Synergistic Effects of Depression and Poor Impulse Control on Physical Partner Violence ; A National Study in the United States

Background and aims: Although major depressive disorder (MDD) also increases the risk of intimate partner violence (IPV) perpetration, poor mental health has been mostly conceptualized as a consequence of IPV victimization rather than a risk factor for IPV perpetration. Poor impulse control is another IPV risk factor. Building on a risk/risk framework, this study investigated additive and multiplicative effects of male partners’ MDD and poor impulse control on physical IPV victimization reported by female partners. Methods: This national longitudinal study followed a random sample of 2500 male and female cohabiting partners for 2 years in the United States. Data came from the Fragile Families and Child Wellbeing Study (FFCWS). Male partner’s MDD was the independent variable. Baseline socio-demographic factors (age, relation status, education level, income, and minority status) and generalized anxiety disorder (GAD), and Alcohol Use Disorder (AUD) were covariates. Outcome was female partner’s physical IPV victimization measured at baseline and 2 years later. Male partner’s impulse control was the moderator. Model I tested independent effects of MDD and impulse control on physical IPV. Model II used multi-group structural equation model to test the effect of MDD in those with low and poor impulse control. All participants provided written consent. SPSS and AMOS were used for data analysis. Results: Model 1 that tested additive effects of socioeconomic status (SES), GAD, AUD, MDD, and poor impulse control did not show main effects of MDD or impulse control on an increase in perpetration of physical IPV over 2 years. Model 2 showed that MDD predicts an increase in perpetration of physical IPV by men with poor impulse control, but not men with high impulse control. Conclusion: Based on the findings, poor impulse control and MDD have synergistic effects on IPV committed by men. Given the synergistic effects of psychological determinants of IPV, there is a need for prevention of IPV in male partners who have multiple risk factors such as MDD and poor impulse control. Self-regulation trainings of depressed men with poor impulse control may reduce IPV risk among men.


Introduction
Most previous studies have conceptualized depression as a consequence not a risk factor for intimate partner violence (IPV). 1,2Depression, however, may increase the risk of IPV perpetration, as well. 3,4Depression may predict IPV perpetration, independent of alcohol use disorder (AUD) and illicit drug use. 3In a study of about 2000 males and females between ages of 15 and 26, high depressive symptoms of men predicted IPV (adjusted odds ratio [AOR] = 1.50) and rape (AOR = 1.81) perpetration. 5In a national household sample of American couples, high depressive symptoms increased risk of male-to-female and female-to-male IPV perpetration. 6mong women, depression may more strongly correlate with IPV victimization than perpetration. 7mong men, however, depression may be a stronger predictor of IPV perpetration than victimization. 7lthough severe depression is linked to both onesided and bidirectional aggression, depression has a stronger link to one-sided aggression. 7Most of these studies, however, have done a cross-sectional study.More longitudinal research is needed on depression as a potential risk factor for IPV perpetration.

Assari
][10] Based on a study in California, high impulse control was associated with high rates of IPV, net of other risk factors such as problem drinking. 11Researchers have argued that disinhibition and poor impulse control may be a mechanism by which psychiatric disorders increase risk of partner violence. 12nfortunately, very few studies -if any -have tested multiplicative effects of poor impulse control and psychiatric disorders such as major depressive disorder (MDD) on IPV.We believe MDD and impulse control may have synergistic effects as IPV risk factors. 124][15] There is still a need to study multiplicative effects of risk factors of IPV.
Built on a risk/risk framework, this study studied how men's poor impulse control, MDD, and AUD interact in predicting perpetration of physical IPV over time.Similar to the risk and resilience model, [16][17][18] risk/risk model suggests that there is a need to study multiple risk and positive factors.Risk/risk framework suggests that the effect of one risk factor on the negative outcome may depend on the presence or absence of other risk or protective factors. 19Based on this framework, while resilient factors potentially interrupt paths from risk factors to undesired outcomes, presence of other risk factors may exacerbate their effects. 20here are 2 strategies to study risk-risk frameworks. 20n Model 1, risk factors operate independent of other risk factors.This model measures the additive effects of multiple risk factors. 21In Model 2, the second risk factor moderates the effect of first risk factor on outcome. 22This model conceptualizes a multiplicative effect among various risk factors. 23lthough risk-risk frameworks have been previously applied to study IPV perpetration, we could not find any previous study considering the multiplicative effects of MDD, AUD, and poor impulse control on physical IPV perpetration.As mentioned earlier, most of our knowledge in this area comes from cross sectional studies.Using a longitudinal design and national sample, we tested additive and multiplicative effects of MDD, AUD, and poor impulse control of male partner on changes in physical IPV reported by other partner over time.

Design and Setting
This study used data from waves 2 and 3 of Fragile Families and Child Wellbeing Study (FFCWS), a national population-based longitudinal study that started in 1998 in the United States.
The study used a random national sample.Participants were male and female partners who were selected from 20 US cities with populations of 200 000 or more.The cities were chosen by a stratified random sampling method to maximize cross city variation in economic and politic environments.Baseline data collected consisted of 4898 families selected from 75 hospitals across 20 cities. From this number, 3712 couples were unmarried and 1186 couples were married.More information regarding the FFCWS sampling strategy and interview protocol is available elsewhere. 24ata was collected during core interviews and the add-on In-Home Longitudinal Study of Pre-School Aged Children.Participants were interviewed at baseline (near the time of the target child's birth) and again at 1 (Wave 2, child age 1) and 3 (Wave 3, child age 3) following birth.As an add-on to the core interview, a subset of mothers (n = 3288) participated in the In-Home Study when the index child was 3 years old; male partners were not eligible to participate in the In-Home Study.
FFCWS has oversampled non-married couples. 24As non-marital unions are at greater risk for relationship instability, a large number of male partners were not living with the mother in waves 2 or 3.For instance, by the time the study target child was 3 years old, less than half of male partners resided in the home with the female partner.

Measures
Most variables were based on male partner's selfreport; female partner's report was used when male partner's report was unavailable or inappropriate to use.Male-to-female interpersonal violence was based on female partner's self-report.
Outcomes Female partner's victimization of physical IPV.Physical IPV was assessed by asking mothers 2 questions, on a 3point scale ("never", "sometimes", or "often"), regarding how often fathers carried out behaviors toward the mother, (e.g., slapping, kicking, hitting) which were adopted from the Conflict Tactics Scale (CTS-2) for adults. 25,26The original and revised CTS 25,26 have been the most commonly used research measures of domestic violence.Version 1996 measures psychological dimensions, physical violence, sexual violence, and financial control. 27

Major Depression. The Composite International
Diagnostic Interview -Short Form (CIDI-SF), Section A 28 was used to measure major depression disorder.The CIDI-SF is a standardized instrument that is consistent with the criteria set forth in the Diagnostic and Statistical Manual of Mental Disorders -Third Edition -Revised (DSM-III-R). 29This instrument has good reliability and validity). 28The CIDI-SF uses the criteria set forth in the Diagnostic and Statistical Manual of Mental Disorders (DSM-IV) to determine the probability that the respondent would be diagnosed with major depression if given the full CIDI interview.Major depression is indicated by feelings of depression or anhedonia experienced for most of the day, every day, for at least 2 weeks.Participants were classified as likely to have major depression if they endorsed the screening items and 3 or more depressive symptoms (e.g., losing interest, feeling tired, change in weight) (0 = no, 1 = yes).
Major Depression and Generalized Anxiety Disorder (GAD).The CIDI-SF was used to measure GAD. 28his diagnosis is based on DSM-III-R (American Psychiatric Association, 1994).CIDI0-SF has good reliability and validity. 28GAD is defined as 6 months or more of feeling excessively worried or anxious about more than one thing, more days than not, and having difficulty controlling their worries.Some of the common symptoms include being keyed up or on edge, irritability, restlessness, having trouble falling asleep, getting tired easily, difficulty concentrating and tense or aching muscles.Individuals were classified as having GAD if they met full DSM-III-R diagnostic criteria (0 = no, 1 = yes).
Alcohol Abuse.Heavy alcohol use was defined as 4 or more drinks (coded "1") or 0-3 drinks (coded "0") consumed in a single day in the past 12 months.This measure reflects the National Institute on Alcohol and Alcoholism (NIAAA) definition of heavy drinking days, 30 defined as 5 or more drinks in a single day for men.
Covariates.Control variables which were measured at baseline interview (Wave 1) included education level (1 = less than high school, 2 = high school degree or GED, 3 = some college/technical school or higher), age, income, minority status, and relationship status (1 = married, 2 = cohabiting, 3 = not married or cohabiting).Demographic variables (marital status, family income) and psychosocial variables (depression, alcohol abuse, anxiety) were measured at Wave 2, when the child was 1 year old.

Analysis Plan
We used PASW 20.0 (SPSS Inc, Chicago, IL) for our univariate and bivariate analysis.We used independent samples t test to compare impulse control, IPV, and control variables among those with and without depression, and anxiety.We also used Pearson correlation test for bivariate associations between IPV, impulse control, frequency of alcohol use and also control variables.
We used AMOS (SPSS Inc, Chicago, IL) for our multivariable analysis.We run 6 different structural equation models based on the type of IPV and IPV time point: (1) physical aggression at baseline, (2) physical IPV 2 years after baseline, (3) change in physical IPV from baseline to the next 2 years, (4) psychological IPV at baseline, (5) psychological IPV 2 years after baseline, and (6) change in psychological IPV from baseline to 2 years later.
In Model 1, after controlling for anxiety, and alcohol, and SES, we tested paths from depression and impulse control to IPV.We also tested Model 2 that tested the effect of depression on IPV in 2 groups of participants based on impulse control (Figure 1 and Figure 2).
The chi-square test, comparative fit index (CFI), chi-square to degree of freedom ratio, and also root mean square error of approximation (RMSEA) were considered as fit indices.A CFI of higher than 0.95, RMSEA of lower than 0.06, and chi-square test to degree of freedom ratio less than 2 were indicative of good fit. 31Variables measured at baseline did not have missing values.Variables measured at Wave 2 had less than 10% of missing value.

Results
Only 37.8% (n = 944) of couples were married at the time of survey.From all, 73% (n = 1836) of male and 71% (n = 1778) of female partners belonged to minority groups.By the means of race, 38% (n = 944) of male and 36% (n=888) of female partners were African American, while 31% of male (n = 774) and 33% (n = 777) of female partners were Hispanic.
Mean household income was $39 720 with a standard deviation of $35 436.Mean age of male partners were 28 ±7, with a range of 15-43 years.Mean and range of age of female partners were 26 ± 6 and 16-53 years, respectively (Table 1).
At year 1, 44% (n = 1,089) and 2% (n = 44) of mothers reported they experienced some degrees of psychological and physical IPV, respectively.Two years later, 34% (n = 837) and 1% (n = 33) of mothers reported that they experienced some degrees of psychological and physical IPV, respectively.Descriptive statistics of physical and psychological IPV at baseline and follow-up and their difference are reported in Table 2.
Our bivariate analysis did not show any direct association between depression and physical IPV at baseline, 2 years later, or their difference (Table 3).while baseline physical IPV, socioeconomic status (SES), GAD, AUD, and MDD were also in the model, impulse control was not associated with change of physical IPV over the 2-year follow-up.
Based on a model with excellent fit indices [X2 = 520.310,df = 130, P = .000,CFI=0.913,CMIN/ DF = 3.922, RMSEA = 0.032, 90% CI = 0.032-.037]which had SES, baseline physical IPV, anxiety, and AUD in the model, MDD was only associated with change in physical IPV among individuals with poor impulse control.Such association was not found between individuals with high impulse control.

Discussion
Based on our analysis, impulse control and MDD did not have main effects on the change in risk of perpetration of physical IPV, while the effects of SES, AUD, and HAD were controlled.However, poor baseline impulse control and MDD had synergistic effects on changes in physical IPV over time.That said, only depressed men with low impulse control showed an increase in physical IPV over time.National sample, large sample size, longitudinal design, and considering SES, MDD, GAD, and AUD as predictors of IPV all should be considered as strengths of this study.
Based on our findings, the effects of poor impulse control and depression on future change in perpetration of physical IPV are not additive but multiplicative.Poor impulse control increases rapid and unplanned reactions to internal or external stimuli without considering the negative consequences of such rapid reactions. 32Impulse control also is low sensitivity to negative consequences of behaviors and reactions to stimuli before complete processing of information. 32Impulsivity reduces attention to longterm consequences of own behaviors. 32Poor impulse control is a stable personality trait, a propensity to emit a certain response to stimuli. 33n 1988, Zuckerman et al introduced a personality dimension composed of traits such as impulsivity, sensation seeking, and anti-sociality and named it as behavioral under-control.In an 11-year longitudinal study of 489 young adults, behavioral under-control predicted relationship aggression net of gender, marital satisfaction, family history of AUD, and drinking patterns. 34Zuckerman (1993) argued that aggression, conduct disorder, and criminality which are associated with both heavy drinking 35 and the risk of AUD 36 may be in fact dis-inhibitory behaviors.
From a biological perspective, the link between AUD and IPV perpetration has been attributed to psychopharmacologic influence of alcohol on brain, cognitive processing, and lack of inhibitions.AUD reduces inhibitory effect of prefrontal cortex, specifically the orbital and ventromedial PFC, which has implication for behavioral control of impulsive aggression. 37,38It has been shown that AUD + participants with IPV have more severe deficits in tasks of impulse control, attention, concentration, and cognitive flexibility compared to controls. 39They argued that understanding how impulse control and related neurocognitive deficits in AUD leads to IPV may improve effective treatments and preventions.Future research should also test if impulsiveness, attention, concentration, and cognitive flexibility mediate the link between mental health and IPV perpetrartion. 39Other types of IPV should also be included in future research.
The findings presented in this study have implication for prevention of male to female physical IPV.Although both genders use violent behaviors against their partners, male perpetrator/female victim is a common type of IPV. 40Based on World Report on Violence and Health, between 10% and 69% of women in the community setting have experienced physical abuse during their lifetime. 41In a community sample of adult women in the United States, 29% have experienced physical violence, rape, and/or stalking by an intimate partner, 24% have been a victim of severe physical violence by a partner, and 15% have been injured as a result of IPV in their lifetime. 42,43The rate of severe physical aggression victimization among US women reported in the National Comorbidity Survey was reported low (to 6.5%). 44Based on US Department of Justice, The Federal Bureau of Investigation (FBI, 2011), about 100 women are murdered by IPV each year. 45,46ur finding is limited to physical IPV, and future research should be conducted on other types of IPV.Different types of violence tend to co-occur in some relations. 47We can hypothesize that the interaction between MDD and impulse control may vary based on the type of IPV.Although some studies have shown that the pattern of risk factors may be mostly the same for different IPV types, 48 there are studies which have reported different types of IPV may be different in their risk factors.For instance, a study 49 showed that anxiety predicted physical aggression perpetration but not perpetration or victimization of verbal aggression.In 2012, Basile and Hall argued that there is a need for future studies to better understand how distinct risk factors affect perpetration of various types of IPV by men. 50ur findings suggest that MDD and impulse control should be jointly evaluated and addressed for the treatment of physical IPV among men.Men with MDD may largely benefit from self-regulation trainings and impulse control enhancement.Joint interventions for impulse control and MDD should be regarded an important part of treatment of male perpetrators of physical IPV.Screening of impulse control and MDD among male IPV perpetrators may identify individuals that can benefit from such combined interventions.Future research should test efficacy of such approach for treatment of male IPV perpetrators.
To conclude, poor impulse control and MDD have synergistic effects on change in IPV perpetration by men.Further future research is needed in this regard.

Ethical Approval
All study procedures were approved by Institutional Review Board Committees at Princeton University and Columbia University.Verbal and written informed consent was obtained from participants at each interview, and all participants were compensated for their involvement in the study.

Figure 1 .
Figure 1.Diagram of Structural Equation Model Testing Model 1.Based on this model, after controlling physical intimate partner violence perpetration at baseline, SES, baseline general anxiety disorder (GAD), alcohol use disorder (AUD), and impulse control, major depressive disorder (MDD) was not correlated with physical intimate partner violence 2 years later.Dotted path is non-significant.

Figure 2 .
Figure 2. Diagram of Structural Equation Model Testing Model 2. Based on this model, major depressive disorder (MDD) was only associated with perpetration of physical IPV among men with poor impulse control, but not men with high impulse control.That means impulse control moderates the effect of depression on physical aggression 2 years later.

Table 1 .
Descriptive Statistics of Male and Female Partners Included in This Study (2500 Couples Who Were Either Married or Cohabiting at Baseline)

Table 2 .
Characteristics of Male and Female Partners Enrolled to This Study

Table 3 .
Correlation Between Baseline Variables and Physical Intimate Partner Violence