75)		0.96 (0.93–0.99)		169.56	<0.0001
Sexual (0–4)		0.42 (0.38–0.45)		0.38 (0.35–0.42)		0.16 (0.11–0.20)		0.89 (0.68–1.10)		0.38 (0.36–0.40)		122.07	<0.0001
Household environment (0–6)		0.45 (0.42–0.48)		0.48 (0.45–0.51)		0.17 (0.13–0.21)		0.78 (0.62–0.95)		0.53 (0.51–0.55)		185.08	<0.0001
Mother’s abuse (0–4)		0.54 (0.50–0.58)		0.52 (0.48–0.56)		0.29 (0.23–0.36)		0.75 (0.59–0.91)		0.40 (0.38–0.42)		127.85	<0.0001
SSNR (0–5) (mean score, 95% CLM)		4.64 (4.61–4.68)		4.52 (4.48–4.57)		4.72 (4.67–4.78)		4.45 (4.31–4.59)		4.59 (4.56–4.61)		58.52	<0.0001
Demographic		col %		col %		col %		col %		col %		x2	p
Ever experienced physical IPV												74.21	<0.0001
Yes		10.42		9.53		3.47		20.50		9.49
No		89.58		90.47		96.53		79.50		90.51
Age group (years)												550.72	<0.0001
18–19		25.53		28.94		26.05		18.31		18.21
30–39		17.74		23.05		18.76		19.60		14.28
40–49		18.83		20.07		20.93		27.61		16.80
50–59		17.74		13.56		16.28		20.75		19.31
60+		20.15		14.37		17.97		13.72		31.38
Educational attainment												1171.28	<0.0001
Less than high school		15.15		29.92		12.22		13.95		8.30
High school completion		29.41		26.24		15.55		23.72		24.27
Some college		37.56		29.75		22.97		43.61		36.05
College completion		17.88		14.08		49.25		18.71		31.36
Self-reported health												237.75	<0.0001
Excellent/very good		44.16		45.72		58.02		37.48		56.78
Good		31.65		32.77		29.03		30.64		27.67
Fair/poor		24.18		21.51		12.95		31.88		15.55
Lifetime major depressive episode												231.28	<0.0001
Yes		19.82		22.25		15.30		40.53		31.10
No		80.18		77.75		84.70		59.47		68.90
Household Income												683.18	<0.0001
<$20k		36.77		28.68		16.00		29.18		19.31
$20–40k		28.52		32.06		18.57		21.74		23.03
$40–60k		14.34		15.17		16.49		19.34		15.92
>$60k		20.36		24.07		48.92		29.75		41.74
Urbanicity												727.03	<0.0001
Urban		88.47		94.95		96.27		70.31		71.69
Rural		11.53		5.05		3.73		29.69		28.31
		Men (n=15,596)
		Black		Latino		Asian/NH/PI		AI/AN		White		ANOVA SS	p
		n=3,073 (10.9%)		n=3,049 (15.3%)		n=837 (5.8%)		n=201 (1.3%)		n=8,436 (66.7%)
ACES
ACE typologies (mean score, 95% CLM)
Physical (0–4)		1.23 (1.62–1.29)		1.04 (0.98–1.10)		1.04 (0.93–1.15)		1.62 (1.39–1.84)		1.15 (1.12–1.19)		74.83	<0.0001
Psychological (0–6)		1.04 (0.98–1.10)		1.16 (1.10–1.23)		1.02 (0.91–1.12)		1.46 (1.23–1.68)		1.05 (1.02–1.08)		73.63	<0.0001
Sexual (0–4)		0.18 (0.15–0.22)		0.13 (0.11–0.15)		0.10 (0.06–0.15)		0.17 (0.08–0.27)		0.12 (0.11–0.13)		6.17	0.01
Household environment (0–6)		0.46 (0.42–0.50)		0.41 (0.38–0.44)		0.19 (0.13–0.24)		0.71 (0.57–0.84)		0.43 (0.41–0.45)		78.73	<0.0001
Mother’s abuse (0–4)		0.44 (0.39–0.49)		0.42 (0.37–0.46)		0.25 (0.20–0.31)		0.61 (0.45–0.76)		0.32 (0.30–0.34)		54.83	<0.0001
SSNR (0–5) (mean score, 95% CLM)		4.67 (4.63–4.71)		4.58 (4.54–4.63)		4.77 (4.70–4.83)		4.62 (4.49–4.75)		4.67 (4.65–4.70)		27.82	<0.0001
Demographic		col %		col %		col %		col %		col %		x2	p
Ever experienced physical IPV												7.07	0.13
Yes		2.42		1.49		1.01		1.92		1.77
No		97.58		98.51		98.99		98.08		98.23
Age group (years)												476.14	<0.0001
18–19		27.55		31.23		25.24		18.79		19.57
30–39		17.93		24.51		19.94		18.85		14.97
40–49		18.96		19.34		20.22		16.27		17.68
50–59		18.49		13.30		16.22		24.90		19.83
60+		17.04		11.61		18.36		21.19		27.95
Educational attainment												1142.41	<0.0001
Less than high school		17.19		30.86		8.74		16.83		8.82
High school completion		34.50		29.78		14.45		28.88		25.88
Some college		33.35		26.52		27.80		37.60		32.20
College completion		14.94		12.83		49.00		16.68		33.10
Self-reported health												55.77	<0.0001
Excellent/very good		51.48		53.38		57.57		45.03		57.19
Good		27.80		28.93		25.43		27.26		27.67
Fair/poor		20.70		17.67		16.98		27.70		15.13
Lifetime major depressive episode												81.41	<0.0001
Yes		12.13		13.08		10.04		20.34		18.02
No		87.86		86.92		89.96		79.66		81.98
Household Income												574.85	<0.0001
<$20k		30.44		22.04		16.21		33.92		14.76
$20–40k		26.07		31.06		18.91		23.26		20.54
$40–60k		16.55		18.11		13.74		12.75		16.51
>$60k		26.94		28.79		51.22		30.06		48.20
Urbanicity												539.37	<0.0001
Urban		87.28		94.02		97.93		70.00		72.82
Rural		12.72		5.98		2.07		30.00		27.18

The adjusted predicted probabilities of ever experiencing physical IPV increased with each additional ACE score for all five ACE typologies among women (Figures 1A–1E) and men (Figures 2A–2E). The plotted point estimates and their 95% confidence intervals, as well as beta coefficients, standard errors, and p-values for the predicted probability regression estimates, are presented in Supplemental Table 2. Racial/ethnic variation in predicted IPV was more pronounced among women. Also among women, strong associations were identified for the household environment ACE category (Figure 1D) among Blacks (β=0.053, p<0.0001) and Latinas (β=0.052, p<0.0001), followed by Whites (β=0.042, p<0.0001) and Asian/NH/PI (β=0.038, p<0.0001). AI/AN women had the highest predicted IPV probability estimates at the highest possible ACE score for three of the five ACE categories, including physical abuse (Figure 1A; 31.0), psychological abuse (Figure 1B; 35.6), and witness of abuse towards a mother (Figure 1E; 37.0).

Figure 1:

WOMEN – Adjusted predicted probability of ever experiencing IPV by (a) physical, (b) psychological, (c) childhood sexual abuse, (d) household environment, (e) abuse towards mother ACE scores, and (f) familial SSNR score – NESARC-III (2012–2013, n=20,018)

Figure 2:

MEN – Adjusted predicted probability of ever experiencing IPV by (a) physical, (b) psychological, (c) childhood sexual abuse, (d) household environment, (e) abuse towards mother ACE scores, and (f) familial SSNR score – NESARC-III (2012–2013, n=15,596)

Note: The y-axis for women ranges between 0–0.40. The y-axis for men ranges between 0–0.10.

A negative association was observed in the relationship between SSNR and predicted IPV among men and women (Figures 1F and 2F), such that the probability of IPV decreased with each subsequent score among all but one group. The strongest negative associations were identified among White women (β=−0.031, p<0.0001) and AI/AN women (β=−0.029, p<0.0001). Only Black men demonstrated a positive relationship between SSNR and IPV (β=0.001, p<0.0001), although the magnitude of the association was notably weaker than all other SSNR associations.

For physical and psychological abuse, witnessing abuse towards a mother, and SSNR, there was a clustering of racial/ethnic groups among women (Figures 1A, 1B, 1E, and 1F). First, AI/AN women had higher probability of experiencing IPV compared to all other groups across each of these scores. Second, Black, Latina, and White women had similar trends and probabilities of IPV and clustered in the middle of the figures. Third, Asian/NH/PI women had the lowest predicted IPV probability across all ACE and SSNR scores.

Discussion

Using a nationally representative sample, we assessed multiple ACE typologies and SSNR among racial/ethnic and gender groups to provide a narrative of physical IPV disparities in the United States. Physical IPV was much higher among women than men. A greater proportion of AI/AN women reported physical IPV compared to any other group, while AI/AN men reported the most experience with physical abuse as children. We identified racial/ethnic and gender differences in the predicted probability of ever experiencing physical IPV for ACEs and SSNR, adjusted for several factors likely associated with IPV risk as an adult. Positive relationships between ACEs and physical IPV – as well as negative relationships between SSNR and physical IPV – were particularly identifiable among women.

Our findings are broadly aligned with other reports that some ACEs are associated with varying types of IPV as an adult. Specifically, experiencing physical abuse as a child and witnessing parental violence [7], family dysfunction [36], and family violence [6] have been shown to be positively associated with experiences of IPV in adulthood. Our research adds to the extant literature by disaggregating results by race/ethnicity and gender, which allows for a more granular assessment of social and health disparities. Furthermore, we accounted for the influence of characteristics associated with health disparity populations, including education, socio-economic status, mental health diagnoses, and underrepresented rural geographies. Importantly, we acknowledge and appreciate the heterogeneity of racial/ethnic and gender groups across the US. At the same time, using a nationally representative sample is valuable in assessing general trends within groups and identifying higher level, systematic inequities experienced disproportionately by some groups.

While we identified several differential associations by race/ethnicity and gender, our focal narrative is on AI/AN, as both AI/AN women and men demonstrated particularly high risks for IPV and ACEs. AI/AN women demonstrated the highest predicted probabilities of physical IPV for all ACE typologies and across the majority of possible ACE scores. Importantly, AI/AN women also demonstrated the highest prevalence of multiple factors that likely compound the effects of ACEs and are bi-directionally related to IPV risk, including poor self-reported health and lifetime major depressive episodes. We noted that AI/AN women demonstrated the weakest significant association between IPV and ACEs related to the household environment compared to all other women. At the same time, AI/AN men demonstrated the strongest association between IPV and household environment ACEs compared to other men. One specific household characteristic – the incarceration of a parent – tends to affect boys more negatively than girls [37], and boys are more likely to lose a male role model than girls are to lose a female role model [38]. Given that the rate of jail incarceration is highest among AI/AN [39], the strong association between household environment and IPV among AI/AN men is further supported.

We found no associations between predicted IPV and childhood sexual abuse or witnessing abuse towards a mother among AI/AN men. This was a challenging finding to interpret within our overarching narrative of differential associations between racial/ethnic groups. While the mean sexual abuse and abuse towards a mother scores were low among all men, they were relatively higher among AI/AN men. Similarly, IPV prevalence was also low among men. We recommend that future research addresses specific ACE typologies and violence-related outcomes among AI/AN in general, and AI/AN men in particular.

Conversely, Asian/NH/PI women demonstrated the lowest predicted IPV probabilities, although the direction of the relationships between IPV and ACE and SSNR scores was similar to other female groups. Notably, racial/ethnic differences between White, Black, and Latina women were less clear, as the plotted predicted probability lines between these three groups were very close and often overlapping. Compared to varying intercepts, varying slopes may be more telling of disparities in the association between ACE score and IPV. Of note, detecting statistical significance is less likely with smaller sample sizes like our AI/AN women group, and p-values derived from regression equations should be interpreted in the context of other useful tools in assessing relationships, including trend lines, confidence intervals, and expert interpretation of the meaningfulness of group differences. Evaluating the clinical meaningfulness of group differences may lead to important and relevant conclusions about disparities that may otherwise be missed in evaluations based solely on binary alpha cutoffs. The presentation of our results is in line with a recent call by Ward and colleagues [40] for researchers to interpret racial disparities in health through multiple and comprehensive analytic strategies aimed at better understanding inequities in the patterns of health and exposures. In this sense, there may be some underlying latent social factors that affect White, Black, and Latina women similarly, and at the same time are particularly detrimental to AI/AN women. Broadly, access to social, legal, and health services is limited for AI/AN women [41–43], which may in turn exacerbate the effect of ACEs and enable the reoccurrence of IPV.

Our inclusion of SSNR in the context of ACEs and physical IPV sheds light on the need for public health research to focus on positive, non-deficit individual characteristics. We found that higher SSNR scores were related to a reduction in predicted IPV probability, particularly among White and AI/AN women. However, even though SSNR appears to be protective against IPV, AI/AN women still maintained a higher overall predicted probability compared to other racial/ethnic groups. Protective factors have been studied sporadically in relation to experiencing IPV [44, 45]. This is the first study that models protective childhood characteristics and their variation across racial/ethnic and gender groups, adjusted for experiences of ACEs, in order to more fully understand competing correlates of physical IPV as an adult. This is a key factor in investigations into differential associations between racial/ethnic groups, as focusing solely on naming deficiencies among mostly minority groups may serve to create or reinforce misconceptions of cultural differences. Our findings should be considered alongside latent factors that also affect exposure to IPV, including racially and economically oppressive systems that disproportionately inhibit access to IPV preventive and treatment services among racial/ethnic minority groups. Critically, families operating within these systems of oppression may face additional barriers to create and maintain nurturing home environments. The absence of adversity should not qualify as an indication of the presence of safety and support, as many individuals reporting no experience with child abuse may also report low levels of familial support or childhood nurturing [20]. We stress the importance of considering differences between racial/ethnic groups in public health research, as it allows for a more nuanced understanding of the social and systemic forces that shape health.

Finally, it is important to note that our estimates of ACEs and IPV do not account for heterogeneity within racial/ethnic groups or for non-binary gender identities. AI/AN cultures vary widely, which may affect the size and direction of associations, as well as the applicability of epidemiologic research on culturally diverse populations [46]. Additionally, binary gender indicators fail to account for the disproportionately high risk of IPV among non-binary and transgender individuals [47, 48]. Future research investigating more granular details of high-risk populations may better inform interventions aimed at ACEs and IPV prevention.

Limitations

Given the sensitive nature of questions addressing physical IPV and ACEs, survey responses endorsing experiences with physical IPV or ACEs may be underreported. The underreporting of IPV is broadly attributable to purposeful concealment, fear of retaliation or arrest of the perpetrator, embarrassment or shame, and the possible misunderstanding of IPV-related questions [49, 50]. Underreporting could bias our results in either direction, depending on the variable. NESARC survey staff were trained in comprehensive protocols designed to minimize discomfort or awkwardness during the course of sensitive personal questions [30], but there is no indication in the NESARC documentation that survey staff underwent IPV-specific training.

NESARC-III is a cross-sectional survey, which requires self-reporting of past events. The possibility of recall bias as it pertains to NESARC-III questions of childhood adversity is also a limitation of this study. The ability to accurately recall abusive childhood events is influenced by multiple factors throughout the lifespan, which affects the validity of measurement based on retrospective reports [24, 51]. Experiences of IPV were also subject to recall bias and/or underreporting, as the survey question addressing trauma and IPV was framed around any lifetime exposure. The measurement of ACEs and IPV would be more reliable in longitudinal or natural studies that capture experiences of child abuse. In addition, such longitudinal study designs that also capture health effects in adulthood are ideal for testing hypotheses of causal inference.

There are many forms of IPV, but we were limited in the use of physical IPV, specifically addressing being “beaten up by a spouse or romantic partner”. This was a limitation of available NESARC data. Narrowly defining IPV in its physical form, as well as using an unconventional physical IPV metric, limits the generalizability of the current findings. Future studies may investigate interactions between ACEs and SSNR using other or multiple definitions of IPV. In addition, we stress the importance of national surveyors – in particular those under the auspices of the National Institutes of Health – to consider more robust definitions of IPV in accordance with established and validated scales.

Conclusion

Women in general, and AI/AN women in particular, demonstrated disproportionate experiences with physical IPV in adulthood. Using five ACE typologies, we observed a clear and upward relationship between ACE scores and the predicted probability of experiencing physical IPV, and these relationships differed by race/ethnicity and gender. SSNR may be protective against physical IPV. The linear trajectories of predicted IPV probability across ACEs and SSNR scores were clearly strong and unique among women, with exceedingly high probabilities of IPV observed among AI/AN women. The juxtaposition of SSNR alongside ACEs helps to frame the overarching narrative more towards recognizing the inherent positive qualities across groups, which is often lacking in studies comparing effects between racial/ethnic groups. Advocating for SSNR in family-based interventions is a strategy of health promotion; its prevention-based counterpart necessarily involves naming characteristics of deficiency. Health services aimed at strengthening health throughout the life course would benefit by continuing to promote supportive and nurturing childhood relationships, while also acknowledging the broader social contexts in which families may struggle to provide that nurturing. Our investigation of differential associations by race/ethnicity was a unique contribution to the public health and violence literature, and we encourage future investigations to address more granular findings within population subgroups.