Characteristics of Men Who Perpetrate Intimate Partner Violence

Sampling

We analyzed the 2001 to 2003 National Comorbidity Survey–Replication (NCS-R), which used a nationally representative, multistage, clustered sample of households. This probability sampling obtained respondents ages 18 or older living in the noninstitutionalized civilian household population of the United States.¹⁸ Harvard Medical School and the University of Michigan provided institutional review board approval for the original data collection.¹⁸ The University of Michigan institutional review board determined this current study to be not regulated because it is research using a publicly available data set.

We selected potential IPV perpetration risk domains based on prior findings^6⇓⇓–9 and included demographics (age, race, education, health insurance, employment, and income); mental illness (depression, posttraumatic stress, panic, and anxiety disorders); substance use history (abuse or dependence of alcohol or drugs); and prior family violence (IPV victimization from age 21, physical abuse as a child, and exposure to parental IPV). We selected physical symptoms or health conditions of chronic pain, epilepsy or seizures, injury, and headaches, which may cause men to seek primary care and have been associated with IPV perpetration in previous work.⁹ We chose IBS and insomnia symptoms given their association with IPV victimization.^14,15 For relevance to identification in clinical settings, we selected health services use measures that can describe primary care and ED utilization.

Intimate Partner Violence

Physical abuse measures from the Conflict Tactics Scale¹⁹ were adapted in the NCS-R to construct the IPV perpetration variable. In brief, the dependent variable IPV perpetration²⁰ was assessed by asking: “Over the course of your relationship, how often have you ever done any of these things (pushed, grabbed, or shoved; threw something; slapped or hit; kicked, bit, or hit with a fist; beat up; choked; burned or scalded; threatened with a knife or gun) to your current spouse/partner?” Responses included often, sometimes, rarely, or never. We dichotomized responses into any/none.

Experience with Violence as a Child

The Conflict Tactics Scale measures of perpetration of parent-to-child physical abuse²¹ were adapted for experience with violence as a child. Victim of childhood family violence was measured by the question: “When you were growing up, how often did someone in your household do any of these things (push, grab, or shove; throw something; slap or hit) to you?” Responses were often, sometimes, rarely, or never. Witnessed childhood family violence was assessed by asking: “How often did (your parents/the people who raised you) do any of these things (push, grab, or shove; throw something; slap or hit) to each other while you were growing up?” Responses were often, sometimes, rarely, or never. We dichotomized responses into any/none.

Demographics

We dichotomized race between non-Hispanic white and all other race and ethnicity categories. We dichotomized education at high school completion. We dichotomized health insurance and employment as any/none. We divided income in dollars by $10,000 increments.

Health Services Use and Physical Symptoms or Health Conditions

We assessed having a regular medical doctor for routine care and ≥1 outpatient or ED visit per year. Responses were dichotomized. Men were asked whether they ever had chronic pain from arthritis, back or neck problems, or any other source; epilepsy or seizures; or injury. IBS symptom was measured by the question, “Have you ever had a period lasting 12 months or longer when at least 1 week each month you had frequent pain or discomfort in your stomach or lower abdomen that was relieved when you had a bowel movement?” We dichotomized IBS symptoms into yes/no. Insomnia symptoms were measured by the question, “How often do you have trouble getting to sleep or staying asleep?” Headache symptoms were measured by the question, “How often are you bothered by headaches or pains in the head?” The response set for insomnia and headache symptoms were nearly all the time (1), pretty often (2), not very much (3), and never (4). We analyzed insomnia and headache responses using a mean frequency of positive report of symptoms.

Substance Use and Mental Health Diagnoses

Constructed variables using Diagnostic and Statistical Manual of Mental Disorders IV criteria²² assessed the following lifetime diagnoses: substance abuse/dependence (defined as alcohol abuse or dependence or drug abuse or dependence), major depressive disorder, PTSD, and anxiety disorders (defined as panic disorder or generalized anxiety disorder).²⁰ We chose these mental health variables because of IPV perpetration associations in previous research.^8,20 We dichotomized the responses.

Statistical Analysis

We conducted data analyses using STATA/SE software version 11 for Windows (StataCorp., College Station, TX). We analyzed age and income demographic variables, as well as insomnia and headache symptom physical health conditions, as continuous variables. Survey weights were applied in all analyses. First, we describe IPV perpetration among married or cohabiting men ≥21 years old. Second, we present demographic characteristics of and prior family violence among the full sample, as well as univariate characteristics of patients who do and do not perpetrate IPV (Table 1). We also present bivariate logistic regression associations evaluating participants who do and do not perpetrate IPV (Table 1). We chose multivariate logistic regression models (Table 2) based on which bivariate associations were significant and theoretically grounded while adjusting for demographics and health insurance. The first model included only demographics, health services use, and physical symptoms because most primary care physicians do not assess Diagnostic and Statistical Manual of Mental Disorders–based substance use diagnoses, nor do they routinely obtain history of victimization from or witnessing childhood family violence. The second model added substance abuse and dependence to evaluate whether the associations with physical symptoms remained significant. The third model added prior family violence to understand the relationship of substance abuse and history of family violence to IPV. No multicollinearity of variables was observed in any of the 3 models.