Problem Drinking and Marijuana Use as Risks for Unidirectional and Bidirectional Partner Violence.

Abstract

Despite evidence that most who perpetrate intimate partner violence (IPV) also report victimization, little is known about bidirectional IPV among Emergency Department (ED) patients and its association with problem drinking and marijuana use. We conducted an observational, cross-sectional survey among low- and moderate-acuity patients at a Northern California safety-net ED. Physical IPV was measured with the Revised Conflict Tactics Scale (CTS2). We recorded patient’s frequency of intoxication and marijuana use. Spouse/partner’s problem drinking and marijuana use were measured dichotomously. Odds Ratios [ORs] and 95% confidence intervals [CIs] were estimated using multinomial logistic regression models of unidirectional and bidirectional IPV. Among 1,037 patients (53% female), perpetration only, victimization only, and bidirectional IPV were reported by 3.8%, 6.2%, and 13.3% of the sample, respectively. Frequency of intoxication was associated with perpetration (OR 1.50; 95% CI 1.18 to 1.92) and bidirectional IPV (OR=1.34; 95% CI 1.13 to 1.58). Days of marijuana use were associated with bidirectional IPV (OR=1.15; 95% CI 1.03 to 1.28). Patients whose partners were problem drinkers were at risk for victimization (OR=2.56; 95% CI=1.38, 4.76) and bidirectional IPV (OR=1.97; 95% CI 1.18, 3.27). Among patients who reported any past-year IPV, most experienced bidirectional aggression. ED staff should consider asking patients who are married, cohabiting, or in a dating relationship about their experience with past-year IPV and inquire about their substance use patterns and those of their romantic partner, to share information about potential linkages. Medical and recreational marijuana legalization trends underscore the importance of further research on IPV and marijuana.

INTRODUCTION

Intimate partner violence (IPV) remains an international public health issue. Estimates from the 2004-2005 U.S. National Epidemiologic Study on Alcohol and Related Conditions (NESARC) indicate that among respondents involved in a past-year romantic relationship, the prevalence of IPV victimization, perpetration and bidirectional (reciprocal) violence for women were 1.6%, 3.1%, and 3.9% respectively. For men, the rates were 2.7%, 1.1%, and 3.1% (McMahon et al., 2015). Among young adult participants in a U.S. national longitudinal study, nearly 24% reported past-year IPV; most of these violent incidents were bidirectional (Whitaker, Haileyesus, Swahn, & Saltzman, 2007). IPV-involved men and women are at risk for numerous mental and physical health problems. For example, those who experience IPV are more likely to develop an Axis I psychiatric disorder (e.g., any substance use or mood disorder) than those who have not experienced IPV victimization (Okuda et al., 2011).

Prior studies show that urban emergency department (ED) patients have elevated rates of IPV compared to those in the general household population. Among male and female patients screened in ED-based studies, rates of past-year IPV ranged from 8.7% to 37% (Rhodes et al., 2009; Walton et al., 2009). The observed higher IPV prevalence among patients who present at urban safety-net EDs can be partially explained by their social disadvantage and other characteristics of the patient population. Compared to the general population, urban ED patients have increased rates of substance use problems, unemployment, and depression (Booth et al., 2011; Rhodes et al., 2009), and are more often exposed to aspects of the social environment that are linked with IPV risk, such as neighborhood poverty (Cunradi, Mair, Ponicki, & Remer, 2012). The heightened prevalence of IPV and co-occurrence of related problems among ED patients underscores the importance of conducting ED-based research to identify those at risk, and to formulate and test interventions aimed at reducing these health disparities.

Alcohol use features prominently as a contributor to both the occurrence and severity of partner aggression (Leonard & Quigley, 2017). A meta-analysis conducted by Foran and O’Leary (2008) found a small to moderate effect size for the association between alcohol use or abuse and male-perpetrated IPV, and a small effect size for female-perpetrated IPV. Results from the international GENACIS study indicate that self-reported IPV severity was significantly higher for incidents in which one or both partners had been drinking. These findings were consistent for men and women and across respondents from 13 countries (Graham, Bernards, Wilsnack, & Gmel, 2011). A potential mechanism underlying the alcohol-IPV association is disinhibition that occurs via the psychopharmacologic effects of alcohol on cognitive processing. The alcohol myopia model (Steele & Joseph, 1990) proposes that the pharmacological properties of alcohol cause individuals to narrow their focus on the most salient aspects of their environment. When these environmental cues are negative, alcohol will increase this myopic effect. Thus, under conditions of provocation or conflict, IPV may be facilitated by focusing attention on the most salient (e.g., provocative) aspects of the situation, and not to less salient (e.g., inhibitory) signals (Giancola, 2002). Personality factors such as impulsivity may further exacerbate the impact of alcohol on aggression (Mair, Cunradi, & Todd, 2012).

While the alcohol-IPV research evidence is strong, findings for the association between marijuana use and IPV have been inconsistent. For example, a meta-analysis found a small effect size (d=.22) for the association between marijuana and partner violence (Moore et al., 2008). A subsequent 9-year longitudinal study among a community-based sample of newlywed couples found that more frequent marijuana use was inversely related to IPV perpetration (Smith et al., 2014). Wives’ marijuana use, however, predicted more frequent female-to-male IPV perpetration among wives who had perpetrated IPV prior to marriage. Additionally, among a sample of those arrested for domestic violence, Stuart et al. (2013) found that women were less likely to perpetrate IPV on days in which marijuana was used. The mechanisms underlying these associations remain poorly elucidated, especially distal versus proximal effects of cannabis use on partner aggression (Testa & Brown, 2015). Shorey and colleagues (2018) suggest that the mechanisms underlying the alcohol-IPV association as explained by the alcohol myopia model may be similar to the processes for marijuana-related IPV.

The purpose of this study is to determine the associations of problem drinking and marijuana use with risk for unidirectional (perpetration or victimization only) and bidirectional (perpetration and victimization) IPV among a sample of urban ED patients. Understanding these relationships are important for several reasons. First, there is considerable evidence from national and community-based surveys that a large proportion of those who report IPV perpetration also report IPV victimization (Langhinrichsen-Rohling, Misra, Selwyn, & Rohling, 2012). Bidirectional IPV is associated with greater severity and injury (Whitaker et al., 2007), yet few ED-based studies have sought to estimate its prevalence (Bazargan-Hejazi et al., 2014). It is unknown, for example, whether patterns observed among general population samples in which most IPV-involved study participants report both perpetration and victimization will be observed among a safety-net ED-based sample. Although most IPV-related injuries seen in the ED are overwhelmingly among female patients (Davidov, Larrabee, & Davis, 2015), the prevalence of unidirectional and bidirectional IPV among low-income patients seeking primary care at urban EDs may resemble the prevalence seen in the general population. Second, trends towards legalization of medical marijuana and recreational marijuana underscore the importance of examining the contribution of marijuana use to public health problems (Wilkinson, Yarnell, Radhakrishnan, Ball, & D'Souza, 2016), such as IPV. Towards this end, substance use researchers have advocated for additional studies that focus on the association between marijuana and IPV, and the combined impact of alcohol and marijuana on IPV (Shorey, Haynes, Strauss, Temple, & Stuart, 2017; Testa & Brown, 2015). Lastly, it is important to account for the spouse/partner’s problem drinking and marijuana use as these behaviors could serve as a source of couple conflict for the dyad and thereafter increase the likelihood of IPV (Cunradi, Todd, & Mair, 2015). Because the ED visit can present an opportunity for a “teachable moment” in which behavior change may be more likely to occur (Bernstein & D'Onofrio, 2009), quantifying the extent to which problem drinking and marijuana use are associated with unidirectional and bidirectional IPV can aid in identifying those at risk and could advance prevention and treatment efforts in ED settings.

The current study builds upon findings from previous ED-based IPV studies (Lipsky, Caetano, Field, & Bazargan, 2005; Rhodes et al., 2009; Sutherland, Fantasia, & McClain, 2013; Walton et al., 2009), community-based studies of reciprocal and non-reciprocal IPV (Caetano, Ramisetty-Mikler, & Field, 2005; Melander, Noel, & Tyler, 2010; Reingle Gonzalez, Connell, Businelle, Jennings, & Chartier, 2014; Whitaker et al., 2007), and the conceptual framework developed through our research (Cunradi, Ames, & Duke, 2011; Mair et al., 2012). Our conceptual framework is informed by IPV-related health disparities (e.g., racial/ethnic and socioeconomic status differences in IPV prevalence) (Cunradi, Caetano, & Schafer, 2002b), the contribution of childhood trauma and its sequelae (Oral et al., 2016), and the role of alcohol-related problems and drug use (Cunradi, Caetano, & Schafer, 2002a). We therefore investigated how demographic characteristics (gender, marital status, education, race/ethnicity, survey language, age, and food insufficiency), psychosocial variables (depression, impulsivity, PTSD, adverse childhood experiences) and substance use factors (patient’s frequency of intoxication and days of marijuana use, spouse/partner’s problem drinking and marijuana use) may differ according to the type of violence involvement (i.e., victimization only, perpetration only, bidirectional IPV) among an urban ED sample. We hypothesized that (1) frequency of intoxication and days of marijuana use would be associated with each type of IPV; (2) spouse/partner’s problem drinking and past-year marijuana use would be associated with each type of IPV; and (3) psychosocial variables would be associated with each type of IPV. We did not have a priori hypotheses as to how the strength of the associations would differ among each type of IPV. Finally, we expected male gender to be inversely associated with perpetration only, and positively associated with victimization only. We also expected non-white race/ethnicity and food insufficiency to be positively associated with each type of IPV, and age to be inversely related.

METHOD

Participants

This cross-sectional, observational study is based at a single ED of an urban Level I trauma center in Northern California. The hospital is part of a county-wide integrated public health care system. The ED has an annual caseload of 72,000, and serves as the county’s safety-net provider, with 61% of visits covered by Medicaid, and another 17% uninsured. Approximately 41% of ED patients are African American, and 33% are Hispanic. Study inclusion criteria were: 18-50 years old; English or Spanish speaker; resident of the county in which the hospital is located; and married, cohabiting, or in a romantic (dating) relationship for the past 12 months. Patients who were intoxicated, experiencing acute psychosis or suicidal or homicidal ideation, were cognitively/psychologically impaired and unable to provide informed consent, in custody by law enforcement, or in need of immediate medical attention (i.e., Emergency Severity Index [ESI] (Agency for Healthcare Research and Quality, 2014) level 1 or 2) were excluded. Males comprised nearly 47% of the study sample. Approximately 50% of participants were Hispanic, and 29% were African American. Mean age was 35.2 years (SD 8.5). Nearly one third had not completed high school; less than 10% had graduated from college. Almost half reported that they sometimes or often ran out of food during the past 12 months and didn’t have enough money to get more.

Procedure

The project was approved by the hospital’s Institutional Review Board. A team of bilingual, B.A.-level Research Assistants (RAs) pilot tested the survey with 41 non-emergent patients. The purpose of the pilot test process was to identify obstacles to study recruitment, refine data collection procedures, and provide the research team with estimates of average survey interview length. Following completion of the pilot testing, data collection with the finalized survey instrument was conducted from February 27^th through December 15^th, 2017. Staffing constraints prevented us from proportionately recruiting patients from all ED shifts. Instead, 2 interviewers per shift staffed the ED during weekday peak volume hours (9am – 9pm) to recruit eligible patients to the study. The RAs identified potentially eligible patients through a multi-step process (Figure 1): 1. They searched the ED’s electronic patient information system (Wellsoft™) for currently registered ED patients ages 18-50 (n=3,386) who had been triaged at ESI levels 3-5. 2. They located and conducted face-to-face screening with patients in the ED waiting room or in a treatment cubicle (n=2,212). 3. They offered eligible patients the opportunity to participate in a confidential, face-to-face survey interview for which they would receive a $30 grocery store gift card incentive (n=1,184). Informed consent was obtained in a private area adjacent to the ED waiting room, or in the patient’s room without others present (n=1,066). Twenty-nine patients terminated the survey interview before completion, usually due to interruption for medical services (e.g., patient transported to ultrasound or X-ray). Thus, 1,037 patients (53% female) completed the survey interview (87.5% participation rate). Patient survey data was collected using computer assisted personal interview (CAPI) techniques with tablet computers running the Qualtrics platform.

Figure 1:

Study sample recruitment

Measures

Unidirectional and Bidirectional Intimate Partner Violence:

Past-12 month physical IPV was measured with the 12-item physical assault subscale in the Revised Conflict Tactics Scale (Straus, Hamby, Boney-McCoy, & Sugarman, 1996). Patients were asked about violent behaviors they may have perpetrated against their spouse/partner, and that their spouse/partner may have perpetrated against them. This allows for identification of the patient as a perpetrator of violence, a victim of violence, or in a relationship with bidirectional violence. Cronbach’s α for the scale in the dataset under analysis was .85.

Frequency of intoxication:

We asked patients how often they had any kind of alcoholic beverage in the last 12 months. Those who consumed 1 or more drinks were asked additional questions about their alcohol consumption. To measure frequency of intoxication, they were asked, “During the past 12 months, about how many times did you drink enough to feel intoxicated or drunk, that is, when your speech was slurred, you felt unsteady on your feet, or you had blurred vision?” This question has been used in previous studies of IPV and drinking (Cunradi, Mair, Todd, & Remer, 2012).

Marijuana use:

We asked patients, “How many times during the past 12 months, or 365 days, did you use marijuana or hashish (weed, pot, hash) without a doctor's instruction?” Amount of marijuana use was not measured. Days of marijuana use (0-365) were recorded. We also asked, “During the past 12 months, or 365 days, did your spouse or partner use marijuana or hashish (weed, pot, hash) without a doctor's instruction (yes/no)?” A dichotomous variable was created, coded ‘1’ for those who reported any past-year marijuana use and ‘0’ for those with no past-year use.

Partner problem drinking:

We used the 3-item AUDIT-C (Alcohol Use Disorders Identification Test-Consumption) to measure the patient's assessment of his/her spouse/partner's drinking (Bradley et al., 2003; Bush, Kivlahan, McDonell, Fihn, & Bradley, 1998). The three questions cover how often the partner drinks, how many standard drinks the partner has on a typical day, and how often the partner has 6 or more drinks on one occasion. Sum of scores in the 3 items range from 0 to 12, and male/female partners with a score above 4 or 3, respectively, were considered hazardous drinkers. Cronbach’s α = .81.

Adverse Childhood Experiences (ACE):

The modified ACE measures exposure to six adverse experiences the patient may have had "while they were growing up during their first 18 years of life:" (1) exposure to a mentally ill person in the home; (2) parent/caregiver alcoholism; (3) sexual abuse; (4) physical abuse; (5) psychological abuse; and (6) violence directed against the respondent's mother (Cabrera, Hoge, Bliese, Castro, & Messer, 2007). These six exposures were summed to create the ACE variable (range=0-6; Cronbach’s α = .74).

Impulsivity:

This construct was measured with a 3-item scale that originated in the National Alcohol Survey (Schafer, 1994), and has been used in prior IPV studies (Caetano, Cunradi, Schafer, & Clark, 2000; Cunradi, Todd, Duke, & Ames, 2009). Items were scored from 1-4, with a higher score representing greater impulsivity, and averaged to create a composite score. Cronbach’s α = .79.

Depression:

We measured depression with 7 items from the Hospital Anxiety and Depression Scale (Zigmond & Snaith, 1983). Items were scored from 1-4, with a higher score representing greater depression, and averaged to create a composite score (Cronbach’s α = .69). Following Brennan et al. (2010) a score greater than 8 was used as the cutoff point to identify positives. This cut point gives sensitivity of .82, and specificity of .74, for depression.

Post-Traumatic Stress Disorder (PTSD):

This 4-item measure is from the Primary Care Screener for PTSD (Prins et al., 2003). Answers were coded yes/no; a score of 3 or more is considered positive. Cronbach’s α = .83.

Sociodemographic factors:

Gender. We asked patients to self-report their gender. Three patients identified as transgender; due to small numbers, these cases were excluded from the analysis. We then created a dichotomous variable coded as male and female (reference category). Race/ethnicity: Patients were asked to name the racial or ethnic group(s) that best describes them from the following list: Native American Indian or Alaska Native; Filipino; Asian (not including Filipino); Black, African American; Latino, Hispanic; Native Hawaiian or other Pacific Islander (not including Filipino); White, Caucasian; some other race (specify). Those who selected more than one category were categorized as multiethnic. For the analyses, these groups were recoded into a 5-category race/ethnicity variable: Hispanic; African American; multiethnic; other; and white (reference category). Age. Patient age was used as a continuous variable. Level of education. We asked patients about the highest level of education they completed, which we coded into four education categories: a) less than high school; b) completed high school or GED; c) some college or technical or vocational school; d) completed 4-year college or higher (reference category). Marital status: Patients were asked to describe their current marital status (married; living with partner; single; separated; divorced; widowed). This was recoded into a 3-category variable: a) married; b) living with partner; c) single, separated or divorced (reference category). Survey language: The patient’s survey language preference (English/Spanish) was recorded. Household food insufficiency: Patients were asked their level of agreement with the statement, “In the past 12 months, the food we bought ran out and we didn’t have money to get more.” Response categories were: never; sometimes true; often true. In accord with Okechukwu et al. (2012) we dichotomized and compared those who responded “sometimes” or “often” to those who responded “never” (reference category).

Data Analytic Strategy

The study’s initial sample size estimate called for the enrollment of 800 married, cohabiting, or dating adults (50% female). This was based on calculations that using linear regression analyses, power would be 80% to detect a small overall effect (R² = .02) with 20 predictors, α = .05. Power would be 85% to detect small incremental changes of adding single variables to the regression equations (ΔR² = .01) with 19 prior predictors, a prior R² of .10, and α = .05.

Analyses were conducted with IBM SPSS Statistics v. 25. We calculated frequencies for categorical variables and means and standard deviations for continuous variables. We estimated Odds Ratios [ORs] and 95% Confidence Intervals [CIs] with multinomial regression models using SPSS’s “nomreg” procedure. The dependent variable had 4 categories: perpetration only, victimization only, bidirectional violence, and no violence (reference category). Models included substance use, demographic and psychosocial variables. Abstainers (those who reported no past-year drinking) and drinkers (those who had at least one drink in the past year) were included in the analysis. A recoded ‘frequency of intoxication’ variable was created in which abstainers were coded as ‘0,’ and the intoxication frequency values for all patients were log transformed due to skewed distribution. A new ‘days of marijuana use’ variable was created in which non-users were coded as ‘0,’ and the days of marijuana use values for all patients were log transformed due to skewness. Missing data ranged from 0 – 1.6% for the variables in the study and were dropped from the analysis through listwise deletion. The multinomial analysis is based on complete data from 449 males and 515 females.

RESULTS

About 23% of the sample reported either unidirectional or bidirectional IPV (Table 1). Among those reporting any IPV (n=241), 16.2% reported perpetration, 26.5% reported victimization, and 57.3% reported bidirectional IPV. Bivariate chi-square analysis (type of IPV by gender; X²=24.58, 3 df; p < 0.001) showed significant gender differences in IPV type. A greater proportion of females than males reported perpetration (6.0% vs. 1.2%), and a greater proportion of males reported victimization (8.8% vs. 4.0%). There were no gender differences in the proportions of those reporting mutual IPV.

Table 1.

Sample characteristics (N=1037)

	% or M, SD
Gender:
Male	46.7
Female	53.0
Missing (3)*	0.3
Marital status:
Married	40.7
Cohabiting	31.4
Single, separated, divorced	27.6
Missing (3)	0.3
Education:
Less than high school	32.5
High school graduate/GED	35.4
Some college	21.5
College graduate+	9.0
Missing (17)	1.6
Race/ethnicity:
Hispanic	50.1
African American	28.8
Multiracial	5.4
Other	9.1
White	6.6
Missing (0)	0.0
Survey language:
English	64.8
Spanish	35.2
Missing (0)	0.0
Age (range 18-50; missing=0)	35.2 (8.5)
Food Insufficiency:
Sometimes/often	49.7
Never	49.8
Missing (6)	0.6
Depression screen
Positive	17.0
Negative	82.9
Missing (1)	0.1
Impulsivity score (range 3-12; missing=3)	5.3 (2.5)
PTSD screen
Positive	25.1
Negative	74.7
Missing (2)	0.2
Adverse childhood experiences (range 0-6; missing=2)	1.3 (1.5)
Past-year frequency of intoxication (range 0-365; missing=13)	8.3 (39.9)
Past-year days of marijuana use (range 0-365; missing=12)	51.9 (119.4)
Partner’s AUDIT-C
Positive	21.2
Negative	78.8
Missing (0)	0.0
Partner’s past-year marijuana use
Yes	22.1
No	76.4
Missing (16)	1.5
Past-year intimate partner violence:
None	76.0
Perpetration only	3.8
Victimization only	6.2
Bidirectional	13.3
Missing (8)	0.8

^*3 patients identified as transgender and were excluded from the analysis due to small numbers.

Multinomial regression results (Table 2) show that compared to females, males were less likely to report perpetration (OR 0.22; 95% CI 0.08 to 0.58), and more likely to report victimization (OR 3.25; 95% CI 1.75 to 6.04). Compared to white patients, African American and multiracial patients were at elevated risk for bidirectional IPV (ORs 2.81 and 3.67, respectively). Age was inversely associated with risk for bidirectional IPV (OR 0.96; 95% CI 0.93 to 0.99). Those who reported experiencing food insufficiency were more than twice as likely to report bidirectional IPV compared to those who didn’t experience food insufficiency (OR 2.62; 95% CI 1.59 to 4.31). There was no association between marital status, education, or survey language and risk for any type of IPV involvement.

Table 2.

Multinomial logistic regression results

	Perpetration only vs. no IPV	Victimization only vs. no IPV	Bidirectional IPV vs. no IPV
	OR (95% CI)	OR (95% CI)	OR (95% CI)
Gender:
Male	0.22 (0.08, 0.58)**	3.25 (1.75, 6.04)***	1.37 (0.84, 2.21)
Female (ref.)	---	---	---
Marital status:
Married	1.36 (0.51, 3.63)	0.55 (0.25, 1.20)	1.17 (0.64, 2.14)
Cohabiting	1.29 (0.54, 3.07)	0.94 (0.49, 1.80)	1.13 (0.66, 1.94)
Single, separate, divorced (ref.)	---	---	---
Education:
Less than high school	4.32 (0.46, 40.24)	1.29 (0.37, 4.49)	1.49 (0.54, 4.12)
High school graduate/GED	5.40 (0.65, 44.86)	1.39 (0.43, 4.54)	1.48 (0.56, 3.91)
Some college	4.00 (0.48, 33.55)	1.29 (0.38, 4.37)	1.72 (0.65, 4.59)
College graduate+ (ref.)	---	---	---
Race/ethnicity:
Hispanic	0.97 (0.23, 4.09)	0.87 (0.27, 2.79)	2.11 (0.72, 6.21)
African American	1.42 (0.40, 5.01)	0.91 (0.32, 2.57)	2.81 (1.06, 7.47)*
Multiracial	1.37 (0.28, 6.65)	1.32 (0.35, 5.01)	3.67 (1.15, 11.65)*
Other	1.05 (0.16, 6.91)	0.92 (0.24, 3.62)	3.00 (0.91, 9.92)
White (ref.)	---	---	---
Survey language:
Spanish	0.36 (0.09, 1.42)	0.66 (0.25, 1.70)	0.55 (0.25, 1.22)
English (ref.)	---	---	---
Age	0.97 (0.92, 1.01)	0.98 (0.95, 1.01)	0.96 (0.93, 0.99)**
Food Insufficiency:
Sometimes/often	1.18 (0.54, 2.56)	1.77 (0.99, 3.16)	2.62 (1.59, 4.31)***
Never (ref.)	---	---	---
Depression screen	0.63 (0.24, 1.68)	1.18 (0.59, 2.38)	1.34 (0.79, 2.28)
Impulsivity	1.21 (1.05, 1.38)**	1.13 (1.01, 1.26)*	1.17 (1.07, 1.28)***
PTSD screen
Positive	1.54 (0.69, 3.42)	2.03 (1.09, 3.79)*	1.67 (1.02, 2.74)*
Negative (ref.)	---	---	---
Adverse childhood experiences	1.30 (1.04, 1.62)*	0. 98 (0.80, 1.20)	1.21 (1.04, 1.39)*
Frequency of intoxication (logged)	1.51 (1.18, 1.93)**	1.00 (0.78, 1.27)	1.33 (1.13, 1.57)**
Days of marijuana use (logged)	1.06 (0.90, 1.26)	1.04 (0.90, 1.19)	1.15 (1.03, 1.28)*
Partner’s AUDIT-C:
Positive	1.10 (0.47, 2.61)	2.53 (1.36, 4.70)**	1.92 (1.16, 3.20)*
Negative (ref.)	---	---	---
Partner’s marijuana use:
Yes	0.82 (0.33, 2.00)	1.47 (0.71, 3.03)	1.56 (0.89, 2.72)
No (ref.)	---	---	---

^*p<0.05

^**p<0.01

^***p<0.001

Regarding psychosocial factors, impulsivity was positively associated with risk for all types of IPV, with similar ORs (range 1.13-1.21). Patients who screened positive for PTSD were more likely to report victimization (OR 2.03; 95% CI 1.09 to 3.79) and bidirectional IPV (OR 1.67; 95% CI 1.02 to 2.74) than those with a negative PTSD screen. Adverse childhood experiences were positively associated with risk for perpetration (OR 1.30; 95% CI 1.04 to 1.62) and bidirectional IPV (OR 1.21; 95% CI 1.04 to 1.39). Those with a positive depression screen were not at increased risk for any type of IPV compared to those without.

In terms of substance use, the patient’s frequency of intoxication was associated with perpetration (OR 1.51; 95% CI 1.18 to 1.93) and bidirectional IPV (OR 1.33; 95% CI 1.13 to 1.57). Days of marijuana use were associated with bidirectional violence (OR 1.15; 95% CI 1.03 to 1.28). Patients whose partners were problem drinkers were more likely to report victimization (OR 2.53; 95% CI 1.36 to 4.70) and bidirectional violence (OR 1.92; 95% CI 1.16 to 3.20) compared to those whose partners were not problem drinkers. Partner’s marijuana use was not associated with any type of IPV.

DISCUSSION

Among a diverse, low-socioeconomic status urban ED population, a substantial proportion of the sample (about 23%) reported past-year IPV involvement; among these patients, most experienced bidirectional IPV. This is of concern since there is evidence that bidirectional IPV is associated with greater injury compared to unidirectional IPV (Whitaker et al., 2007). Our hypothesis that frequency of intoxication and days of marijuana use would be associated with each type of IPV was partially confirmed. Specifically, the findings showed that the patient’s frequency of intoxication was associated with risk for perpetration and bidirectional IPV; their days of marijuana use was associated with bidirectional IPV. Similarly, our hypothesis that the spouse/partner’s problem drinking and past-year marijuana use would be associated with each type of IPV was only partially confirmed. Here the findings showed that spouse/partner problem drinking was associated with victimization and bidirectional IPV; no associations were seen for past-year marijuana use. ED providers should consider asking patients who are married, cohabiting, or in a dating relationship about their experience with past-year IPV perpetration and victimization, and inquire about their substance use patterns, as well as those of their spouse or romantic partner, to share information in a non-judgmental way about the potential linkages. Although the findings showed no association between the partner’s marijuana and any IPV involvement, this may be a result of measuring past-year marijuana use in a dichotomous manner.

Our hypotheses concerning the associations between male gender and victimization and perpetration were confirmed. The analysis found that compared to female patients, male patients were less likely to report perpetration, more likely to report victimization, and neither more nor less likely to report bidirectional IPV. Interestingly, these findings mirror those reported by Melander and colleagues (2010) regarding gender differences in risk for unidirectional and bidirectional IPV among a large sample of young adults. Based on a multinomial analysis, they found that females were more likely than males to report perpetration, and less likely to report victimization; no gender differences were seen for bidirectional IPV. Similarly, in a multinomial analysis of 2004-5 NESARC data, Reingle et al. (2014) found that compared to females, males were more likely to report victimization, less likely to report perpetration, and less likely to report bidirectional IPV. In contrast, Bazargan-Hejazi et al. (2014) reported no bivariate gender differences regarding IPV type among an ED sample; their study, however, did not contain a multivariable analysis. The findings underscore the importance of ED providers asking both male and female patients about their experience with partner aggression. It should not be assumed that males will always be the perpetrators and females will be the victims.

In terms of race/ethnicity, our hypothesis concerning greater IPV risk for non-white patients was partially confirmed. The findings indicate that African American and multiracial patients were at elevated risk for bidirectional IPV compared to white patients even after adjusting for many sociodemographic and psychosocial factors. Previous non-ED population-based studies have found that African Americans are approximately twice as likely to report bidirectional IPV than whites (Melander et al., 2010), and more likely than whites to report unidirectional IPV (Reingle et al., 2014). These results suggest that other unmeasured factors may mediate the association between race/ethnicity and IPV observed in the current sample and warrants further study. Our hypothesis regarding the inverse association between age and each type of IPV was confirmed only for bidirectional IPV.

Food insufficiency, sometimes referred to as food insecurity (e.g., concern about not having enough to eat, food not lasting, need to cut or skip meals, going hungry), has been identified as a significant IPV risk factor (Schwab-Reese, Peek-Asa, & Parker, 2016). To our knowledge this is the first study to report an association between food insufficiency and bidirectional IPV among an ED sample. Thus, our hypothesis regarding food insufficiency and each type of IPV was partially confirmed. The mechanism underlying this finding remains unknown. A potential explanation is that lack of food and its attendant stress contributes to partner conflict and thereafter increase the likelihood of physical aggression. Clinically, it may be important for ED providers to inquire about whether the patient is experiencing food insufficiency, and if so, make appropriate referrals that offer short- and long-term solutions. Future IPV prevention programs or recruitment efforts for IPV intervention studies may want to target food banks or soup kitchens to reach those with food insufficiency who are at risk for IPV.

The results showed that most of the psychosocial factors included in the multinomial model were associated with unidirectional and bidirectional IPV, thus largely confirming our hypothesis. For example, ACE score was positively related to perpetration and bidirectional IPV. This is in accord with multinomial results reported by Melander et al (2010). Impulsivity was associated with greater likelihood of perpetration, victimization, and bidirectional IPV, findings that are in agreement with the results of a couples analysis reported by Cunradi et al. (2011). Those with a positive PTSD screen were more likely to report victimization and bidirectional IPV, results that are consistent with those reported by Reingle et al. (2014). Unexpectedly, no association was seen for depression, a finding that is in contrast with much of the IPV literature and studies that link depressive symptoms with bidirectional IPV (Melander et al., 2010; Reingle et al., 2014), victimization (Reingle et al., 2014), and IPV severity (Caetano, Cunradi, Alter, Mair, & Yau, 2019 Jan 31 [Epub ahead of print]).

Several limitations should be considered when interpreting the study’s findings. First, the cross-sectional design precludes making causal inferences regarding the study’s observed associations. Furthermore, the sample was obtained from a single Northern California urban ED, which may limit generalizability. Second, due to survey time constraints, no data were collected concerning psychological abuse, injury, or sexual coercion among the patients. Moreover, the spouse’s/partner’s past-year marijuana use was measured dichotomously, which doesn’t capture disordered use or dependence. The impact of this variable on IPV risk may therefore be underestimated in the analysis. Third, the patients’ spouses and romantic partners weren’t interviewed; lack of concurrent dyadic reports on the occurrence of IPV may result in an underestimation of IPV prevalence (Cunradi, Bersamin, & Ames, 2009). Lastly, recall bias may have affected patients’ estimation of events over the previous 12 months.

In summary, this study contributes to the literature by providing prevalence estimates for unidirectional and bidirectional IPV among a sample of urban ED patients and a multivariate analysis of its correlates. Among those who reported any past-year IPV, most reported bidirectional violence. This suggests that the IPV prevalence patterns seen among an at-risk urban ED sample are similar to that observed in the general population. In addition, the patient’s frequency of intoxication and days of marijuana use, and their spouse/partner’s problem drinking, are differentially linked with risk for these outcomes. The role of each partner’s substance use merits further investigation using longitudinal methods, such as ecological momentary assessment. In addition to substance use, the study identifies specific sociodemographic and psychosocial characteristics of ED patients that are associated with unidirectional and bidirectional IPV. Addressing food insufficiency among underserved ED populations may be an avenue of IPV prevention. ED providers should be aware that most patients who screen positive for any past-year IPV involvement may have experienced both perpetration and victimization.