1 Introduction

Intimate partner violence (IPV) perpetration is a critical and necessary area of examination among college populations due to the heightened rates as well as the correlates particular to this population. Estimates of IPV perpetration among college students in the United States (US) are above 30% during the course of their college career [1]. The prevalence of IPV perpetration among college students has also been increasing over the last 10 years, with men reporting 12-month perpetration prevalence of 6.49% in 2018–2019 and women reporting a 3.5% perpetration prevalence in the same year [2]. IPV perpetration consistently links to other negative health risk factors and consequences and psychiatric disorders such as anxiety, depression, post-traumatic stress disorder, and alcohol and drug abuse [3]. Understanding the pathways between IPV perpetration and eating disorder (ED) symptomatology as well as some of these correlates (such as alcohol use and salience of alcohol) is important to prevention and intervention strategies.

Eating disorders (EDs) have a strong comorbidity with alcohol use and are similarly connected with IPV [4, 5]. These disorders are more prevalent during college years, although the majority who are diagnosed do not receive treatment and less than 6% are asked about their diagnosis by a healthcare professional [6, 7]. This contributes to a situation wherein many students may have clinically significant symptoms but may have either limited treatment options or face barriers to treatment including their own skepticism or lack of urgency to address the issue [7]. Additionally, there is burgeoning literature showing increased odds of physical violence perpetration among those screening at-risk for an ED [2]. Specifically, college students who screen positive (using the 5-item SCOFF screening tool) for an ED had 48% increased odds of physical violence perpetration.

Although a variety of risk factors of IPV perpetration have been identified using Dutton’s [8] nested ecological theory on partner violence [9], alcohol use or alcohol use disorder (AUD; the impaired ability to stop or control drinking despite adverse consequences) is recognized as both a primary risk factor, consequence, and leading correlate of IPV perpetration [10]. Among a population of IPV offenders, the majority (82.4%) were diagnosed with alcohol dependence and 17.6% were diagnosed with alcohol abuse [11]. Another study found that individuals with AUD were 2.57 times more likely to perpetrate IPV than individuals than individuals without AUD [12].

Research also suggests that an individual’s perception of the centrality of alcohol to their college experience is an independent predictor of drinking behavior over time, even when controlling for baseline drinking [13, 14]. Consequently, since alcohol becomes a more normative and socially accepted experience, especially in the college years, the relationship between alcohol use and IPV may actually also be associated with an individual’s perception of alcohol salience to their college experience, making this question important to understand. Thus, alcohol salience may help explain why the association between alcohol and IPV is not exactly linear [15]. Recent qualitative research provides alcohol-related themes that illustrate its role in perpetration against victims of physical or sexual violence. Specific themes that arose including perpetrators’ use of alcohol as a way to control the victim or commit targeted rape [16].

There are a variety of gender differences related to the connections between IPV perpetration and alcohol use in college populations. A longitudinal study conducted among women between the ages of 13–19 found that women were significantly more likely to report perpetrating physical violence as opposed to men (38% vs. 19%) [17]. The type and severity of injuries also exhibit some distinctions, with college student women reporting higher rates of psychological and physical IPV perpetration, whereas college student men who perpetrate may leave more serious injuries than women [18]. Additionally, in a study conducted among college students utilizing electronic diary technology to consider psychological aggression with drinking or non-drinking days, men’s odds of psychological aggression were substantially exaggerated on drinking days compared to women (men’s odds ratio = 7.03; women’s odds ratio = 1.60 [19]. Meta-analyses have indicated stronger effects for men perpetrating IPV with alcohol use than for women’s alcohol use and IPV perpetration [20]. There are also unique gender differences in the relationships between EDs/ED symptomatology and IPV as well. Increased odds are found between screening positive for an ED and physical violence perpetration among men college students, but not women college students [2].

Objectification theory is an underpinning theory of ED symptomatology as well as body image because it explains the experience where individuals (typically women, but also all genders) are reduced to their separate or a summation of their body parts [21]. When individuals are consistently treated in this reductionist way, they internalize this objectification and may also self-objectify. Objectification has been consistently linked to IPV perpetration [22, 23], and some findings suggest individuals (specifically men) who score high in objectification are less sympathetic to victims and assign less blame to perpetrators [24]. Objectification and self-objectification have also been connected with EDs, violence, and substance use, and in particular alcohol use [23, 25]. Additionally, alcohol use has been identified as a mediator between sexual objectification and sexual assault perpetration for college men [26]. There is uncertainty still about the directionality of IPV perpetration and EDs, yet there is plausibility for the connection with objectification and self-objectification in both directions. Especially among perpetrators who identify as men, high trait objectification is associated with more favorable attitudes towards IPV perpetration, including assigning less responsibility to perpetrators and minimizing the degree of harm to victims and feeling less sympathetic towards them. For directionality in the opposite direction, objectification after perpetration may occur to reduce the cognitive dissonance between their values and behaviors. If they perceive their victims as an object, they are better able to condone their behavior. This objectification in turn is often internalized and associated with EDs [27].

To our knowledge, no studies have explored the relationship between IPV perpetration, alcohol use and alcohol salience to the college experience, and ED symptomatology. Exploring these pathways is critical to understanding the unique ways interventions could prevent and address perpetration among at-risk college students using alcohol. Therefore, the purpose of this study is to model the path between IPV perpetration and ED symptomatology among college students who identify as men or women. Our study was guided by three aims. Our first aim was to determine the association between IPV perpetration and ED symptomatology, hypothesizing a positive association between the two. Our second aim examined the connection between alcohol use and alcohol salience to college life with IPV perpetration, again hypothesizing a positive association. The third aim considered the relationship between IPV perpetration, alcohol use and salience, and ED symptomology, with the hypothesis that ED symptomatology would be positively associated with these risk factors. Within this aim, we also wanted to determine differences in the model among those students who identified as a man or a woman. It is hoped this study will strengthen the literature via the inclusion of both men and women-identified college students to better understand these issues within a broader population than has been previously studied.

2 Methods

2.1 Procedures

All full-time undergraduate students (N = 15,536) at the main campus at a U.S. midsized midwestern university received an email invitation for the Student Health online survey. The anonymous survey was open for 2 weeks during March of 2018. After completing the survey, the students immediately received a $3 e-gift card for a local coffee shop. A sample of 3917 full-time students (25.21% response rate) completed the survey. For this study, measures were presented in random order to mitigate ordered effects in responses. In addition, some measures were only given to a random subset of the sample (i.e., planned missingness). Due to one of the outcome measures (IPV perpetration) falling into this category, the final sample included 1580 participants (40%) due only participants receiving this measure were included in the final sample. Mental health resources were provided as well as information related to sexual violence reporting in the consent and debriefing forms. This study was approved by the referent university’s Institutional Review Board (IRB#: 01191r).

2.2 Participants

A majority of the participants (61.4%, n = 970) identified as a woman. The respondents also identified as heterosexual (85.5%, n = 1382); 5.8% (n = 91) as bisexual, and 1.8% (n = 28) identified as gay. Most identified as White (88.9%, n = 1404); 7.8% (n = 124) indicated that they were Asian or Asian American; and 4.3% (n = 66) reported being Black or African American. In order to allow students to identify accurately, participants were able to select multiple races/ethnicities in the survey. There was a slight over-sampling of 1st year students (31.7%, n = 501); 23.2% (n = 367) of the sample reported being 2nd year students; 23.4% (n = 369) of the sample was in their 3rd year; and 20.2% (n = 319) were in their 4th year. The average age was 20.02 (SD = 1.63) years. Additionally, 10.1% (n = 159) students reported being international students. With respect to indicators of social economic status, 11.9% (n = 188) are first generation students and 12.2% (n = 193) are eligible for Pell grants. Approximately 46.5% (n = 734) indicated that they were members of a social sorority or fraternity. In addition, 52.7% (n = 832) of the participants lived on campus. Most participant characteristics are similar to those of the student population from which it was sampled. The exception is an oversampling of participants who identify as women.

2.3 Measures

Participants received a series of measures. Means, standard deviations, and Cronbach’s alpha estimates are provided in Table 1 for each study measure and are detailed below.

Table 1 Correlations between IPV Perpetration Types, ED Symptomatology, and Alcohol
Full size table

2.3.1 ED symptomatology measured by the SCOFF (Sick, Control, One, Fat, Food) questionnaire [28]

The SCOFF is used to identify potential cases of anorexia nervosa (AN) or bulimia nervosa (BN) before a more in-depth clinical assessment is conducted [28]. There are a total of five questions which are (1) “Do you make yourself Sick because you feel uncomfortably full?”; (2) “Do you worry that you have lost Control over how much you eat?”; (3) “Have you recently lost more than 14 lb (adapted from ‘One stone’ in original version) in a 3-month period?”; (4) “Do you believe yourself to be Fat when others say you are too thin?”; and (5) “Would you say that Food dominates your life?”. Answers are dichotomous with yes/no responses; answering ‘yes’ to two or more questions on the SCOFF indicates a need for further clinical assessment for either AN or BN. The SCOFF has a sensitivity of 86% and specificity of 83% across pooled estimates from a recent meta-analysis [29]. See Table 1 for the alpha, mean, and standard deviation.

2.3.2 IPV perpetration

Participants were asked if they had a committed/casual partner in the past 12 months. If participants indicated “yes” to the first question, they received the IPV scale. This six-item measure included both verbal and physical IPV perpetration items that occurred within the past 12 months and employed wording adapted from the Conflict Tactic Scale [30]. The Conflict-Tactic Scale-2 is empirically-based for use in a college student population [30]. The question prompted: “All couples have disagreements and conflicts. Please identify if any of the following behaviors occurred during any disagreements in the last 12 months.” Three items assessed verbal perpetration (e.g., “I shouted or yelled at my partner”) and three addressed physical perpetration (e.g., “I kicked or hit something belonging to my partner”). Using a screening version of the response options, participants selected “yes” or “no” to each item. Scales were then summed across subscale items (i.e., verbal perpetration). A similar version of this assessment was used when examining victimization in college students [31]. The KR20 is 0.69 (M = 0.47, SD = 0.88).

2.3.3 Alcohol use

After the definition of a standard drink, three items were used to measure alcohol consumption. The item assessing frequency of alcohol consumption asked, “In a typical week, how many days do you have at least one drink containing alcohol?” The other two items assessed number of drinks consumed in one occasion. The questions asked, “How many drinks do you have on a typical day when you are drinking?” and “During the last 30 days, what was the highest number of drinks that you drank on any one occasion?” These items are commonly used in the alcohol literature [32].

2.3.4 College life alcohol salience scale (CLASS) [13]

The CLASS assessed participants’ beliefs regarding the centrality of alcohol consumption within the college context [13]. The CLASS consists of 15 items; each item includes the prompt “To what extent do you agree with the following statements based on alcohol use during college?” followed by statements such as “To become drunk is a college rite of passage” and “College is a time for experimentation with alcohol.” Participants responded to items using a 5-point Likert scale with options including, “Strongly Disagree,” “Disagree,” “Neither Agree nor Disagree,” “Agree,” and “Strongly Agree.” The Cronbach’s alpha is 0.91 (M = 39.48, SD = 10.99).

2.4 Data analysis

Before fitting the structural equation model, several steps were taken as preliminary procedures. Specifically, patterns of missingness and correlations between variables were inspected, and these procedures revealed the data appeared to be missing at random (< 5% missing overall outside of planned missingness). For the t-tests, the total scores of the scale were used. Structural equation models were run to examine the relationship between experiences with IPV perpetration, alcohol use, alcohol salience, and ED symptoms. The models were run using WLSMV estimation using MPlus version 8.6 [33]. In the models, latent variables were denoted using circles/ovals (IPV perpetration, alcohol use), and manifest variables are symbolized by rectangles (alcohol salience and eating disorder symptoms). The following criteria were used to examine the model: (1) theoretical relevance (i.e., being informed by the previous research), (2) global fit indices (i.e., chi-square, CFI, and TLI), (3) microfit indices (i.e., RMSEA), and (4) parsimony. A non-significant chi-square suggests that the data does not significantly differ from the hypotheses represented by the model; for CFI and TLI, fit indices of above 0.90 (and preferably above 0.95) indicated a well-fitting model [34]. Browne and Cudeck[35] suggest a RMSEA of less than 0.05 indicates a well-fitting model. The SCOFF was treated as a manifest variable in the model due to its scoring as a screener.

3 Results

3.1 SCOFF—eating disorder symptoms

Approximately 16.7% (n = 264) of the sample endorsed two or more questions on the SCOFF indicating that further clinical evaluation for EDs was warranted. Of those who endorsed two or more items, the sample primarily identified as a woman (74.8%, n = 199 women; men, 22.9%, n = 61).

3.2 IPV perpetration

Rates of IPV perpetration across gender are presented in Table 2. The comparison across genders is limited to those who identify as men or women due to insufficient sample sizes in other gender identities. Approximately 28.4% (n = 448) indicated that they perpetrated some aspect of IPV. Women were statistically significantly more likely to report perpetration of IPV, as well as more likely to report both physical and verbal perpetration. However, the effect sizes indicate that there is only a small difference between those who identify as men and women with regard to the amount of IPV perpetrated.

Table 2 IPV perpetration rates across reported Gender Identity
Full size table

3.3 Alcohol use

In a typical week, participants reported drinking on average 1.54 (SD = 1.40) days per week. With regards to their typical consumption, they indicated that they drank an average of 3.28 (SD = 2.60) standard drinks. The participants also indicated that they drank on average 5.18 (SD = 4.70) standard drinks on their highest drinking occasion in the past 30 days.

3.4 Structural equation models

The first study aim examined the relationship between IPV perpetration and ED symptomology. The model fit the data, χ2 (n = 1588, 14) = 27.29, CFI = 0.99, TLI = 0.98, RMSEA = 0.02. However, IPV perpetration was not significantly associated with ED symptoms (β = 0.07, p > 0.05). See Fig. 1.

Fig. 1
figure 1

Structural Equation model between IPV perpetration and ED symptoms, Examining the associations between ED, CLASS, Alcohol Use, IPV Perpetration. ED SCOFF eating disorder screener, IPV Intimate Partner Violence Perpetration, CLASS College Alcohol Salience; Women (man); *p < 0.05; **p < 0.01; ***p < 0.001

Full size image

The second study aim examined the relationship between IPV perpetration, alcohol use, and alcohol salience. The model fit the data well, χ2 (n = 1588, 33) = 43.19, p = 0.11, CFI > 0.99, TLI > 0.99, RMSEA = 0.01. IPV perpetration was significantly associated with alcohol use (β = 0.13, p < 0.01) and alcohol salience (β = 0.11, p < 0.01). See Fig. 2.

Fig. 2
figure 2

Structural Equation model between IPV perpetration, ED symptoms, and Alcohol Use, Examining the associations between ED, CLASS, Alcohol Use, IPV Perpetration. ED SCOFF eating disorder screener, IPV Intimate Partner Violence Perpetration, CLASS College Alcohol Salience; Women (man); *p < 0.05; **p < 0.01; ***p < 0.001

Full size image

The final study aim examined the relationship between IPV perpetration, alcohol use, alcohol salience, and ED symptomology. Multiple models (e.g., exploring each variable as a dependent and independent variable) were explored due to the cross-sectional nature of the data. Due to systematic missing data from the randomization of the measures (i.e., planned missingness), the sample also decreased for the final model (n = 1067). The final model presented is supported by the literature and fits the data well, χ2 (n = 1067, 40) = 54.26, p = 0.07, CFI > 0.99, TLI > 0.99, RMSEA = 0.02. See Fig. 3. However, alcohol salience, alcohol use, and IPV perpetration was not significantly associated with ED symptoms. There were also no significant indirect effects.

Fig. 3
figure 3

Structural Equation model between IPV perpetration, ED symptoms, Alcohol Use, and College Life Alcohol Salience, Examining the associations between ED, CLASS, Alcohol Use, IPV Perpetration. ED SCOFF eating disorder screener, IPV Intimate Partner Violence Perpetration, CLASS College Alcohol Salience; Women (man); *p < 0.05; **p < 0.01; ***p < 0.001

Full size image

Due to the literature supporting gender differences with regard to all of the constructs, we examined the final aim 3 model across participants who identified as a man or as a woman. The model fit the data, χ2 (n = 1067, 88) = 152.47, CFI = 0.97, TLI = 0.97, RMSEA = 0.04 indicating that the model varied across gender identity. The path from IPV perpetration to CLASS was significant for people who identified as a woman but was non-significant for people who identify as a man, indicating gender differences, which may be due to the higher proportion of women who reported perpetration. See Fig. 4.

Fig. 4
figure 4

Structural Equation model between IPV perpetration, ED symptoms, Alcohol Use, and College Life Alcohol Salience by Gender, Examining the associations between ED, CLASS, Alcohol Use, IPV Perpetration. ED SCOFF eating disorder screener, IPV Intimate Partner Violence Perpetration, CLASS College Alcohol Salience; Women (man); *p < 0.05; **p < 0.01; ***p < 0.001

Full size image

4 Discussion

This study’s purpose was to model the path between IPV perpetration and ED symptomatology among college students, taking into consideration the important correlates of alcohol use and salience of alcohol to the college experience. Our aims were to: (1) determine the association between IPV perpetration and ED symptomatology; (2) determine the relationship between alcohol use, alcohol salience to college life, and IPV perpetration; and (3) explore the indirect effects of IPV perpetration through both alcohol variables to ED symptomatology. Our first aim considered the association between IPV perpetration and ED symptomatology, hypothesizing a positive association between the two. Our second aim examined the connection between alcohol use and alcohol salience to college life with IPV perpetration, again hypothesizing a positive association. A third aim considered the relationship between IPV perpetration, alcohol use and salience, and ED symptomology. Our first hypothesis was not confirmed; although the model fit well, IPV perpetration was not significantly associated with ED symptomatology. Our second hypothesis was supported, with the model fitting the data well and IPV being significantly associated with alcohol use and alcohol salience. The findings from our third aim showed that although the model fit well, neither alcohol salience, alcohol use, nor IPV perpetration significantly predicted ED symptomatology.

There were several other findings of note in this study. First, college women reported perpetrating IPV more than men, which is consistent with research that suggests women tend to report perpetrating violence more often than men [17, 36, 37]. This finding should be considered within the context of the measures used in this study, which were focused on specific behaviors and did not inquire about the overarching context, including any power and control dynamics, that may or may not be present in relationships where IPV perpetration occurs. For example, it is possible some IPV perpetration reported by participants was retaliatory or used in self-defense. It is also likely that some participants reported situational couple violence—physical violence that arises within the context of a specific argument or situation and that is not nested within a larger controlling pattern in the relationship [38, 39]. There is also an indication that current measures of IPV perpetration do not fully incorporate harmful gender norms and practices, which may inhibit accurate estimates and understanding of factors related to perpetration [40].

Second, although the relationship between alcohol use and IPV perpetration is well established [10,11,12, 41]. We extended the literature by linking perpetration to the salience of alcohol to the college experience. Participants in the current study who reported higher levels of perpetration also believed that alcohol was more central to their college experience. This may indicate a belief in alcohol-fueled social interactions, although additional research is needed to explore this possibility. Finally, although the effect size was small, the invariance test suggests that the relationship between perpetration and CLASS varies across gender. However, the invariance test is consistent with our global findings—and the extant literature—suggesting that women tend to report perpetrating violence in relationships more than men [17, 36, 37].

Finally, ED symptomatology was not associated with perpetration, alcohol use, or alcohol salience. This may be due to a measurement limitation with the SCOFF, which is a screener and not a more in-depth questionnaire for symptoms or diagnosis. Additionally, this only accounts for symptoms that fall in line with AN or BN diagnoses, rather than reflecting a broad range of ED symptoms. Given the strong comorbidity between EDs and alcohol abuse [42], future research might consider other measures of EDs, such as the Eating Pathology Symptom Inventory (EPSI), which is a more comprehensive approach on symptomatology associated with a range of EDs [43]. Previous research has shown a significant relationship between IPV victimization and ED symptomatology in this same population; thus, it may be that victimization is more strongly tied to ED symptomatology than perpetration [31]. This may be due to the increasing self-focused attention that is seen in some eating disorders, particularly anorexia nervosa [44], which may contribute to greater self-criticism but may also reduce focus on others. These relationships are still not fully understood and further exploration is needed.

It is important to note study limitations. First, there was a low prevalence of IPV perpetration reported, so we combined both physical and verbal perpetration into one category. Using a community-based sample with a higher prevalence of perpetration would be useful to do in the future in order to be able to discern the individual effects of verbal and physical perpetration. Second, the study utilized self-report data, which are susceptible to biases. Third, temporal relationships cannot be inferred from our cross-sectional study design. Fourth, we assessed the perpetration behavior, but because we do not understand the context surrounding that behavior, there could be conflation between IPV, situational couple violence, retaliatory violence, and self-defense. Fifth, diversity is limited in some ways in this sample due to the lack of racial and age diversity. Sixth, due to the length of the survey, certain scales were only given to a random subset of the participants. This process leads to some planned missingness. However, the sample is still sufficiently powered for the exploratory analysis. Finally, we do not have data about the victims of the perpetrators, which would be useful to a more thorough understanding of the context surrounding the violence.

5 Implications

5.1 Future research

This preliminary study offers insight into the association between alcohol use, IPV perpetration, and ED symptomatology, as well as a unique variable, the salience of alcohol to the college experience. Future research should examine this relationship at various college environments to determine if the salience of alcohol to the college environment changes in different contexts. This would also be important to understand for generalizability purposes. It would also be useful to explore this area using an IPV measure that captures the context behind perpetration violence. In addition, future research might consider incorporating measures of mental health. Previous literature suggests that mental health is associated with ED symptoms [45], IPV, and alcohol use [16]. Additionally, since this interrelationship represents an emerging field, it would be critical to conduct qualitative or mixed methods research to further explore this from a lived experience. Finally, given the association between victimization and perpetration, future studies might explore the overlap of these experiences in relation to the alcohol and ED symptoms.

6 Future interventions

Translating these findings into areas of prevention and intervention is a key element to ensure it helps improve public health. Some recommendations for future interventions should be to integrate intervention efforts rather than treating addressing as siloes. Given that college students face many demands on their time and have competing interests, the interventions that will be most effective will be those that address multiple risk factors or negative health outcomes. Instead of a targeted IPV intervention, a more comprehensive intervention addressing personal safety and health could be utilized. Additionally, these findings indicate that alcohol use and its salience to college are strongly associated with IPV perpetration. This gives more incentive for universities to work with collegiate recovery centers to sponsor more sober events and social opportunities that give students an alternative to alcohol-fueled social gatherings.