Post-Concussion Symptoms in Women With Head Injury Due to Intimate Partner Violence

Abstract

Limited research has examined the symptom sequelae of head injuries in women survivors of intimate partner violence (IPV), despite this community being at increased risk for neurotrauma due to partner abuse. The current study compared post-concussion symptom severity between women with and without IPV-related head injuries. Women were recruited from court jurisdictions in Kentucky, USA, after receiving a protective order for partner abuse. The sample included 268 women with no prior head injuries (age: M[standard deviation (SD)] = 31.8[9.8], 77.2% White) and 251 women with lifetime IPV-related head injuries (age: M[SD] = 31.8[9.8], 88.0% White). Women with IPV-related head injuries were slightly older (t = 2.46, p = 0.014) with lower education (χ² = 5.81, p = 0.016), were more frequently unemployed (χ² = 9.23, p = 0.002), and had a higher likelihood of residing in a rural setting (χ² = 30.16, p < 0.001). Women with IPV-related head injuries were also more often White (χ² = 10.47, p = 0.001), but this group difference was almost entirely related to rural versus urban residence. Women with IPV-related head injuries reported a higher severity of lifetime physical IPV (t = 7.27, p < 0.001, d = 0.64, 95% confidence interval [CI]: [.46, .82]) and sexual IPV (t = 4.65, p < 0.001, d = 0.41 [0.24, 0.59]). A three-factor model of post-concussion symptoms, inclusive of cognitive, physical, and emotional symptoms, fit well (χ² = 368.99, p < 0.0001, comparative fit index [CFI] = 0.974, Tucker-Lewis index [TLI] = 0.968, root mean square error of approximation [RMSEA] = 0.079 [0.071, 0.087]), and showed evidence for strong measurement invariance across women with and without IPV-related head injuries. The subscale and total scores each had acceptable reliability: cognitive (ω = 0.88 [0.86, 0.90]), physical (ω = 0.74 [0.70, 0.77]), and emotional (ω = 0.88 [0.86, 0.89]), and total score (ω = 0.93 [0.92, 0.95]). Women with IPV-related head injuries reported all individual post-concussion symptoms at a significantly higher frequency, with medium group differences in cognitive (t = 7.57, p < 0.001, d = 0.67 [0.50, 0.85]) and physical symptoms (t = 7.73, p < 0.001, d = 0.68 [0.51, 0.86]) and large group differences in emotional (t = 8.51, p < 0.001, d = 0.75 [0.57, 0.93]) and total symptoms (t = 9.07, p < 0.001, d = 0.80 [0.62, 0.98]). All sociodemographic characteristics were independently associated with post-concussion symptoms, as were physical IPV (total score: r = 0.28 [0.19, 0.35], p < 0.001) and sexual IPV severity (total score: r = 0.22 [0.13, 0.30], p < 0.001). In hierarchical regression analyses, controlling for sociodemographic characteristics (i.e., age, race/ethnicity, education, unemployment, and rural/urban residence) and physical and sexual IPV severity, IPV-related head injury was independently significant and accounted for significant additional variance when predicting cognitive (ΔR² = 0.05, p < 0.001), physical (ΔR² = 0.03, p < 0.001), emotional (ΔR² = 0.07, p < 0.001), and total symptoms (ΔR² = 0.06, p < 0.001). Negative-binomial regression resulted in similar findings. This study demonstrates that multiple sociodemographic and IPV history variables are related to post-concussion symptom severity, but IPV-related head injury was independently associated with greater symptom severity. Women with IPV-related head injuries may be at increased risk for unaddressed health problems spanning cognitive, physical, and emotional domains. Future research is needed to psychometrically evaluate assessment instruments for this population and to assess efficacy of interventions to address their unique health care needs.

Introduction

Traumatic brain injury (TBI) due to intimate partner violence (IPV) represents a significant public health issue,¹ but an understudied area in neurotrauma research.^2–4 The majority of TBI research has focused on athletes and service members and veterans,^5–7 involving predominantly male samples.^8,9 Although estimates vary significantly,¹⁰ 28–100% of women survivors of IPV may have experienced a TBI in their lifetime,¹¹ with repetitive mild TBI (mTBI) common in prior samples.^12,13 Guidelines on the assessment of concussion and mTBI typically include the evaluations of multiple domains following injury, including cognitive functioning, postural control, vestibular ocular functioning, and post-concussion symptoms.^14,15 Among variables assessed following mTBI, post-concussion symptoms have been associated with the largest effect sizes in the acute phase following injury (e.g., within 1–14 days)^16,17 and are the eponymic feature of persistent post-concussion symptoms, which include symptoms that persist more than 3 months following an mTBI.¹⁸ Among adults presenting to the emergency department with a head injury, an estimated 18–31% experience persistent post-concussion symptoms; and, of those with persistent symptoms, 30–50% experience disability at 3–6 months post-injury.¹⁹ An understanding of persistent post-concussion symptoms in women survivors of IPV may help determine those at risk for disability and in need of treatment within specific symptom domains.

Persistent post-concussion symptoms remain controversial, in that their etiology has been historically debated in research literature,²⁰ they are difficult to define,^18,21 and they are non-specific, occurring in healthy adults²² and individuals with preexisting conditions and no personal history of mTBI or concussion.^23,24 The misattribution of symptoms to an mTBI can also contribute to longer recovery following injury,²⁵ indicating the importance of ensuring that the symptoms are not attributable to another cause in need of treatment (e.g., preexisting migraines, mental health conditions). In service members and veterans, post-traumatic stress disorder (PTSD) has been associated with greater symptom severity.^26,27 Women with IPV-related TBI have elevated PTSD symptoms compared with women survivors of IPV without TBI,²⁸ indicating that post-concussion symptoms in this population may be partly attributable to difficulty processing and recovering from cumulative traumatic experiences.

Relatively few studies have examined post-concussion symptoms among women with IPV-related TBIs. Among the few studies, one found that, among women survivors of IPV, a brain injury severity score significantly predicted post-concussion symptoms in women with IPV-related brain injuries after controlling for age, education, childhood trauma, and partner abuse severity.²⁹ Another study also found a significant association between TBI history and post-concussion symptoms among women, even after controlling for age, IPV exposure, and positive screens for depression and post-traumatic stress disorder.³⁰ Three studies have examined the frequency of post-concussion symptom reporting by women survivors of IPV, indicating that women typically reported multiple symptoms that most commonly include physical (e.g., headache), emotional (e.g., depression, irritability), and cognitive symptoms (e.g., forgetfulness, inattention).^13,31,32 Thus, although there is limited research, these few studies collectively demonstrate that women with IPV-related TBIs have consistently reported post-concussion symptoms that do not appear fully attributable to other causes.

Prior research on post-concussion symptoms in women with IPV-related TBIs has provided evidence that symptoms serve as a meaningful outcome following injury and provide insight into the ongoing problems these women may experience. The current study builds on prior research examining post-concussion symptoms in women survivors of IPV by assessing the factor structure of post-concussion symptoms, determining whether symptoms fit within separate domains of injury sequelae, and, through a retrospective cohort design, determining whether a history of IPV-related head injury is related to greater post-concussion symptom severity across symptom domains. No prior research has examined the factor structure of post-concussion symptoms in women with IPV-related head injuries. Factor models can be helpful in summarizing the symptom experiences of individuals with mTBIs to identify specific domains of functioning that may serve as targets for intervention.

The current study tested a hypothesized three-factor structure of post-concussion symptoms, including cognitive, physical, and emotional symptom domains, aligning with prior factor models tested in other populations with mTBI.^33–38 Women survivors with and without IPV-related head injuries were also compared on cognitive, physical, and emotional symptom domains, with and without adjusting for physical and sexual IPV severity. IPV-related head injury was hypothesized to independently predict post-concussion symptom severity across symptom domains, accounting for variance in symptoms beyond that explained by IPV severity alone.

Methods

Participants

Participants (n = 756) included cisgender women from Kentucky, USA, who were 18 years and older (or 17 and emancipated) and had obtained a protective order against an intimate partner within 6 months prior to participating in the study. Participants completed a 1-year follow-up interview, which had a 94% follow-up rate (n = 709). Questions on head injury were only asked at the 1-year follow-up interview. Among the participants who completed the follow-up interview, 641 were asked questions about head injury, of whom 268 reported no head injury and 373 (58.2%) reported experiencing one or more lifetime head injuries and 255 (39.8%) reported experiencing one or more lifetime IPV-related head injuries. Among participants with head injuries, women were excluded if they: 1) reported no head injuries related to IPV (n = 118), and 2) reported having experienced a coma following a prior head injury (n = 4; duration in days: Mdn = 4, range: 3–28), as a coma indicated a severe brain injury with worse outcomes than typical for most cases of IPV-related head injury. Of note, women reporting one or more head injuries due to IPV could also report additional head injuries unrelated to IPV. This resulted in a sample of 268 women survivors of IPV with no reported lifetime head injuries and 251 women survivors with one or more IPV-related head injuries. The sociodemographic characteristics of each group are provided in Table 1.

Table 1.

Demographic Characteristics

	No head injury (n = 268)	IPV-related head injury (n = 251)	t/χ2	P	d/OR [95% CI]
Age, M (SD)	31.8 (9.8)	33.9 (8.9)	t = 2.46	0.014	d = 0.22 [0.04, 0.39]
Race, n (%)	–	–	χ2 = 10.47	0.001	OR = 2.17 [1.35, 3.50]
White	207 (77.2%)	221 (88.0%)
Black	53 (19.8%)	21 (8.4%)
Hispanic/Latina	3 (1.1%)	3 (1.2%)
Asian	2 (0.7%)	0 (0.0%)
Native American	2 (0.7%)	1 (0.4%)
Puerto Rican	0 (0%)	1 (0.4%)
Biracial	0 (0%)	0 (0%)
Education, n (%)	–	–	χ2 = 5.81	0.016	OR = 1.60 [1.09, 2.34]
No formal schooling	1 (0.4%)	0 (0%)
8th grade or less	15 (5.6%)	19 (7.6%)
Less than high school	50 (18.7%)	67 (26.7%)
GED	32 (12.0%)	32 (12.7%)
High school diploma	44 (16.5%)	50 (19.9%)
Trade/Technical school	17 (6.4%)	12 (4.8%)
Some college	73 (27.3%)	53 (21.1%)
College degree	30 (11.2%)	17 (6.8%)
Some graduate school	2 (0.7%)	0 (0%)
Graduate school degree	3 (1.1%)	1 (0.4%)
Employment, n (%)	–	–	χ2 = 9.23	0.002	OR = 1.72 [1.21, 2.44]
Unemployed	131 (49.2%)	156 (62.2%)
Part-time	37 (13.9%)	23 (9.2%)
Full-time	98 (36.8%)	72 (28.7%)
Area, n (%)	–	–	χ2 = 30.16	<0.001	OR = 2.69 [1.88, 3.85]
Urban	153 (57.1%)	83 (33.1%)
Rural	115 (42.9%)	168 (66.9%)
Income (US$), Mdna	$12.5k	$12.5k	t = 1.37	0.173	d = 0.12 [-0.05, 0.29]

^a Six participants were missing on Income, which was the total self-reported annual income from all sources in the past year. The Income range was the same for all groups (<$1,000 to >$50,000). χ² test for race was based on percentage White versus percentage other racial/ethnic identities; for education it was based on high school degree versus less than a high school degree; and for employment it was based on unemployed versus part-time or full-time employed.

CI, confidence interval; IPV, intimate partner violence; OR, odds ratio; SD, standard deviation.

Measures

As part of a larger study examining health among women with recent protective orders, participants were asked about sociodemographic characteristics, lifetime head injury history, and physical and psychological health problems spanning multiple domains. The sociodemographic questions included age, race/ethnicity, highest level of education, employment status, area of residence (i.e., urban or rural), and annual household income.

Head injury history

The head injury questionnaire began by contextualizing head injury as injuries that caused severe pain, being knocked out, coma, being hospitalized, or that resulted in memory problems, visual problems, difficult recovery, or rehabilitation, clarifying that minor bumps to the head or incidents that did not result in serious pain, actual injury, or loss of consciousness should not be recorded. Participants were then asked to answer yes or no to whether they ever received a head injury from the following mechanisms: automobile crash, motorcycle crash, all-terrain vehicle crash, bicycle crash, other vehicle crash, assault/abuse, gunshot wound, sports injury, fall/being pushed or shoved, near drowning, or some other mechanism of injury. They were asked how many times a head injury from each mechanism occurred, whether an intimate partner ever caused an injury within each mechanism category, and how many times the head injury was caused by an intimate partner within each mechanism category. This approach allowed for the quantification of lifetime head injuries both related to and unrelated to IPV. Participants were also asked whether they were ever hospitalized, lost consciousness, or experienced a coma due to any head injuries, and if so, how many times. The head injury interview questions are provided in Supplementary Appendix S1.

Post-concussion symptoms

The questions on health problems did not follow any established post-concussion symptom questionnaire, but many had construct overlap with existing instruments (e.g., cognitive, physical, and emotional).^33–38 These questions were embedded in a longer interview with other questions asking about additional health problems (e.g., chest pain, weakness, breathing problems). Questions aligning with the dimensions measured by existing post-concussion symptom questionnaires were selected from the longer interview materials and categorized as aligning with specific constructs.

The cognitive items included Trouble remembering things, Trouble concentrating, Difficulty making decisions, Your mind going blank, and The idea that something is wrong with your mind. The physical items included Headaches and head pain, Faintness or dizziness, Nausea or upset stomach, Numbness or tingling in parts of your body, and Trouble falling asleep. The emotional items included Nervousness or shakiness inside, Feeling easily annoyed or irritated, Temper outbursts that you could not control, Feeling blue, and Feeling no interest in things. One physical item, Headaches and head pain, was rated on the frequency at which they were experienced during the past 30 days based on a 5-point scale including the following response options: Never (0), Once or twice (1), A few times (2), Fairly often (3), and Very often (4). All other cognitive, physical, and emotional items were rated on how much a specific problem had bothered them in the last 7 days based on a 5-point scale including the following response options: Not at all (0), A little bit (1), Moderately (2), Quite a bit (3), and Extremely (4). The questionnaire is provided in Supplementary Appendix S2.

Physical and sexual IPV severity

Adapted items from the Conflict Tactics Scales^39,40 assessed lifetime physical and sexual IPV history. Individual IPV experiences were assigned weights based on severity, as used in prior research,^41–45 with weighted item values summed to create physical and sexual IPV severity scores for individual participants. For physical violence, the items were as follows, with the weights listed in parentheses: twist arm/hair (1), push/shove (1), grab (1), slap (1), kick (2), bite (2), punch/hit with something that could hurt (2), slam against the wall (2), beat up (5), burn/scald (5), choke/strangle (5), threaten with knife/gun (6), try to run down with car (6), and use knife/gun (8). For sexual violence, the items were as follows, with the weights listed in parentheses: sexual insistence without force (1); used threats to do sexual things, including intercourse (2); used physical force to do sexual things, including intercourse (3).

Procedure

Women survivors of IPV were recruited from four court jurisdictions that spanned three rural areas and one urban area at the time they received a civil protective order for partner abuse.⁴⁵ For each participant, a baseline interview was conducted between February 2001 and March 2003, lasting 3 to 4 h, and a follow-up interview was conducted between October 2001 and December 2004, lasting 2 h. The mean time between interviews was approximately 1 year (M = 11.7 months, standard deviation [SD] = 1.31), with a range of 9–18 months. Participant chose a public setting for the interviews, which included libraries and hospitals. Each participant provided written informed consent before participating in either interview. The study was approved by the University of Kentucky Institutional Review Board.

Statistical analyses

Sociodemographic group comparisons. Participants were compared on age using a t-test and χ² tests for categorical sociodemographic characteristics. Sociodemographic characteristics were dichotomized due to small cell sizes in some categories. Race/ethnicity was recoded as White (1) and other racial/ethnic identities (0), education was recoded as a high school diploma or equivalent (1) and less than a high school diploma (0), employment was recoded as unemployed (1) and part- or full-time employment (0), and residential area was coded as rural residence (1) and urban residence (0).

Confirmatory factor analysis and measurement invariance

A series of confirmatory factor analyses were conducted for post-concussion symptoms using MPlus v8.9⁴⁶ for the full sample and separately for women with and without IPV-related head injuries. A three-factor model was specified, with items loading onto separate cognitive, physical, and emotional symptom factors as categorized above. These analyses evaluated the psychometric properties of the symptom questions, which were not derived from an established post-concussion symptom questionnaire. Confirmatory factor analyses examined whether self-reported symptoms aligned with a hypothesized model of symptoms based on prior research.^33–38

All item responses were on a 5-point ordinal scale (range: 0–4) and were treated as polytomous variables, which involves an alternative estimator: weighted least squares with mean and variance adjusted (WLSMV). The WLSMV estimator assumes a continuous latent response variable underlies categorical responses, calculating threshold parameters that represent the values of this latent variable that distinguish between ordinal categories.^47,48 Model fit was assessed using the χ² goodness-of-fit test, the comparative fit index (CFI), the Tucker-Lewis index (TLI), and the root mean square error of approximation (RMSEA), which work appropriately for polytomous data.⁴⁹ For the χ² test, a non-significant value indicates adequate model fit,⁵⁰ but these values become inflated with large sample sizes leading to increased power to reject the null hypothesis.⁵¹ Alternative fit indices were used to interpret model fit, including the CFI and TLI, for which values ≥0.95 indicate acceptable model fit,⁵² and the RMSEA, for which values ≤0.08 indicate acceptable model fit.⁵³ For the total symptom score and subscale scores based on the three-factor model, omega (ω)⁵⁴ was calculated as an estimate of reliability using JASP,⁵⁵ with cutoffs of 0.65 for acceptable reliability and 0.80 for strong reliability.⁵⁶

Measurement invariance analyses were conducted to ensure the model fit was consistent for women with and without IPV-related head injuries. The establishment of measurement invariance ensures that scores derived from the factor model (i.e., cognitive, physical, and emotional subscale scores) are meaningfully comparable across groups.⁵⁷ For measurement invariance, a series of analyses fit the same model to separate groups and constrain parameters across groups to be equivalent, with each sequential model adding additional constraints. The configural model imposes the same factor model across groups, the weak model constrains the factor loadings to be equivalent, and, when analyzing polytomous data, the strong model constrains item thresholds to be equivalent.⁵⁸ Theta parameterization was used to allow the inclusion of residual variances in the model,⁴⁶ which are constrained to be equivalent for the configural model in order for the model to converge. Models were compared using the Δχ² and ΔCFI tests. For the Δχ² test, non-significant values indicate consistent fit across models, with an adjustment made for categorical data.^59,60 For the ΔCFI test, change ≤0.01 indicates consistent fit across models.⁶¹

Group comparisons in IPV and post-concussion symptoms

Women with and without IPV-related head injuries were compared on the frequencies at which they reported individual experiences of physical and sexual IPV history using χ² tests and the physical and sexual IPV severity scores using t-tests. The distributions of physical and sexual IPV approximated normality based on skewness (i.e., 0.06 and 1.08, respectively) and kurtosis values (i.e., −0.80 and −0.36, respectively), which were well under recommended cutoffs of 3.0 for skewness and 7.0 for kurtosis.⁶² For sexual IPV severity, Levene's test was significant (p < 0.001) and equal variance was not assumed, with Welch's t-test used for this analysis.

For post-concussion symptoms, the items within each domain were summed to create cognitive, physical, and emotional subscale scores (range: 0–20) and all items were summed to create a total symptom score (range: 0–60). Women with and without IPV-related head injuries were also compared on the frequencies at which they endorsed individual post-concussion symptoms using χ² tests. They were also compared on subscale scores and the total symptom score using t-tests. The distributions approximated normality for the total score (skewness = 0.74; kurtosis = −0.31) and the subscale scores: cognitive (skewness = 0.95; kurtosis = −0.06), physical (skewness = 0.65; kurtosis = −0.31), and emotional (skewness = 0.68; kurtosis = −0.59). For all t-tests examining symptom scores, Levene's test was significant and equal variance was not assumed, with Welch's t-tests used for these analyses.

Covariate evaluation

Sociodemographic characteristics and IPV severity were evaluated in relation to post-concussion symptoms, both subscale and total scores. For categorical covariates (i.e., race/ethnicity, education, employment, and area of residence), participants were compared on symptom scores using t-tests. For some comparisons, Levene's test for equality of variance was significant (p < 0.05), with equal variance not assumed through use of Welch's t-tests. For continuous covariates (i.e., age, physical and sexual IPV severity), parametric correlations (r) were calculated between each variable and symptom scores with a corresponding 95% confidence interval (CI).

Regression models

A series of univariate ordinary least squares hierarchical linear regression analyses was conducted, with separate sets of models predicting the cognitive, physical, and emotional subscales and the total symptom score. Three sequential models were tested, with each model introducing new predictors. Model 1 included sociodemographic covariates found to either differ significantly between women with and without IPV-related head injuries or to be related to the dependent variable. Model 2 included the sociodemographic covariates and two additional variables: sexual IPV severity and physical IPV severity. Model 3 included all covariates from prior models and introduced a dummy-coded IPV-related head injury variable (0 = No head injury; 1 = IPV-related head injury). A post hoc fourth regression model was also conducted using the total symptom score as a dependent variable, introducing two interaction terms into the model: an interaction between physical IPV and IPV-related head injury and an interaction between sexual IPV and IPV-related head injury. Multi-collinearity was assessed using cutoffs of <2.5 for the variance inflation factor⁶³ and >0.20 for tolerance.⁶⁴ Models were compared by examining the change in the F-statistic for significance (ΔR² as the effect size), determining whether the introduction of new predictors accounted for significant additional variance in the dependent variable.

Although skewness and kurtosis values approximated normality for the symptom scores, there were slightly zero-inflated distributions for the cognitive subscale (17.7% scored zero) and affective subscale (10.0% scored zero), but not the physical subscale (6.0% scored zero) or total score (2.3% scored zero). A series of negative-binomial regression models were conducted to account for this distribution type for the dependent variables. The same models completed for the ordinary least squares hierarchical regression analyses were specified using the generalized linear model, with model comparisons conducted using the Akaike Information Criterion (AIC).⁶⁵ The AIC penalizes for model complexity, with values closer to zero representing better fit. Models were compared based on the difference in AIC (i.e., ΔAIC), with reductions in AIC ≤ −2 indicating improved model fit.⁶⁶

Missing data

Data for three participants were missing on individual post-concussion symptom items, all three of which were substituted with the rounded average of the remaining items within the missing item's symptom domain. Data for three participants were missing on individual items asking about physical and sexual IPV experiences, which were not imputed because these items represented self-reported events. These three participants were deleted listwise from any analysis involving a comparison of those individual events or IPV severity variables, which included some group comparisons and the hierarchical linear regression models.

Power analysis

The recommended sample size for a confirmatory factor analysis with three factors and medium loadings (λ = 0.65) was n = 220 for a model with three indicators per factor (1-β = 0.87) and n = 130 for a model with five indicators per factor (1-β = 0.82),⁶⁷ meaning the separate samples of women with IPV-related head injury (n = 251) and without prior head injury (n = 268) had sufficient power for testing the measurement model. Per a prior study, post-concussion symptoms showed a roughly medium-sized group difference (d = 0.41) between women survivors with IPV-related brain injuries (n = 74) and without IPV-related brain injuries (n = 25), although the result was non-significant due to an underpowered analysis.²⁹ For all analyses in the current study, p < 0.05 was set as the threshold for a significant finding. For an independent samples t-test, the current sample size (n = 519) had sufficient power (1-β = 0.99) to detect a medium effect size (d = 0.50) at α = 0.05; and for the hierarchical linear regression analyses, the sample size had sufficient power (1-β = 0.99) to detect a medium effect size (f² = 0.15) at α = 0.05 for an R² increase and an individual beta-weight with eight predictors in the regression model.⁶⁸

Results

Participants with and without IPV-related head injuries were significantly different in terms of mean age, racial composition, educational attainment, employment status, and area of residence. Women with IPV-related head injuries were slightly older (p = 0.014, d = 0.22 [0.04, 0.39]) and more often were White (p = 0.001, OR = 2.17 [1.35, 3.50]), had less than a high school education (p = 0.016, odds ratio [OR] = 1.60 [1.09, 2.34]), were unemployed (p = 0.002, OR = 1.72 [1.21, 2.44]), and resided in a rural area (p < 0.001, OR = 2.69 [1.88, 3.85]). The two groups did not differ in annual household income (p = 0.173, d = 0.12 [–0.05, 0.29]). The racial difference between groups was most likely attributable to area of residence, in that 65.0% of White participants (n = 278) lived in rural settings compared with 5.5% of participants of other racial/ethnic identities (n = 5), χ² = 107.00, p < 0.001, OR = 31.88 [12.66, 80.25].

The frequencies of individual physical and sexual IPV experiences are reported in Table 2 for women with and without IPV-related head injury. Women with IPV-related head injury reported all physical and sexual IPV experiences at a higher rate than women with no head injury history, although there was no significant group difference in the frequency of being pushed or shoved (p = 0.052) or being bitten by an intimate partner (p = 0.095). The total physical and sexual IPV severity scores were higher among women with IPV-related head injuries.

Table 2.

Intimate Partner Violence History

Variable (weight in parentheses)	No head injury (n = 268)	IPV-related head injury (n = 251)	t/χ2	P	d/OR [95% CI]
Physical violence, MW (SD)	17.1 (9.8)	23.8 (10.9)	t = 7.27	<0.001	d = 0.64 [0.46, 0.82]
Twist arm or hair (1), n (%)	138 (51.7%)	169 (67.3%)	χ2 = 13.11	<0.001	OR = 1.93 [1.35, 2.75]
Push or shove (1), n (%)	215 (80.5%)	218 (86.9%)	χ2 = 3.78	0.052	OR = 1.60 [0.99, 2.57]
Grab (1), n (%)	214 (80.1%)	223 (88.8%)	χ2 = 7.41	0.006	OR = 1.97 [1.20, 3.24]
Slap (1), n (%)	121 (45.3%)	156 (62.4%)	χ2 = 15.15	<0.001	OR = 2.00 [1.41, 2.85]
Kick (2), n (%)	78 (29.2%)	133 (53.2%)	χ2 = 30.75	<0.001	OR = 2.75 [1.92, 3.96]
Bite (2), n (%)	33 (12.4%)	44 (17.6%)	χ2 = 2.80	0.095	OR = 1.52 [0.93, 2.47]
Punch or hit with something (2), n (%)	131 (49.1%)	180 (71.7%)	χ2 = 27.66	<0.001	OR = 2.63 [1.83, 3.79]
Slam against the wall (2), n (%)	135 (50.6%)	161 (64.1%)	χ2 = 9.75	0.002	OR = 1.75 [1.23, 2.49]
Beat up (5), n (%)	153 (57.1%)	200 (79.7%)	χ2 = 30.41	<0.001	OR = 2.95 [1.99, 4.36]
Burn or scald on purpose (5), n (%)	10 (3.7%)	30 (12.0%)	χ2 = 12.32	<0.001	OR = 3.51 [1.68, 7.33]
Choke (5), n (%)	152 (56.9%)	178 (70.9%)	χ2 = 10.95	<0.001	OR = 1.85 [1.28, 2.66]
Threaten with a knife or gun (6), n (%)	132 (49.3%)	157 (63.1%)	χ2 = 9.97	0.002	OR = 1.76 [1.24, 2.50]
Try to run down with a car (6), n (%)	34 (12.7%)	71 (28.4%)	χ2 = 19.76	<0.001	OR = 2.73 [1.74, 4.29]
Used a knife or fired a gun on (8), n (%)	95 (35.4%)	126 (50.8%)	χ2 = 12.41	<0.001	OR = 1.88 [1.32, 2.68]
Sexual violence, MW (SD)	1.2 (1.8)	2.1 (2.4)	t = 4.65	<0.001	d = 0.41 [0.24, 0.59]
Sexual insistence (1), n (%)	127 (47.4%)	146 (58.4%)	χ2 = 6.29	0.012	OR = 1.56 [1.10, 2.21]
Threats to do sexual things including intercourse (2), n (%)	36 (13.4%)	68 (27.3%)	χ2 = 15.47	<0.001	OR = 2.42 [1.55, 3.79]
Physical force to do sexual things including intercourse (3), n (%)	45 (16.8%)	82 (32.9%)	χ2 = 18.15	<0.001	OR = 2.43 [1.61, 3.69]

For the comparison of Sexual violence (M_W), equal variance was not assumed and Welch's t-test was used. Four participants were missing data on at least one Physical violence item, and two participants were missing data on at least one Sexual violence item.

CI, confidence interval; IPV, intimate partner violence; Mw, weighed mean; OR, odds ratio; SD, standard deviation.

Among women with IPV-related head injuries, 55.8% (n = 140) reported having experienced loss of consciousness (LOC) due to a head injury and 55.4% (n = 139) reported being hospitalized following at least one head injury. They reported, on average, 17.2 (SD = 50.5) lifetime head injuries of any mechanism (Mdn = 4, range: 1–515), 14.7 (SD = 40.3) lifetime head injuries due to IPV (Mdn = 3, range: 1–515), 3.3 (SD = 5.4) lifetime head injuries involving LOC (Mdn = 1, range: 1–35), and 2.2 (SD = 5.2) lifetime head injuries resulting in hospitalization (Mdn = 1, range: 1–60). The most common mechanism of injury was assault/abuse (n = 218, 86.9%), followed by falling or being pushed (n = 137, 54.6%) and automobile accidents (n = 99, 39.4%). Among women reporting falling or being pushed, 85.6% reported this being caused by an intimate partner.

Confirmatory factor analysis and measurement invariance

The three-factor model fit well in the total sample (χ² = 368.99, p < 0.0001, CFI = 0.974, TLI = 0.968, RMSEA = 0.079 [0.071, 0.087]), and separately in participants with no head injuries (χ² = 169.78, p < 0.0001, CFI = 0.980, TLI = 0.976, RMSEA = 0.060 [0.046, 0.073]) and participants with IPV-related head injuries (χ² = 280.20, p < 0.0001, CFI = 0.958, TLI = 0.950, RMSEA = 0.094 [0.082, 0.106]). The factor loadings for the three-factor model in each group are presented in Table 3 along with the reliability estimates for each symptom subscale and the total symptom score. All factor loadings exceeded conventional cutoffs (i.e., >0.40)⁶⁹ and omega estimates indicated strong reliability (i.e., ≥0.80) for the cognitive and emotional subscales and total symptom score and acceptable reliability for the physical subscale (i.e., ≥0.65).⁵⁶

Table 3.

Factor Loadings, Reliability Estimates, and Inter-Factor Correlations for the Post-Concussion Symptom Model

	Total sample (n = 519)		No head injury (n = 268)		IPV-related head injury (n = 251)
	λ	e	λ	e	λ	e
Cognitive symptoms	ω = 0.88 [0.86, 0.90]		ω = 0.85 [0.83, 0.88]		ω = 0.87 [0.85, 0.90]
Trouble remembering things	0.74	0.45	0.68	0.54	0.76	0.43
Trouble concentrating	0.90	0.19	0.89	0.21	0.91	0.18
Difficulty making decisions	0.85	0.28	0.83	0.32	0.83	0.31
Mind going blank	0.82	0.33	0.82	0.33	0.78	0.40
Something wrong with mind	0.82	0.32	0.82	0.32	0.78	0.39
Physical symptoms	ω = 0.74 [0.70, 0.77]		ω = 0.68 [0.62, 0.74]		ω = 0.72 [0.67, 0.78]
Headaches and head pain	0.52	0.73	0.43	0.82	0.56	0.69
Faintness or dizziness	0.73	0.47	0.74	0.45	0.68	0.54
Nausea or upset stomach	0.72	0.48	0.74	0.46	0.67	0.56
Numbness or tingling	0.68	0.54	0.66	0.56	0.65	0.58
Trouble falling asleep	0.70	0.51	0.62	0.62	0.69	0.52
Emotional symptoms	ω = 0.88 [0.86, 0.89]		ω = 0.87 [0.85, 0.90]		ω = 0.85 [0.83, 0.88]
Nervousness or shakiness	0.81	0.34	0.80	0.35	0.76	0.42
Easily annoyed or irritated	0.81	0.34	0.82	0.33	0.78	0.39
Temper outbursts	0.70	0.51	0.70	0.52	0.63	0.60
Feeling blue	0.85	0.28	0.83	0.31	0.84	0.30
Feeling no interest in things	0.89	0.20	0.90	0.20	0.88	0.22
Inter-factor correlations	r	SE	r	SE	r	SE
Cognitive-Physical	0.82	0.03	0.82	0.04	0.77	0.04
Cognitive-Emotional	0.87	0.02	0.87	0.03	0.85	0.03
Physical-Emotional	0.86	0.02	0.92	0.03	0.76	0.04
Total symptoms	ω = 0.93 [0.92, 0.95]		ω = 0.91 [0.90, 0.93]		ω = 0.92 [0.90, 0.93]

IPV, intimate partner violence; SE, standard error of the mean.

In measurement invariance analyses, the configural model fit well (χ² = 443.58, p < 0.0001, CFI = 0.971, TLI = 0.968, RMSEA = 0.072 [0.063, 0.081]), along with the weak model (χ² = 385.36, p < 0.0001, CFI = 0.979, TLI = 0.978, RMSEA = 0.059 [0.050, 0.068]) and strong model (χ² = 452.19, p < 0.0001, CFI = 0.978, TLI = 0.980, RMSEA = 0.057 [0.049, 0.065]). The change in CFI was below the threshold (i.e., ≤0.01) from the configural to weak (ΔCFI = 0.008) and weak to strong (ΔCFI = −0.001) model comparisons. The change in χ² was non-significant in the configural to weak model comparison, Δχ² = 9.13, p = 0.6921, but was significant in the weak to strong model comparison, Δχ² = 84.84, p = 0.0003. These findings collectively indicate that the cognitive, physical, and emotional subscales and the total symptom score have acceptable-to-strong reliability and that the means of the subscale scores can be meaningfully interpreted and compared across groups.

Symptom endorsement frequency and severity comparisons

The frequencies of individual post-concussion symptom endorsements for women with and without IPV-related head injury are provided in Table 4, along with average scores for each symptom subscale and the total symptom score. Women with IPV-related head injuries endorsed all symptoms at a higher rate than women with no head injuries. The most commonly endorsed symptom for women in both groups was Headaches and head pain, endorsed by 89.6% of women with IPV-related head injuries and 81.3% of women with no prior head injuries. For individual symptoms, the largest group differences were observed for Nervousness or shakiness (OR = 4.01 [2.75, 5.86]), Temper outbursts (OR = 3.21 [2.22, 4.65]), and Trouble falling asleep (OR = 3.02 [2.04, 4.48]). Group differences in total symptom severity were large (d = 0.80 [0.62, 0.98]) and medium-to-large for symptom subscales. The largest subscale difference was observed for emotional symptoms (d = 0.75 [0.57, 0.93]), with similar medium-sized group differences observed for the cognitive subscale (d = 0.67 [0.50, 0.85]) and physical subscale (d = 0.68 [0.51, 0.86]).

Table 4.

Comparison of Symptom Severity and Individual Symptom Endorsement by Head Injury History

	No head injury (n = 268)		IPV-related head injury (n = 251)		t/χ2	P	d/OR [95% CI]
Cognitive symptoms, M (SD)	3.8	4.0	7.1	5.6	t = 7.57	<0.001	d = 0.67 [0.50, 0.85]
Trouble remembering things, n (%)	171	63.8%	183	72.9%	χ2 = 4.95	0.026	OR = 1.53 [1.05, 2.22]
Trouble concentrating, n (%)	139	51.9%	182	72.5%	χ2 = 23.41	<0.001	OR = 2.45 [1.70, 3.53]
Difficulty making decisions, n (%)	117	43.7%	157	62.5%	χ2 = 18.56	<0.001	OR = 2.16 [1.52, 3.06]
Mind going blank, n (%)	119	44.4%	168	66.9%	χ2 = 26.61	<0.001	OR = 2.53 [1.77, 3.62]
Something wrong with mind, n (%)	66	24.6%	116	46.2%	χ2 = 26.53	<0.001	OR = 2.63 [1.81, 3.82]
Physical symptoms, M (SD)	5.0	3.9	7.9	4.5	t = 7.73	<0.001	d = 0.68 [0.51, 0.86]
Headaches and head pain, n (%)	218	81.3%	225	89.6%	χ2 = 7.14	0.008	OR = 1.99 [1.19, 3.30]
Faintness or dizziness, n (%)	70	26.1%	113	45.0%	χ2 = 20.28	<0.001	OR = 2.32 [1.60, 3.35]
Nausea or upset stomach, n (%)	126	47.0%	167	66.5%	χ2 = 20.09	<0.001	OR = 2.24 [1.57, 3.20]
Numbness or tingling, n (%)	88	32.8%	131	52.2%	χ2 = 19.91	<0.001	OR = 2.23 [1.57, 3.19]
Trouble falling asleep, n (%)	153	57.1%	201	80.1%	χ2 = 31.59	<0.001	OR = 3.02 [2.04, 4.48]
Emotional symptoms, M (SD)	4.6	4.6	8.4	5.4	t = 8.51	<0.001	d = 0.75 [0.57, 0.93]
Nervousness or shakiness, n (%)	120	44.8%	192	76.5%	χ2 = 54.38	<0.001	OR = 4.01 [2.75, 5.86]
Easily annoyed or irritated, n (%)	203	75.7%	218	86.9%	χ2 = 10.43	0.001	OR = 2.12 [1.34, 3.35]
Temper outbursts, n (%)	68	25.4%	131	52.2%	χ2 = 39.43	<0.001	OR = 3.21 [2.22, 4.65]
Feeling blue, n (%)	147	54.9%	194	77.3%	χ2 = 28.96	<0.001	OR = 2.80 [1.91, 4.10]
Feeling no interest in things, n (%)	119	44.4%	170	67.7%	χ2 = 28.58	<0.001	OR = 2.63 [1.84, 3.76]
Total symptoms, M (SD)	13.5	11.2	23.4	13.6	t = 9.07	<0.001	d = 0.80 [0.62, 0.98]

Levene's test for equality of variance was significant (p < 0.05) for all t-tests; and alternatively, Welch's t-tests were used with equal variance not assumed for all subscales and total symptom comparisons. A symptom was considered endorsed if it was rated as 1 or greater on a 5-point scale (range: 0–4).

CI, confidence interval; IPV, intimate partner violence; OR, odds ratio; SD, standard deviation.

Covariate assessment

The relationship between sociodemographic variables and IPV severity with symptom subscales and the total symptom score were evaluated. Age had a minimal-to-small correlation (95% CI reported in brackets) with all subscales and the total score, which was significant for the cognitive subscale and the total score: cognitive: r = 0.10 [0.01, 0.18], p = 0.028; physical: r = 0.09 [0.00, 0.17], p < 0.052; emotional: r = 0.06 [–0.03, 0.15], p = 0.167; and total score: r = 0.09 [0.00, 0.17], p = 0.041). Physical IPV had a small-to-moderate correlation with all subscales and the total score, all of which were significant: cognitive: r = 0.23 [0.14, 0.31], p < 0.001; physical: r = 0.30 [0.22, 0.37], p < 0.001; emotional: r = 0.22 [0.14, 0.30], p < 0.001; and total score: r = 0.28 [0.19, 0.35], p < 0.001. Sexual IPV had a small correlation with all subscales and the total score, all of which were significant: cognitive: r = 0.20 [0.12, 0.29], p < 0.001; physical: r = 0.21 [0.13, 0.29], p < 0.001; emotional: r = 0.17 [0.09, 0.25], p < 0.001; and total score: r = 0.22 [0.13, 0.30], p < 0.001.

Bivariate covariates were evaluated in relation to the symptom subscales and total symptom score in Table 5. All binary sociodemographic covariates were related to the total symptom score, meaning participants who identified as White, had less than a high school education, were unemployed, and lived in rural settings had higher total symptom severity. There was some variation in the relationship between the covariates and specific subscales. Participants identifying as White had significantly higher cognitive and physical symptoms, but not emotional symptoms. Participants with less than a high school education had significantly higher physical and emotional symptoms, but not cognitive symptoms. Participants who were unemployed and resided in rural settings had higher scores on all subscales.

Table 5.

Comparison of Symptoms by Binary Covariates

	M	SD	t	P	d
Cognitive symptoms
White (n = 428)	5.6	5.3	2.03	0.044	0.20 [-0.02, 0.43]
Other racial/ethnic identities (n = 91)	4.5	4.2
Less than high school education (n = 152)	6.0	5.3	1.84	0.066	0.18 [-0.01. 0.37]
High school graduate or greater (n = 367)	5.1	5.0
Employed (n = 232)	4.1	4.3	5.64	<0.001	0.49 [0.31, 0.66]
Unemployed (n = 287)	6.5	5.4
Urban (n = 236)	4.8	4.6	2.41	0.018	0.21 [0.04, 0.38]
Rural (n = 283)	5.9	5.4
Physical symptoms
White (n = 428)	6.7	4.5	4.02	<0.001	0.42 [0.19, 0.65]
Other racial/ethnic identities (n = 91)	4.9	3.9
Less than high school education (n = 152)	7.4	4.8	2.94	0.004	0.30 [0.11, 0.49]
High school graduate or greater (n = 367)	6.0	4.2
Employed (n = 232)	5.2	3.9	6.03	<0.001	0.52 [0.35, 0.70]
Unemployed (n = 287)	7.4	4.6
Urban (n = 236)	5.1	3.7	6.45	<0.001	0.56 [0.38, 0.73]
Rural (n = 283)	7.5	4.7
Emotional symptoms
White (n = 428)	6.7	5.4	1.86	0.064	0.21 [-0.01, 0.44]
Other racial/ethnic identities (n = 91)	5.5	5.0
Less than high school education (n = 152)	7.6	5.5	3.05	0.002	.29 [.10, .48]
High school graduate or greater (n = 367)	6.0	5.2
Employed (n = 232)	5.2	5.1	4.89	<0.001	0.43 [0.26, 0.61]
Unemployed (n = 287)	7.5	5.3
Urban (n = 236)	5.5	4.9	3.67	<0.001	0.32 [0.15, 0.49]
Rural (n = 283)	7.2	5.5
Total symptoms
White (n = 428)	19.0	13.7	2.98	0.003	0.30 [0.08, 0.53]
Other racial/ethnic identities (n = 91)	15.0	11.2
Less than high school education (n = 152)	20.9	13.8	2.96	0.003	0.29 [0.10, 0.48]
High school graduate or greater (n = 367)	17.2	13.0
Employed (n = 232)	14.4	11.9	6.16	<0.001	0.71 [0.36, 0.71]
Unemployed (n = 287)	21.4	13.7
Urban (n = 236)	15.5	11.7	4.52	<0.001	0.36 [0.19, 0.54]
Rural (n = 283)	20.6	14.2

For some comparisons, Levene's test for equality of variance was significant (p < 0.05), with equal variance not assumed for the following comparisons (i.e., Welch's t-tests were used): cognitive and physical subscales and total symptom comparisons based on racial/ethnic identity, physical subscale comparisons based on education, cognitive and physical subscales and total symptom comparisons based on employment, and all comparisons based on rural versus urban residence.

SD, standard deviation.

Hierarchical regression models

All tolerance values were above 0.70 and all variance inflation factors were below 1.5, indicating no concerns regarding multi-collinearity.^63,64 Residuals from the regression models had normal distributions and were homoscedastic. The results of each regression model, including F-values, individual beta-weights, and effect sizes, are provided in Table 6.

Table 6.

Ordinary Least Squares Hierarchical Regression and Negative-Binomial Regression Results

	Cognitive symptoms			Physical symptoms			Emotional symptoms			Total symptoms
	Model 1	Model 2	Model 3	Model 1	Model 2	Model 3	Model 1	Model 2	Model 3	Model 1	Model 2	Model 3
Ordinary least squares
Age	0.05*	0.06**	0.04	0.04*	0.05*	0.04	0.04	0.04	0.02	0.13*	0.15**	0.10
Race/Ethnicity	0.36	0.21	0.12	0.41	0.25	0.19	0.07	-0.08	-0.19	0.85	0.37	0.11
Education	-0.29	-0.15	-0.07	-0.53	-0.41	-0.35	-0.92	-0.81	-0.71	-1.74	-1.37	-1.13
Unemployed	2.20***	2.14***	1.95***	1.70***	1.63***	1.50***	1.77***	1.71***	1.48***	5.67***	5.48***	4.93***
Rural/Urban	0.45	0.04	-0.32	1.83***	1.43***	1.17**	1.15*	0.76	0.32	3.43**	2.24	1.18
Physical IPV	–	0.07**	0.04*	–	0.08***	0.06**	–	0.07**	0.04	–	0.22***	0.15**
Sexual IPV	–	0.32**	0.27*	–	0.22*	0.19*	–	0.23	0.17	–	0.77**	0.62*
IPV-related HI	–	–	2.48***	–	–	1.76***	–	–	3.03***	–	–	7.26***
F	7.41***	10.15***	13.19***	14.33***	16.55***	17.77***	7.18***	8.61***	13.32***	10.96***	13.91***	18.23***
R2	0.07	0.12	0.17	0.12	0.19	0.22	0.07	0.11	0.17	0.10	0.16	0.22
ΔR2	–	0.06***	0.05***	–	0.06***	0.03***	–	0.04***	0.07***	–	0.06***	0.06***
Negative-Binomial
Age	0.01***	0.01*	0.01	0.01	0.01	0.00	0.01	0.01	0.00	0.01	0.01	0.00
Race/Ethnicity	0.02	0.05	0.02	0.07	0.07	0.06	-0.01	-0.01	-0.01	0.03	0.04	0.02
Education	-0.06	-0.02	-0.02	-0.06	-0.03	-0.02	-0.14	-0.12	-0.12	-0.09	-0.06	-0.06
Unemployed	0.42***	0.41***	0.34**	0.27**	0.25*	0.22*	0.29**	0.28**	0.22*	0.32***	0.30**	0.26**
Rural/Urban	0.10	0.03	-0.05	0.29*	0.22	0.19	0.19	0.13	0.06	0.20	0.13	0.07
Physical IPV	–	0.01*	0.01	–	0.01*	0.01	–	0.01*	0.01	–	0.01*	0.01
Sexual IPV	–	0.06*	0.04	–	0.04	0.03	–	0.04	0.03	–	0.04	0.03
IPV-related HI	–	–	0.46***	–	–	0.28**	–	–	0.48***	–	–	0.40***
LRT ꭓ2	26.87***	47.09***	65.72***	26.16***	39.15***	46.75***	20.50***	33.32***	55.39***	25.59***	42.07***	58.99***
AIC	2845.15	2828.93	2812.65	3010.78	3001.78	2996.19	3023.05	3014.23	2994.16	4040.77	4028.29	4013.36
ΔAIC	–	-16.22	-16.28	–	-9.00	-5.59	–	-8.82	-20.07	–	-12.48	-14.93

^* p < 0.05, ^**p < 0.01, ^***p < 0.001.

Race/Ethnicity was coded White = 1 and Other racial/ethnic identities = 0. Education was coded High school diploma/GED or greater = 1 and Less than high school = 0. Unemployed was coded 1 = Unemployed and 0 = Employed either part-time or full-time. Rural/Urban was coded 1 = Rural residence and 0 = Urban residence.

AIC, Akaike information criterion; HI, head injury; IPV, intimate partner violence; LRT, likelihood ratio test.

Model 1 predicted each symptom domain and the total symptom score with sociodemographic characteristics that differed between groups and/or were independently related to the dependent variables. This model was significant for each dependent variable. In terms of individual beta-weights, the cognitive subscale was independently predicted by greater age and unemployment; the physical subscale was independently predicted by greater age, unemployment, and rural residence; and the emotional subscale was independently predicted by unemployment and rural residence. The total symptom score was predicted by greater age, unemployment, and rural residence. Model 2 introduced physical and sexual IPV as additional independent variables in the model, with this model explaining significantly more variance in all subscales and the total symptom score (ΔR² range: 0.04–0.06, ps < 0.001). Both physical and sexual IPV were independently predictive of the cognitive and physical subscales and the total symptom score, but only physical IPV was independently predictive of the emotional subscale. Model 3 introduced IPV-related head injury as an additional predictor, which accounted for significantly more variance in all symptom subscales and the total symptom score (ΔR² range: 0.03–0.07, ps < 0.001). Physical and sexual IPV remained significant predictors in the model for cognitive and physical subscales and the total symptom score, but neither were independently significant predictors of the emotional subscale.

Negative-binomial regression models

The results of the negative-binomial regression models, including likelihood ratio ꭓ² values, individual beta-weights, and AIC values are provided in Table 6. The results were mostly comparable to the findings from the hierarchical linear regression models. IPV-related head injury was an independently significant predictor of all symptom subscales and the total symptom score, and the AIC values were lowest for the models including IPV-related head injury as a predictor. There were also some notable differences. Sexual IPV was only independently predictive of cognitive symptoms, and physical and sexual IPV were not independently predictive of any symptom subscale or the total symptom scale when IPV-related head injury was added to the model. There were some potential issues with variable suppression for sociodemographic covariates in the hierarchical linear regression models, in that rural residence changed sign as a predictor of cognitive symptoms (i.e., the beta-weight went from a positive to a negative value) and race/ethnicity changed sign as a predictor of emotional symptoms with the addition of new predictors into the model, but these changes in sign were not observed in the negative-binomial models.

Post hoc analyses

Examining non-fatal strangulation and symptom severity. Women who experienced IPV-related head injuries reported a higher frequency of being choked than women with no head injuries (70.9% vs. 56.9%, p < 0.001, OR = 1.85 [1.28, 2.66]). That said, the majority of women with no reported head injuries experienced this form of IPV, which may account for symptoms in this sample. The exact wording of the variable was “Did an intimate partner ever choke, strangle, or put his hands on your throat?” and women were categorized based on whether they reported this experience or not.

Women without reported head injury (n = 152 with possible non-fatal strangulation and n = 115 without non-fatal strangulation) were compared using t-tests on cognitive (t = 1.73, p = 0.085, d = 0.20 [–0.04, 0.45]), physical (t = 1.86, p = 0.065, d = 0.23 [–0.01, 0.47]), emotional (t = 2.14, p = 0.033, d = 0.26 [0.01, 0.50]), and total symptoms (t = 2.18, p = 0.030, d = 0.26 [0.02, 0.50]), finding small group differences in emotional and total symptoms. Women with IPV-related head injuries (n = 178 with possible non-fatal strangulation and n = 73 without non-fatal strangulation) did not differ on cognitive (t = 1.16, p = 0.249, d = 0.16 [–0.11, 0.43]), physical (t = 0.77, p = 0.444, d = 0.11 [–0.17, 0.38]), emotional (t = 0.26, p = 0.795, d = 0.04 [–0.24, 0.31]), or total symptoms (t = 0.83, p = 0.407, d = 0.12 [–0.16, 0.39]). For the full sample, women with possible non-fatal strangulation (n = 330) and women without non-fatal strangulation (n = 188) differed significantly on cognitive (t = 3.00, p = 0.003, d = 0.26 [0.08, 0.44]), physical (t = 2.80, p = 0.005, d = 0.26 [0.08, 0.44]), emotional (t = 2.74, p = 0.006, d = 0.24 [0.07, 0.42]), and total symptoms (t = 3.21, p = 0.001, d = 0.28 [0.10, 0.46]), with all differences associated with small effect sizes.

Exploring interactions between head injury and IPV severity. An additional regression model was conducted, introducing a fourth model into the hierarchical linear regression analysis of the total symptom score. In this fourth model, two interaction terms were added to the model: an interaction between physical IPV and IPV-related head injury and an interaction between sexual IPV and IPV-related head injury. The physical and sexual IPV variables were centered to deal with issues of multi-collinearity associated with each term (i.e., tolerance values below 0.20 and variance inflation factors above 2.5), although even with centering the sexual IPV by IPV-related head injury interaction had a variance inflation factor of 3.12. In this model, there was no significant change in variance accounted for symptom severity (i.e., ΔR² = 0.01, p = 0.146). The individual beta-weight for the interaction between IPV-related head injury and sexual IPV was non-significant (B = 0.04, p = 0.520), and the interaction between IPV-related head injury and physical IPV was at the threshold for significance, but not below it (B = −0.13, p = 0.050). Based on a negative-binomial model, there was no improvement in model fit (AIC = 4013.82, ΔAIC = 0.46) with the introduction of the interaction terms, and the interaction terms between IPV-related head injury and physical IPV (B = −0.02, p = 0.081) and sexual IPV (B = 0.002, p = 0.967) were non-significant.

Discussion

The current study assessed post-concussion symptoms in cisgender adult women from Kentucky, USA who had a protective order against an intimate partner in the prior year. Within this sample, 39.8% of women reported a lifetime history of at least one IPV-related head injury; these women reported, on average, 14.7 (SD = 40.3) lifetime IPV-related head injuries and 17.2 (SD = 50.5) lifetime head injuries due to any mechanism. Compared with women survivors of IPV with no head injury history, women with IPV-related head injuries experienced multiple variables associated with greater socioeconomic disadvantage, including lower rates of high school completion, higher rates of unemployment, and higher rates of rural residence. IPV-related head injuries were also associated with a significantly higher lifetime severity of physical and sexual IPV.

As with prior research, post-concussion symptoms fit within a three-factor framework,^34,37 inclusive of cognitive, physical, and emotional symptom domains. The severity of symptoms within each of these domains was related to variables aside from IPV-related head injury, including sociodemographic variables associated with socioeconomic disadvantage and lifetime IPV severity. Socioeconomic disadvantage has been associated with a greater cognitive, physical, and emotional symptom burden in athletes.⁷⁰ Many variables approximating socioeconomic disadvantage in the current sample (e.g., rural residence, unemployment) may increase risk for poor health in general. Rural residence reduces access to some healthcare specialties that may be only available in urban centers^71,72 and unemployment may contribute to lack of health insurance, further limiting access to health care and worsening overall health.

The overall prevalence of symptoms among women without a reported head injury indicates the non-specific nature of symptoms often attributed to neurotrauma, with rates of individual symptom endorsements ranging from 24.6% to 81.3% for this group. It is possible that these symptoms could be attributable, in part, to concurrent injuries that did not involve blunt force trauma to the head, specifically non-fatal strangulation resulting in possible cervical injuries and/or anoxic-hypoxic brain injuries. A growing body of literature has begun examining non-fatal strangulation among women survivors of IPV,⁷³ finding that women disclosing strangulation reported headache and neck pain,⁷⁴ worse mental health and substance use,^75,76 and problems with speaking, swallowing, and neurological symptoms.^76,77

The majority of women with no head injuries reported being choked or strangled (i.e., 56.9%) and post hoc analyses examined whether symptoms differed based on this experience. For the full sample, there were small group differences in all symptom domains and total symptom severity based on self-reported history of non-fatal strangulation (d range: 0.24 to 0.28). Among women with no head injuries, there were slightly greater emotional and total symptoms (d = 0.26) in women reporting a history of non-fatal strangulation. There were no differences in symptom severity based on a history of non-fatal strangulation among women reporting IPV-related head injuries. These findings indicate a potential relationship between non-fatal strangulation and neurobehavioral symptoms in women survivors of IPV, indicating need for further research into this injury type and its sequelae.

The relationship between IPV and post-concussion symptoms, independent of head injury itself, aligns with prior research.³⁰ Post-concussion symptoms commonly occur in the absence of a recent mTBI, because the symptoms are non-specific. They have been consistently observed among individuals with a variety of preexisting conditions in the absence of head injury,⁷⁸ including mental health conditions.^23,79–81 The independent correspondence between IPV severity and post-concussion symptoms may be attributable to cumulative trauma exposure, both physical and emotional. Greater IPV severity has corresponded with greater PTSD symptoms,⁸² which have aligned with greater post-concussion symptoms as well.²⁹ Women survivors of IPV, regardless of history of head injury, reported many post-concussion-like symptoms. This finding has implications for the assessment and treatment of this population. First, if compared with normative data for symptom severity, women with IPV in general may report elevated symptoms; and likewise, among women with IPV-related head injuries elevated symptoms may or may not be attributable to a head injury. Second, women survivors of IPV without head injury may have interest in and benefit from interventions for cognitive, physical, and emotional symptoms, meaning interventions tailored for this community could be more inclusive of women with and without head injury histories.

Although IPV severity explains some portion of variance in post-concussion symptom reporting, at least two prior studies have shown that IPV-related head injuries are independently related to post-concussion symptoms.^29,30 In the current study, a personal history of IPV-related head injury was independently associated with all domains of post-concussion symptoms, even after controlling for sociodemographic characteristics and IPV severity. This result indicates that post-concussion symptoms are not fully attributable to individual differences in sociodemographic characteristics or the severity of prior IPV experiences.

There was a large group difference for total symptom severity (d = 0.80); and for the separate symptom domains, the largest group difference was observed for emotional symptoms (d = 0.75), with comparable medium-sized group differences in cognitive (d = 0.67) and physical symptoms (d = 0.68). All individual symptoms were endorsed at significantly higher rates by women with IPV-related head injury. Headache and head pain was the most commonly endorsed item by both women with and without IPV-related head injuries (i.e., 89.6% vs. 81.3%). This finding aligns with prior studies finding headache to be one of the most commonly reported post-concussion symptoms.^13,30,32 The largest group differences were observed for nervousness or shakiness, temper outbursts, and trouble falling asleep. IPV-related head injury was independently associated with a collection of health problems that may benefit from intervention.

Many prior researchers have emphasized the need to screen for head injury among women survivors of IPV,¹⁰ and researchers have developed or refined instruments for evaluation of this population^83,84; but many providers serving this population lack knowledge on TBI,⁸⁵ increasing barriers to the detection of neurotrauma in this population. Women with IPV-related head injuries would benefit from more frequent assessment of post-concussion symptoms, as these evaluations may detect health problems in need of treatment. That said, assessment may only be beneficial if relevant referrals and services are available.

Many validated instruments exist for the assessment of post-concussion symptoms,⁸⁶ but none have been psychometrically evaluated for use with women survivors of IPV. These instruments include the Rivermead Post Concussion Symptoms Questionnaire (RPQ), typically administered to patients from the emergency department⁸⁷; the Neurobehavioral Symptom Inventory (NSI), typically administered to military service members and veterans⁸⁸; and the Post-Concussion Symptom Scale and the Sport Concussion Assessment Tool (SCAT), typically administered to athletes.^89,90 There is no established or preferred scale for symptom assessment among women with IPV-related head injuries, although researchers have used the SCAT⁹¹ and RPQ^13,29 and have recommended the use of the RPQ.⁴ Future psychometric research on existing scales or newly devised scales for measuring post-concussion symptoms among women with IPV-related head injuries appears warranted.

Women with IPV-related head injuries may benefit from interventions to address post-concussion symptoms, and some preliminary research has examined interventions with this population.^92,93 Common symptoms reported by the current sample have evidence-based treatments, including headaches,⁹⁴ sleep problems,⁹⁵ cognitive functioning,⁹⁶ and emotional distress.⁹⁷ However, these interventions have not been extensively evaluated among women with IPV-related head injuries, indicating a need for controlled trials to assess intervention efficacy among this population. There have been multiple such studies applied to patients with mTBI from various populations (e.g., outpatients, veterans), finding support for multiple approaches, including trauma-informed care,⁹⁸ cognitive-behavioral therapy,^99,100 cognitive rehabilitation,^101,102 and a combination thereof.¹⁰³ Considering the symptom burden observed for women in the current study and prior research, this population may benefit from an empirical examination of interventions tailored to meet their clinical needs.

This study provides insight into post-concussion symptoms among women with IPV-related head injuries, but also involved limitations. This study relied on retrospective reporting of lifetime number of head injuries, which may lead to inaccurate recall. In some cases, the highest estimated number of lifetime head injuries (maximum = 515), hospitalizations (maximum = 60), and episodes of LOC (maximum = 35) were particularly high and may be overestimates and, at best, a proxy of repetitive head impact exposure. These estimates were based on a single item asking participants to recall the number of these experiences in their lifetime. The information available to gauge brain injury severity was limited, which is why the self-reported injuries were referred to as head injuries. Although participants who reported a coma were excluded, diagnostic information for brain injury, including Glasgow Coma Scale score and duration of LOC and post-traumatic amnesia, were unavailable. Time since injury was also not available. Time since injury is an important predictor of head injury symptomatology, in that individuals with more recent injuries tend to report greater symptom severity.¹⁰⁴

If this sample includes an over-representation of women with acute injuries (e.g., within 3 months), the severity of the symptoms reported may be elevated and not correspond to the typical severity of chronic symptoms experienced by women with more distal injuries. The study did not involve an established post-concussion symptom questionnaire and required the construction of a questionnaire consistent with existing questionnaires based on general symptom reporting. This study was cross-sectional, meaning a causal relationship between prior head injury and self-reported symptoms can be inferred but was not established. The increased severity of post-concussion symptoms among women with IPV-related head injuries may be attributable to another variable not considered in the statistical model. For example, current and prior mental health conditions, which are related to post-concussion-like symptom reporting,¹⁰⁵ were not adjusted for in the statistical model; and conditions, such as depression, anxiety, and trauma-related disorders, could independently affect symptom reporting, or possibly interact with head injury to increase symptom reporting. Future research should address how mental health conditions relate to post-concussion symptom reporting by women survivors of IPV. There were also no corrections made for multiple comparisons, and the current findings should be replicated in future confirmatory research.

This study builds on a growing body of research examining outcomes following IPV-related head injuries. In addition to greater post-concussion symptoms,^13,30,32 published studies have identified reduced diffusivity and functional connectivity on structural and functional neuroimaging,^13,106 worse performances on cognitive testing,^12,107 and greater mental health symptoms^{12,28,108–110} among women with IPV-related head injuries. This largely understudied population^1–4 would benefit from further examination to understand multiple possible health care needs, including physical health, mental health, cognitive functioning, and possible substance use.^10,111 As multiple researchers have emphasized the need to better understand how neurotrauma affects this community,^2–4 future investigations should continue to identify their health care needs, methods by which to identify them, and approaches to train providers working with this community to best address those needs.

Transparency, Rigor, and Reproducibility Summary

This study was not pre-registered because it was based on archival data. The analysis plan was not formally pre-registered. The existing data included a sample size of 756 cisgender adult women from Kentucky, USA, of whom 641 were asked questions about head injury during a follow-up interview. Participants were excluded if they reported head injuries unrelated to IPV (n = 118) or head injuries resulting in a coma (n = 4). The final sample included 268 with no head injuries and 251 with IPV-related head injuries. Before analyses, the sample sizes were assessed for power, with the existing sample having sufficient power (1-β = 0.99) to detect a medium group difference (d = 0.50) at p < 0.05.

Data were collected from 2001 to 2004 via in-person interview. All symptom reporting was based on self-report, with all interview questions described in the text above. Data analyses were performed by investigators who were aware of relevant characteristics of the participants. The software used for analyses included MPlus for factor analysis, JASP for omega reliability calculation, and SPSS for correlations and group comparisons. The reliability and validity of symptom reporting was assessed in this study. The sample sizes and degrees of freedom reflect the number of independent measurements (i.e., number of individual participants). Missing data were minimal with mean imputation for symptom reporting and listwise deletion used for IPV reporting in some analyses. Effect sizes and confidence intervals have been reported in the abstract for primary outcomes and main text for all outcomes. Because IPV-related head injury is a new area in neurotrauma research, corrections were not made for multiple comparisons. No replication or external validation studies have been performed or are planned/ongoing at this time to our knowledge. The data from this study will not be shared because the participants did not consent for their data to be shared, even in a de-identified format, at the time of initial data collection. The authors agree to provide the full content of the manuscript on request by contacting the corresponding author.

Authors' Contributions

Justin E. Karr conceptualized the study, conducted the statistical analyses, and wrote the manuscript. T.K. Logan conceptualized the study, led data collection, and reviewed and edited the manuscript.

Funding Information

This work was supported, in part, by a Building Interdisciplinary Research Careers in Women's Health (BIRCWH) grant (#K12-DA035150) from the National Institute on Drug Abuse (NIDA) of the National Institutes of Health (NIH). The data collection was supported by the National Institute on Alcohol Abuse and Alcoholism (NIAAA) grant (#AA12735-01) and the University of Kentucky General Clinical Research Organization funded by the NIH (#M01RR02602).

Author Disclosure Statement

No conflicting financial interests exist.

Supplementary Material

Supplementary Appendix S1

Supplementary Appendix S2