Children and Youth Services Review 31 (2009) 249256

Contents lists available at ScienceDirect

Children and Youth Services Review

j o u r n a l h o m e p a g e : w w w. e l s ev i e r. c o m / l o c a t e / c h i l d yo u t h

Domestic violence crimes and children: A population-based investigation of direct sensory exposure and the nature of involvement
Rachel A. Fusco a,, John W. Fantuzzo b
a b

School of Social Work, University of Pittsburgh, Pittsburgh, PA, United States Graduate School of Education, University of Pennsylvania, Philadelphia, PA, United States

a r t i c l e

i n f o

a b s t r a c t
Children's exposure to domestic violence is a major national problem. Researchers and policymakers have called for research guided by comprehensive conceptual frameworks to advance understanding of this complex risk to children's well-being [Center for Disease Control and Prevention (2006). Preventing intimate partner violence, sexual violence, and child maltreatment. Retrieved June 3, 2006 from http://www.cdc.gov/ ncipc/pub-res/research_agenda/07_violence.htm.; National Institute of Justice (2007). Adolescents, neighborhoods, and violence: Recent ndings from the Project on Human Development. Retrieved on September 5, 2007 from http://www.ncjrs.gov/pdfles1/nij/217397.pdf). The present study used a developmental epidemiological model to explore the prevalence and nature of children's exposure to and involvement in domestic violence crimes investigated by law enforcement across a population. During the year under study 1581 domestic violence crimes were investigated by law enforcement. Forty-three percent of all domestic violence crimes had children in the household, and nearly all of those children (95%) experienced sensory exposure to the violence. A logistic regression model revealed a relationship between child exposure and domestic violence event characteristics such as victim injury, mutual assault, and perpetrator arrest. This research also examined how children are involved in domestic violence events. Three distinct types of involvement were revealed: children were part of the precipitating event; children called for help; and children were physically involved. Findings highlight the importance of developing a comprehensive surveillance system to ensure children exposed to domestic violence are made visible so they can be referred to appropriate services. 2008 Elsevier Ltd. All rights reserved.

Article history: Received 1 May 2008 Received in revised form 11 July 2008 Accepted 22 July 2008 Available online 27 July 2008 Keywords: Children exposed to domestic violence Children's involvement in domestic violence

1. Introduction In 1994, the Violence Against Women Act made domestic violence a crime and mandated data collection at the national, state, and local levels (VAWA, U.S. Department of Justice, 1994). As a result, all states enacted legislation to provide civil, as well as criminal, penalties for acts of violence in the home. This legislation required investigation and documentation of domestic violence. VAWA did not specically address the issue of children's exposure to domestic violence, although children's exposure to domestic violence has been increasingly recognized as a major problem requiring signicant attention. Researchers and policymakers have called for research guided by comprehensive conceptual frameworks to advance our understanding of how child exposure to domestic violence adversely affects children's physical and psychological well-being (Center for Disease Control and Prevention, 2006; National Institute of Justice, 2007). The National Council of Juvenile and Family Court Judges (1999) published

Corresponding author. E-mail address: raf45@pitt.edu (R.A. Fusco). 0190-7409/$ see front matter 2008 Elsevier Ltd. All rights reserved. doi:10.1016/j.childyouth.2008.07.017

a report emphasizing the importance of identifying families with domestic violence and providing them with appropriate services to support the welfare of both the direct victim and the children exposed to the violence. This report highlighted the importance of developing a comprehensive research agenda to understand the extent of this national problem and the effects of exposure to domestic violence on children's development. A major problem with our current knowledge base is the lack of precise prevalence gures for children exposed to domestic violence. The most commonly cited gures are between 3.3 million (Carlson, 1984) and 10 million (Straus, 1992) children exposed to domestic violence in the U.S. Current prevalence gures are not only gross estimates but have methodological limitations. The studies they are derived from have restricted denitions of domestic violence and often depend on adult recollections of childhood events that may have happened more than 20 years ago. Many researchers have found retrospective reports about emotionally charged memories to be inaccurate or incomplete (e.g., Bradburn, Rips, & Shevell, 1987; Ceci, Ross, & Toglia, 1989; Henry, Moftt, Caspi, Langley, & Silva, 1994). Another problem with many current prevalence estimates is their reliance on parental reports of their children's exposure (Jouriles, McDonald, Norwood, & Ezell, 2001; O'Brien, John, Margolin, & Erel,

250

R.A. Fusco, J.W. Fantuzzo / Children and Youth Services Review 31 (2009) 249256

1994; Tjaden & Thoennes, 2000). Other studies indicate that parents are likely to report that their children were asleep or unaware of the violence, although these same children often provided detailed memories of the violent event when asked directly (Jaffe, Wolfe, & Wilson, 1990; Peled, 1998). These methodological problems have made it difcult to collect accurate statistics. Although there is little known about the prevalence of children exposed to domestic violence, there is a large volume of research on the impact of the problem. This research has provided evidence that children exposed to domestic violence show signicantly more social, emotional, and cognitive problems compared to their non-exposed peers (Jaffe et al, 1990; Jouriles et al, 2001; Margolin & Gordis, 2000; Osofsky, 1999). Although these impact studies contribute to the knowledge base on children's exposure to domestic violence, they reect the concept of urgent knowing. Research on impact has been conducted without an understanding of what children are actually being exposed to within domestic violence events, how children are exposed to violence, and what moderating factors affect outcomes. These studies are not only being conducted in the absence of a sufcient knowledge base, they often have methodological weaknesses that makes their implications suspect. Studies comparing the psychological functioning of children exposed and not exposed to domestic violence have some signicant shortcomings that limit our understanding of the extent and nature of the problem. First, although children's exposure to domestic violence events has been described as a public health problem of epidemic proportions (Glodich, 1998), the current research is not populationbased and relies principally on convenience samples. More specically, samples were often drawn from children who are with their mothers in domestic violence shelters (Jouriles et al., 2001). These samples represent a small percentage of children exposed to domestic violence. Children in shelters are also more likely to have been exposed to the most severe and chronic forms of violence (Ware, Jouriles, Spiller, McDonald, Swank, & Norwood, 2002). The shelter studies have provided some important information, but do not provide an accurate depiction of what the majority of children are exposed to within a domestic violence event. An empirical examination of the prevalence and nature of children's exposure to domestic violence requires standard denitions of the nature of children's exposure to domestic violence. Children are often described as observers or witnesses to domestic violence (Edleson, 1999; Jaffe et al, 1990). However, these terms imply that children are passive observers of this violence; that they are in the background when the violence occurs. This terminology fails to capture the myriad ways children are involved in domestic violence events. These children may be more accurately described as being exposed to domestic violence, a term that is inclusive and incorporates a variety of children's experiences (see Edleson, 1999). Whereas the terms observe and witness imply passivity, exposure to domestic violence includes a wider range of indirect or direct involvement, such as intervening by placing one's self in harm's way or calling 911 (Edleson, Mbilinyi, Beeman, & Hagemeister, 2003; Fantuzzo et al., 1997). The developmental epidemiological framework emphasizes the importance of standard denitions because they allow for greater comparability and generalizability of ndings across studies. Recent studies have effectively used police surveillance to gather information about domestic violence events and associated risk factors. Police ofcers collected data on substantiated domestic violence using standard methods in the Spouse Assault Replication Program (SARP). SARP was a large, cross-city eld experiment assessing the impact of arrest in deterring subsequent misdemeanor domestic violence (Maxwell, Garner, & Fagan, 2001). The SARP database contained information on domestic violence events, individuals present in the household during the events, and associated risk factors across ve municipalities (Milwaukee, Omaha, Atlanta, Charlotte, and Miami). Data were collected at the time of the incident, thus avoiding

the problems of retrospective reports. A secondary analysis of this database by Fantuzzo et al. (1997) showed that children were disproportionately present in households where there was a substantiated incident of domestic violence. Households where domestic violence occurred included higher levels of risk factors to children, such as poverty, single-female headed households, and substance abuse associated with the event. However, from an epidemiological perspective this study was limited in two ways. First, it was not a comprehensive study of events across a population. Only misdemeanor domestic violence cases were included in the study, and cases were excluded if they did not demonstrate male-to-female violence (Maxwell et al., 2001). Second, no data were provided to document the reliability and validity of the use of the standard protocol or police ofcer training. A study by Gjelsvik et al. (2003) utilized the Rhode Island Department of Health Violence Against Women Public Health Surveillance System to examine factors associated with children in the household during police-substantiated domestic violence. Police collected data on the demographic characteristics of the victim, characteristics of the incident, and whether children were in the household. Results showed that 44% of all substantiated domestic violence events had children in the household. These children were more likely to be from ethnic minority households and 47% of them were less than 6 years old. Although this study illustrated police ofcers working as public health sentinels across a xed population, it had several methodological limitations. First, no data on the characteristics of perpetrators were provided. Second, no details were given about the denitions of the domestic violence event variables or on the methods the police used to collect data. Third, no information was given regarding police ofcer training on direct assessment of domestic violence and children in the household. Finally, the reliability and validity of the data collection instruments were not reported. Fantuzzo, Fusco, Mohr and Perry (2007) partnered with police ofcers to develop a standard, validated protocol, the Domestic Violence Event Protocol (DVEP), to gather information on all reported domestic violence events across an entire municipality. Results indicated that children were in the household in almost half of all events, and households experiencing domestic violence were signicantly more likely to have children compared to households in the county at large. A multiple regression model revealed that when children were in the household at the time of the domestic violence event, the event was more likely to have high risk characteristics such as victim injury, the presence of weapons, and substance use during the incident. However, this study only looked at whether children were in the household at the time of the event; it did not provide data on direct sensory exposure to the violence, the number of children exposed, or the children's demographic characteristics. These data are important because the child trauma literature documents that the nature and degree of exposure to traumatic events moderate impact (Rossman, Hughes, & Rosenberg, 1998). Fantuzzo and Fusco (2007) used an enhanced version of the DVEP, which added basic information on child characteristics and the nature of children's sensory exposure to domestic violence. A descriptive picture of all the children in the household during law enforcementinvestigated domestic violence events across an entire municipality for 1 year was provided. This included whether the children in the household had direct sensory exposure to the violence as well as potentially moderating variables such as children's age, gender, race, and relationship to the victim. This study also examined the relationship between characteristics of domestic violence events and children's direct sensory exposure to these events. They found that children were disproportionately represented in household with domestic violence, and these children were more likely to be younger than age six. Domestic violence households with children, compared to overall households with children in the county, were more likely to

R.A. Fusco, J.W. Fantuzzo / Children and Youth Services Review 31 (2009) 249256

251

be nonwhite. The domestic violence households with children were more likely to be headed by single females and living in poverty. Children were also disproportionately likely to be exposed to domestic violence events with high risk characteristics, such as weapon use and victim injury. The present study replicated the exploratory study by Fantuzzo and Fusco (2007) across another full year of domestic violence police investigations. This replication was designed to test the hypothesis that the same pattern the high risk features of domestic violence events would be associated with child direct sensory exposure in the same municipality. The current study also rened the original multiple regression model by conducting meditational analysis to test whether poverty was mediated by race and extended previous research by exploring the concept of children's/involvement/in domestic violence events. This study addressed four primary research questions. First, can children's involvement in domestic violence events be determined reliably from required police ofcer investigation reports? Second, can subtypes of children's involvement in domestic violence events be determined reliably from required reports of police ofcer investigations? Third, what is the overall prevalence of children's involvement in domestic violence events and the prevalence of subtypes of involvement? And fourth, are there signicant differences in children's age, sex, and race across subtypes of involvement? 2. Methods 2.1. Participants Data for this study were obtained from the domestic violence database of a large county police department in the Northeast. The police department consists of six districts and 1074 ofcers whose ethnicity corresponds to the composition of the county. Data were collected on all domestic violence events (DVEs) substantiated by law enforcement across a one-year period using the Domestic Violence Event ProtocolChild Enhanced (DVEP-C). The participating county is suburban with a population of 837,000. County residents are predominantly Caucasian and middle-class. Only 5% of residents live in poverty, and poor families with children under age 5 constitute 6% of the total population (U.S. Bureau of the Census, 2000). Crime rates in the county are below the national average; violent crime is low compared to populations of similar size (U.S. Department of Justice, 2006). Aggravated assaults occurred in nearly 1 out of 1000 residents, and 48% involved violence between intimate partners. 2.2. Measures 2.2.1. Domestic Violence Event ProtocolChild Enhanced (DVEP-C) The DVEP-C is a standard data collection protocol used by police ofcers to capture key characteristics of DVEs (Fantuzzo & Fusco, 2007). It provides a format for collecting demographic information on the victim and perpetrator including race, sex, age, and nature of their relationship. The DVEP-C has checklists to record police ofcers' observations documenting the means of assault by using a continuum of severity (from hand to knife or gun) as well as a checklist on which to record any visible injuries (e.g., lacerations, bruises). The DVEP-C includes information about whether substance use was part of the event, whether the assault was mutual, and whether or not the DVE resulted in an arrest. Police ofcers interviewed every child in the household to assess their exposure (child saw only, heard only, or saw and heard the violence). In addition to the checklist of exposure, police ofcers were required to write a full narrative of the event, including a detailed report of children's experience of the DVE. With young children who are not developmentally able to respond, the ofcer asked the victim about the nature of the child's exposure. The DVEP-C demonstrated high internal consistency (Cronbach's = .87; Fantuzzo et al., 2007) and reliability checks on children's age, gender, and

ethnicity showed intraclass correlations ranging from r = .86 to .95 (see Fantuzzo & Fusco, 2007 for additional information). 2.2.2. Poor neighborhood Neighborhood-level poverty was measured by using U.S. Bureau of the Census (2000) data. In this municipality, more than 99% of all reported DVEs occurred at the victim's residence. ArcView 9 (ESRI, 2004) was used to geocode all addresses where DVEs occurred, and each was assigned to a block group. A poverty rate was established by dividing the number of households below the poverty line by the total number of households in each block group. This rate ranged from 0 to 27% across the county. The rates were then divided into equal quintiles. Block groups in the top quintile of poverty were coded as the poorest neighborhoods in the county. 2.3. Procedures 2.3.1. Data collection Data were collected by police ofcers trained to use the DVEP-C for all reported DVEs. This training was followed by annual continuing education session on dening and understanding DVEs. As part of her regular duties, the police captain reviewed every completed DVEP-C to ensure they were lled out accurately and completely. 2.3.2. Data analysis Logistic regression was used to assess how children's exposure related to the nature of domestic violence events. The previous study by Fantuzzo and Fusco (2007) included seven empirical event characteristics of domestic violence in the model: victim race, victim injury, weapon use, mutual violence, relationship between victim and perpetrator, arrest of the perpetrator, and substance use. The current model added the poor neighborhood variable with those seven variables to test the mediational relationship between race and poor neighborhood. The outcome variable was classied dichotomously since this allows the calculation of odds ratios that are easily interpretable as the degree of likelihood given the explanatory variables on the outcome variable. In order to dichotomize explanatory variables, some data had to be collapsed into binary categories. The following are the sets of dichotomized explanatory variables: Highest neighborhood poverty vs. Lower neighborhood poverty (based on census block group); Victim injury (includes major and minor injuries) vs. No victim injury; Arrest (at the scene of the domestic violence event) vs. No arrest; Weapon (refers to guns, knives, clubs, or any other object used by the perpetrator to inict injury) vs. No weapon; Mutual violence vs. Non-mutual violence (whether the police determined there was one or two aggressors), and Substance use (including both alcohol and drug use at the time of the domestic violence event) vs. No substance use. The DVEP-C collects information about the nature of the relationship between the victim and perpetrator. Coparenting couples only have a relationship through children they have in common. Since coparenting is the only situation where it is known that both victim and perpetrator are the biological parents of the child, the relationship categories were dichotomized to create Coparenting vs. Other. All variables were entered into the model simultaneously. 2.3.3. Child involvement in DVEs To establish whether child involvement can be reliably determined, all 639 DVEP-C forms indicating child exposure were reviewed. The following operational denition was used to study children's involvement in domestic violence events: Involvement in Domestic Violence EventChildren were considered involved in a domestic violence event if in addition to having sensory exposure to the event they were also directly connected to the event by the victim's report or some veriable action. The investigator applied this denition to the police written reports of all 639 DVEP-C reporting forms where children had direct sensory exposure. These DVE reports were separated into two discrete

252

R.A. Fusco, J.W. Fantuzzo / Children and Youth Services Review 31 (2009) 249256

piles based on this denitioninvolved or not involved. The involved reports would be further explored to see if any subtypes of child involvement emerge. The reliability of this identication process was tested by recruiting 12 practicing child welfare professionals to serve as independent raters. Ten raters were women, and two participants had PhDs in addition to MSW degrees. Nine participants identied as White, two as African-American, and one as Latino. Participants had between 3 and 27 years of experience working in child welfare. Chi-square analyses were used to determine whether there were signicant differences in involvement by child sex, child age, and child race/ethnicity. Child sex was examined across the different subtypes that emerged, with boys coded as 0and girls coded as 1 Age differences were examined for signicance, and a dichotomous variable was created through a median age split. Child race/ethnicity was dichotomized as White and nonWhite. 3. Results 3.1. Prevalence of domestic violence events There were 1581 substantiated domestic violence crimes during the year under study. Frequency analyses were used to develop a descriptive picture of the victims and perpetrators of the DVEs, the specic characteristics of the substantiated event, and key contextual variables. The majority of DVEs (73%) involved men as the primary perpetrator against female victims. In the remaining cases, women were the primary aggressor 13% of the time, and 14% of the events were determined to have mutual aggressors both parties were considered equally culpable in starting the violence. Very few cases involved same sex couples (less than 1%). Victims had a mean age of 33 years (SD = 8.6) and perpetrators had a mean age of 34 years (SD = 8.9). Although two-thirds of the county are White, almost twothirds of the DVE victims were minorities (36% African American; 23% Hispanic; 4% Asian; Table 1). Approximately 67% of victims and perpetrators were married or cohabitating at the time of the violent event. During the year under study, less than 5% of the reported DVEs were multiple incidents. Victims were injured in about 70% of the DVEs, and almost half resulted in the perpetrator's arrest. Weapons such as knives, guns, or common objects (such as cellular telephones) were part of the DVE 20% of the time. Substance use by the victim or perpetrator was part of more than one-third of the DVEs. Almost 40% of the DVEs occurred in census block groups with the highest quintile of neighborhood poverty. 3.2. Prevalence of child exposure to DVEs Forty three percent of all DVEs had children in the household at the time of the event (n = 679). Nearly all of these children (95%) had experienced sensory exposure to the violence. Data collected by the investigating ofcer revealed that 22% of the children only heard the violence, 4% only saw the violence (e.g., they were outside and saw the violence through a window), and more than 60% both heard and saw the violence. Three percent of the children (n = 21) were physically injured in addition to both hearing and seeing the domestic violence event. Forty children were in the household but did not experience direct sensory exposure to the DVE (e.g., asleep during the event).
Table 1 Demographic characteristics of victims, perpetrators, and children Victim (%) White African American Hispanic Asian Females Mean age (in years) 37.5 35.6 23.3 3.6 85.5 33.1 Perpetrator (%) 32.4 42.4 22.2 3.0 14.7 34.3 Children (%) 28.9 45.2 21.8 4.1 50.7 4.8

The domestic violence victim was almost invariably the child's mother (94%). Demographic characteristics of all children exposed to domestic violence showed no signicant differences in the sex of exposed children (boys 49%). Children exposed to DVEs were disproportionately younger than age 6 years (52%). The exposed children were 45% African American, 29% White, 22% Hispanic, and 4% Asian. There was an average of 1.4 children exposed per household (SD = 0.7) with a range of one to four children. Exposure Hypothesis 1. High risk features of domestic violence events are more likely to be associated with child sensory exposure. The following variables were included in the multiple logistic regression model: victim injury, weapon use, mutual violence, relationship between victim and perpetrator, and substance use, victim race, and neighborhood poverty. In order to prevent multicollinearity in the model, a correlation matrix was developed with the eight variables. The phi coefcients for these variables were examined to ensure they were all under the level considered acceptable (.70; O'Brien, 1978). Neighborhood poverty and nonwhite victim were highly correlated (phi = .81) indicating multicollinearity if both variables were included in the model. It was hypothesized that this correlation was the result of neighborhood poverty mediating race in this population. Baron and Kenny's (1986) model for mediational effects was used to test this hypothesis. Findings showed that neighborhood poverty was signicantly associated with the victim being nonwhite [Score (1, N = 1581) = 71.42, p b .0001]. Next, these two variables were entered simultaneously into a logistic regression model with the dependent variable, child exposure. In this model, race no longer had an impact on child exposure [Score (2, N = 1581) = 33.58, p b .0001], indicating a perfect mediation of neighborhood poverty on the relationship between race and child exposure. Therefore, the nonwhite variable was dropped from the model. Results indicated that six of the seven explanatory variables in the model had signicant independent relationships with children's direct exposure to domestic violence events [Score (7, N = 1581) =197.85, p b .0001; Table 2]. If the perpetrator and victim were coparents of the child (making the perpetrator the child's biological father) then the child was almost three times as likely to be directly exposed to the domestic violence event. Odds ratios and probability levels indicated that if the perpetrator was arrested and if the violence was mutually aggressive between both parties, children were more than twice as likely to be directly exposed to the violent event. There were signicant, though relatively weaker relationships between child exposure to DVEs if the victim was injured or if the violence occurred in a relatively poor neighborhood. Substance use was also signicant in the model, but there was a decreased likelihood of child exposure when either the victim or perpetrator used drugs or alcohol before or during the DVE. Weapon presence was the only non-signicant variable in the model. 3.3. Involvement Question 1: Can children's involvement in domestic violence events be determined reliably from required reports of police ofcer investigations? DVEPC reports indicating the presence of children were blocked into two piles: involved and not involved. Using the unique number assigned to each case by the police department, a random sample was selected from each pile. Sixty narratives were selected from the involved pile, and 40 were selected from the not involved pile. Training for the raters was provided by the investigator. The purpose of the study was explained, and independent raters were provided with the operational denition of involvement. Participants were asked to determine which narratives indicated children's involvement in domestic violence events by sorting them into involved and not involved-. All identifying information on the child, the victim, and the perpetrator was removed from the written narratives.

R.A. Fusco, J.W. Fantuzzo / Children and Youth Services Review 31 (2009) 249256 Table 2 Odds ratios and characteristics of children's exposure to domestic violence events Outcome variable Explanatory Variable Neighborhood poverty Perpetrator as father Arrest Victim injury Substance use Mutual Weapon Child exposed 1.46 2.69 2.70 1.31 .67 2.85 (95% CI) (1.141.86) (1.983.65) (2.133.44) (1.001.69) (.53.84) (2.103.89)

253

p b .05, p b .01, p. b 001, p b .0001. Note. N = 1581. Signicance is based on Wald chi-square statistics.

A kappa coefcient was calculated to determine inter-rater reliability. Kappa values are characterized into three groupings, and a value greater than .80 (80%) represents strong agreement (DiEugenio & Glass, 2004). The kappa coefcients for the 12 raters had a mean of .95 (SE = .03), with a range from .91 to 1. 3.4. Involvement Question 2: Can subtypes of children's involvement in domestic violence events be determined reliably from required reports of police ofcer investigations? All written reports coded as involved were reviewed to look for types of involvement. They were then sorted in mutually exclusive categories by the investigator. Three themes emerged, labeled child was part of precipitating event, child called for help, and child was physically involved. Post hoc denitions of the three types of involvement were developed: Child was part of the precipitating eventThe victim and/or suspect cite factors related to the children, including parenting practices, as the reason for the violence. There is no intent on the part of the investigator to blame the child for the violence, but to establish how parents may use childrelated issues to instigate arguments. Child called for helpThe child made some effort, either directly or through others, to bring assistance to intervene in the domestic violence. Child was physically involvedThis is a multidimensional concept that includes all the different ways children can actually become a part of the violent event. This ranges from active (such as a child actually pulling the suspect away from the perpetrator) to passive (such as a child being held in it's mother's arms as she is assaulted). This includes children being injured, complaining of pain, or showing marks as a result of the domestic violence. This injury can be intentional, such as the suspect hitting a child when s/he tries to intervene, or unintentional, such as the child getting hit when the intended target was the victim. The same raters who determined involved/not involved in the random sample of police narratives were also asked to subtype those they considered involved into categories. They were provided with the three categories and their operational denitions. Raters were then asked to choose which category most clearly represented the type of involvement. These data were coded nominally and the kappa coefcient was calculated. The mean coefcient across the 12 raters was .94 (SD = .04), and ranged from .89 to 1. 3.5. Involvement Question 3: What is the overall prevalence of children's involvement in domestic violence events and the prevalence of subtypes of involvement? Almost 75% of all children who experienced direct sensory exposure to the DVE were also involved in the violence (n = 467). A descriptive picture of children involved in domestic violence was

developed. There were slightly more girls involved than boys (53% vs. 47%). Involved children were 44% African American, 31% Hispanic, 22% White, and 3% Asian. Mean child age was 7 years old (SD = 5.3). Of all the children involved in domestic violence crimes, the largest percentage was physically involved (37%). More than a third of the involved children were part of the precipitating events (35%), and 28% called for help. Descriptive information was also examined across subtypes. Children in the Child was part of the precipitating event subtype had a mean age of 5.8 (Median = 4.0; SD = 5.0). There was an equal number of boys and girls. The racial/ethnic distribution of children who were part of the precipitating event was: African American 41%, Hispanic 33%, White 23%, and Asian. In the Child called for help subtype, the mean age was 8.4 (Median = 9.0, SD = 5.3). They were more likely to be girls (57%) than boys. These children were 40% African American, 21% Hispanic, 35% White, and 4% Asian. The mean age of children in the Child was physically involved subtype was 5.7, and the median was 3 (SD = 5.2). The children in this subtype were slightly more likely to be girls (52%). African American children made up almost half (49%) of all physically involved children, with the remanding children identied as Hispanic (26%), White (22%), and Asian (3%). More than 10% of the physically involved children were injured during the event. 3.6. Involvement Question 4: Are there signicant differences in children's age, sex, and race across subtypes of involvement? Chisquare analyses were conducted to look for signicant differences between child sex, age, and race across the three types of child involvement in domestic violence. Since the median age of children exposed was 6 years, a dichotomous variable was created with children 06 coded as 0 and children 7-17 coded as 1. Age differences were statistically signicant across the three types of child involvement. Children aged 0 to 6 were signicantly more likely to be part of the precipitating event (2 (1) = 55.9, p b .0001) and to be physically involved (2 (1) = 78.5, p b .0001). Children 7 to 17 years old were more likely to call for help (2 (1) = 28.6, p b .01). In terms of sex, there were no signicant differences between either being part of the precipitating events or being physically involved. However, girls were signicantly more likely to call for help (2 (1) = 13.8, p b .05). No signicant racial differences were found among the three subtypes of domestic violence involvement. 4. Discussion Police ofcers' reports indicated that nearly all children present in the household experienced direct sensory exposure to the violence. These ndings are similar to prior research done on children exposed to domestic violence in this municipality (Fantuzzo & Fusco, 2007). That study revealed that 89% of all children experienced sensory exposure to this violence, and this was most likely to be in the form of both hearing and seeing the violence. The current ndings reveal that domestic violence resulting in physical injury to the victim had a higher likelihood of having children directly exposed than DVEs with no injuries. There is growing evidence that children who witness violence resulting in injuries are more likely to experience post-traumatic stress symptoms (Osofsky, 1999; Osofsky, Wewers, Hann, & Fick, 1993; Pynoos & Nader, 1993). Many of these ndings come from literature on children exposed to community violence, but Fick, Osofsky, and Lewis (1997) found that both parents and police perceive witnessing violence against a parent to have a much greater impact on children than violence against a stranger. When children were directly exposed to DVEs, controlling for the level of violence (dened as victim injuries, presence of weapons, and

254

R.A. Fusco, J.W. Fantuzzo / Children and Youth Services Review 31 (2009) 249256

mutual violence) and neighborhood poverty, the perpetrators were more likely to be arrested. While there is a substantial body of literature on the impact of parental incarceration on a child (e.g., Johnson & Waldfogel, 2002; Parke & Clarke-Stewart, 2002; Takayama, Wolfe, & Coulter, 1998), very few studies have explored the effects of seeing a parent arrested. This experience may be traumatic for children. According to a Child Welfare League of America report (Wright & Seymour, 2000), a child's reaction to seeing a parent arrested may include feelings of helplessness, bitterness about the parent being treated harshly, and anger toward the arresting ofcers. However, it is important to note that many of the parents referred to in their report were arrested for drug or other non violent offenses. Children may feel differently about a parent being arrested for hurting another parent as opposed to other types of crimes where the victim is unknown to the child. Children were more likely to be directly exposed to mutuallyperpetrated domestic violence. Although Straus and Gelles (1990) found in a national survey that almost 50% of domestic violence relationships were categorized by bi-directional violence, children's direct exposure to mutual violence is an understudied variable in the domestic violence literature. Several studies have examined the impact of children's mother as the primary or sole aggressor, but only a few have looked at the effects of mutual violence. These children may be at greater risk of physical harm because they have two violent parents. One study showed that in households with mutual violence, children were abused by one or both parent almost a third of the time (Slep & O'Leary, 2005). The signicance of the coparenting variable indicates that directly exposed children are more likely to witness violent events involving both their mother and their father. Instead of the perpetrators being a stranger or a transient contact, the perpetrator was more likely to be the child's biological father. While the focus in research is on the negative effects of children exposed to a mother's victimization, relatively little is known about the effects of having a father who perpetrates violence. There have been some studies that have shown that quality of fathering is negatively associated with the presence of marital conict. Cummings, GoekeMorey, and Raymond (2004) found that when marital conict is present, the quality of fathering is more likely to be compromised than is the quality of mothering. Their results showed that fathers in conictual intimate relationships demonstrate less engagement with and higher negativity toward their children. A study by McDonald, Jouriles, Norwood, Ware, and Ezell (2000) revealed that paternal marital violence was related to internalizing and externalizing problems in children, controlling for family demographic variables, parent-child aggression, and maternal marital aggression. These results suggest that children directly exposed to domestic violence may also face the added risk of being poorly parented by fathers. An unexpected nding showed that substance use was inversely related to child exposure. When substance use was part of the DVE, children were less likely to be exposed to the violence. Another police surveillance study of children exposed to domestic violence also found that children were less likely to be exposed when substance use was part of the DVE (Gjelsvik, Verhoek-Oftendahl, & Pearlman, 2003). Although this nding looks like it has positive implications for children in domestic violence households, it may actually be indicative of greater problems within the family. Several studies have found that a disproportionately high percentage, from 10% to 67%, of Child Protective Services' (CPS) cases involve parental substance abuse (Harrington, Dubowitz, Black, & Binder, 1995; Famularo, Kinscherff, & Fenton, 1992; Kelley, 1992). Addicted parents may devote so much time, effort, and money to obtaining and using drugs and alcohol that little is left to meet the needs of the children. Therefore, in families with a chronic pattern of substance use, children may not be exposed to DVEs because they have already been removed by CPS (Keller, Cummings, Davis, & Lubke, 2005; Leonard & Roberts, 1998; O'Farrell,

Murphy, Neavins, & Van Hutton, 2000). It is also possible that children who are accustomed to a pattern of substance use and violence in their household learn to anticipate cues to temporarily remove themselves from the environment (Werkerle & Wall, 2002). The ndings from the multiple logistic regression model were similar to the previous study (Fantuzzo & Fusco, 2007) with one exception. In the prior study substance use was not signicantly associated with children's direct sensory exposure to domestic violence; in this study substance use was signicantly correlated with a reduced likelihood of children's direct sensory exposure. As stated above, ndings may indicate that these highly unstable home environments, which contain domestic violence and substance use, are more likely to have children out of the household either voluntarily or involuntarily. This research represents the rst population-based empirical study of child involvement in domestic violence events. Findings showed that almost 75% of all children exposed to domestic violence were also involved in the violence. These results indicate that children are not passive observers, but are active participants in the violence. Domestic violence is not just happening in their home; these children are involved in the incident in very specic ways. Three clear subtypes of involvement were found from the investigation of police reports. The rst type was children as part of the precipitating events, disproportionately composed of children aged 6 and younger. Studies of marital conict have looked at the stress that is added to a relationship when young children are present (DeaterDeckhard, 1998). Very young children are more dependent and have needs that may overtax parents with mental health issues, nancial problems, or other stressors (Rodgers, 1998). Children younger than age seven are probably not sophisticated enough to understand the complexities of interparental violence, and are likely to blame themselves for the ghting. Studies on the effects of parental conict have looked at child-precipitated argument and found that children report a strong sense of guilt when this occurs (Grych & Fincham, 1990; Grych, Seid, & Fincham, 1992). However, these studies have looked only at nonviolent martial conict. The second subtype was child called for help. The effects of a child taking on this responsibility are largely unknown. Peled (1998) conducted qualitative interviews with children living in domestic violence households and found that many reported that they frequently called the police. In a study by Edleson and his colleagues (2003) -mothers reported that their children called for help or physically intervened in a violent event 12% of the time. Girls in the present study were signicantly more likely to call for help than boys. Research on the gender effects of help-seeking behavior shows that females are socialized to ask for help from a young age (Lee, 2002; Newman, 1990). Males are discouraged from seeking help, and view this as feminine behavior (Lane & Addis, 2005; Sheu & Sedlack, 2004). Children calling for help were more likely to be age seven and older. This is probably due to the fact that older child have a better mental and physical capacity to seek help when needed. There is no information in the existing literature on the possible effects of such action. It is possible that the child actively calling for help has a positive effect, because the child may be less likely to feel passive or ineffective. However, this level of responsibility could also be negative for a child, especially if the perpetrator is the child's father even more so if the father is arrested. The third subtype was children's physical involvement in domestic violence. This type included injuries to the child, regardless of whether it is intentional or unintentional. This subtype made up the largest proportion of involved children. There is some literature that indicates that children's physical involvement is a common role in domestic violence events. In Edleson and his colleagues' (2003) study of child involvement, mothers reported that their children physically intervened 23% of the time. Alarmingly, the physically involved children in the current study were more likely to be age 6 and younger.

R.A. Fusco, J.W. Fantuzzo / Children and Youth Services Review 31 (2009) 249256

255

Children may feel empowered by intervening in a situation, but it may also result in greater emotional and physical danger to the child. Eleven percent of children who were physically involved suffered a police-substantiated injury as a result of the involvement. 4.1. Implications for policy and practice This study demonstrated that police ofcers in a police department that provide substantial training in family violence (both in the police academy and through ongoing professional development and supervision) can serve as reliable public health sentinels for children exposed to domestic violence crimes. Police departments in the 21st century have recognized their vital role in often being the rst responders to vulnerable populations (Greene, 2001). This understanding has led many police departments to move toward a model of community policing, which provides a positive context for police to serve as part of a public health surveillance system. Community policing is an organizationwide philosophy and approach that promotes partnerships, proactive problem solving, and community engagement to address crime and other community issues. Improved response to families experiencing domestic violence within this model is supported by the Violence Against Women Act (U.S. Department of Justice, 1994) since it provides funds for increased training of law enforcement on issues of domestic violence to change the police culture from thinking of this issue as a personal family matter- to viewing DVEs as a serious issue that affects families and communities (Bureau of Justice Statistics, 1998; National Coalition Against Domestic Violence, 2005). The high percentage of children exposed to domestic violence in the current study highlights the importance of developing a broad domestic violence surveillance system. Under a community policing model, police can serve as frontline public health sentinels to identify children exposed to domestic violence. However, an effective public health surveillance system requires a host of sentinels to fully capture the problem, which would include domestic violence shelter staff, CPS investigators and health care providers. An effective public health surveillance system must not only identify these children, but also connect them with appropriate public services. More police departments are recognizing the importance of collaborating with social work professionals to assist with cases that involve populations with special needs, such as suspects who are mentally ill or homeless (Patterson, 2004). These individuals often come to the attention of law enforcement but need assistance from social services. Similarly, the police are often the rst point of contact for families experiencing domestic violence. However, the potential needs of the family are likely beyond the scope of law enforcement tasks. Collaboration between social services and law enforcement would greatly increase the likelihood of victims of domestic violence receiving needed services. References
Baron, R. M., & Kenny, D. A. (1986). The moderator-mediator variable distinction in social psychological research: Conceptual, strategic, and statistical considerations. Journal of Personality and Social Psychology, 51(6), 11731182. Bradburn, N., Rips, L. J., & Shevell, S. K. (1987). Answering autobiographical questions: The impact of memory and inference in surveys. Science, 236, 157161. Carlson, B. E. (1984). Children's observations of interpersonal violence. In A. Roberts (Ed.), Battered women and their families (pp. 147167). New York: Springer. Ceci, S. J., Ross, D. F., & Toglia, M. P. (Eds.). (1989). Perspectives on children's testimony New York: Springer-Verlag. Center for Disease Control and Prevention (2006). Preventing intimate partner violence, sexual violence, and child maltreatment. Retrieved June 3, 2006 from http://www. cdc.gov/ncipc/pub-res/research_agenda/07_violence.htm Cummings, E. M., Goeke-Morey, E. M., & Raymond, J. (2004). Fathers in family context: Effects of marital quality and marital conict. In M. Lamb (Ed.), The role of the father in child development (pp. 196221). New York: John Wiley & Sons. Deater-Deckhard, K. (1998). Parenting stress and child adjustment: Some old hypotheses and new questions. Clinical Psychology: Science and Practice, 5, 314332.

DiEugenio, D., & Glass, M. (2004). The kappa statistic: A second look. Computational Linguistics, 30(1), 95101. Edleson, J. L. (1999). Children's witnessing of adult domestic violence. Journal of Interpersonal Violence, 14, 839870. Edleson, J. L., Mbilinyi, L. F., Beeman, S. K., & Hagemeister, A. K. (2003). How children are involved in adult domestic violence. Journal of Interpersonal Violence, 18, 1832. ESRI (2004). ArcView 9. Redlands, CA: ESRI GIS and Mapping Software. Famularo, R., Kinscherff, R., & Fenton, T. (1992). Parental substance abuse and the nature of child maltreatment. Child Abuse & Neglect, 16, 475483. Fantuzzo, J., Boruch, R., Beriama, A., Atkins, M., & Marcus, S. (1997). Domestic violence and children: Prevalence and risk in ve major US cities. Journal of the American Academy of Child and Adolescent Psychiatry, 36, 116122. Fantuzzo, J. W., & Fusco, R. A. (2007). Children's direct sensory exposure to domestic violence crimes. Violence & Victims, 22(2), 158171. Fantuzzo, J. W., Fusco, R. A., Mohr, W. K., & Perry, M. A. (2007). Domestic violence and children's presence: A population-based study of law enforcement surveillance of domestic violence. Journal of Family Violence, 22(6), 331340. Fick, A. C., Osofsky, J. D., & Lewis, M. L. (1997). Perceptions of violence: Children, parents, and police ofcers. In J. D. Osofsky (Ed.), Children in a violent society (pp. 261276). New York: Guilford. Gjelsvik, A., Verhoek-Oftendahl, W., & Pearlman, D. N. (2003). Domestic violence with children witnesses: Findings from Rhode Island surveillance data. Women's Health Issues, 13, 6872. Glodich, A. (1998). Traumatic exposure to violence: A comprehensive review of the child and adolescent literature. Smith College Studies in Social Work, 68, 321345. Greene, J. R. (2001). Community policing in America: Changing the nature, structure and function of the police. Policies, Processes, and Decisions of the Criminal Justice System, 3, 299370. Grych, J. H., & Fincham, F. D. (1990). Marital conict and children's adjustment: A cognitivecontextual framework. Psychological Bulletin, 108, 267290. Grych, J. H., Seid, M., & Fincham, F. D. (1992). Assessing marital conict from the child's perspective: The children's perception of interparental conict scale. Child Development, 63, 558572. Harrington, D., Dubowitz, H., Black, M. M., & Binder, A. (1995). Maternal substance use and neglectful parenting: Relations with children's development. Journal of Clinical Child Psychology, 24, 258263. Henry, B., Moftt, T. E., Caspi, A., Langley, J., & Silva, P. A. (1994). On the remembrance of things past: A longitudinal evaluation of the retrospective method. Psychological Assessment, 6, 92101. Jaffe, P. G., Wolfe, D. A., & Wilson, S. K. (1990). Children of battered women. Thousand Oaks, CA: Sage Publications. Johnson, E. I., & Waldfogel, J. (2002). Parental incarceration: Recent trends and implications for child welfare. Social Service Review, 76, 460479. Jouriles, E., McDonald, R., Norwood, W. D., & Ezell, E. (2001). Issues and controversies in documenting the prevalence of children's exposure to domestic violence. In S. A. Graham-Bermann, & J. L. Edelson (Eds.), Domestic Violence in the Lives of Children (pp. 1334). Washington, DC: American Psychological Association. Keller, P. S., Cummings, E. M., Davies, P. T., & Lubke, G. (2005). Children's behavioural reactions to marital conict as a function of parents' conict behaviours and alcohol problems. European Journal of Developmental Psychology, 4, 157177. Kelley, S. J. (1992). Parenting stress and child maltreatment in drug-exposed children. Child Abuse & Neglect, 16, 317328. Lane, J. M., & Addis, M. E. (2005). Male gender role conict and patterns of help seeking in Costa Rica and the United States. Psychology of Men and Masculinity, 6, 155168. Lee, F. (2002). The social costs of seeking help. Journal of Applied Behavioral Science, 38, 1735. Leonard, K. E., & Roberts, L. J. (1998). The effects of alcohol on the marital interactions of aggressive and non-aggressive husbands and their wives. Journal of Abnormal Psychology, 107, 602615. Margolin, G., & Gordis, E. B. (2000). The effects of family and community violence on children. Annual Review of Psychology, 51, 445479. Maxwell, C., Garner, J., & Fagan, J. (2001). The effects of arrest on intimate partner violence: New evidence from the Spouse Assault Replication Program. Washington, D.C.: Dept. of Justice. McDonald, R., Jouriles, E. N., Norwood, W., Ware, H. S., & Ezell, E. (2000). Husbands' marital violence and the adjustment problems of clinic-referred children: Intimate partner violence. Behavior Therapy, 31, 649665. National Coalition Against Domestic Violence (2005). Public policy update: VAWA 2005. Retrieved on January 8, 2006 from http://www.ncadv.org/les/AugustUpdate.pdf National Council of Juvenile and Family Court Judges (1999). Effective intervention in domestic violence and child maltreatment: Emerging programs for battered mothers and their children. Reno, NV: Author. National Institute of Justice (2007). Adolescents, neighborhoods, and violence: Recent ndings from the Project on Human Development. Retrieved on September 5, 2007 from http://www.ncjrs.gov/pdfles1/nij/217397.pdf Newman, R. (1990). Children's help-seeking in the classroom: The role of motivational factors and attitudes. Journal of Educational Psychology, 82, 7180. O'Brien, R. G. (1978). Robust techniques for testing heterogeneity of variance effects in factorial designs. Psychometrika, 43, 327342. O'Brien, M., John, R., Margolin, G., & Erel, O. (1994). Reliability and diagnostic efcacy of parents' reports regarding children's exposure to marital aggression. Violence and Victims, 9, 4562. O'Farrell, T. J., Murphy, C. M., Neavins, T. M., & Van Hutton, V. (2000). Verbal aggression among male alcoholic patients and their wives in the year before and two years after alcoholism treatment. Journal of Family Violence, 15, 295310. Osofsky, J. (1999). The effects of violence exposure on young children. American Psychologist, 50, 782788.

256

R.A. Fusco, J.W. Fantuzzo / Children and Youth Services Review 31 (2009) 249256 Straus, M. A., & Gelles, R. J. (1990). Physical violence in American families: Risk factors and adaptations to violence in 8,145 families. New Brunswick: Transaction Publishers. Takayama, J. I., Wolfe, E., & Coulter, K. P. (1998). Relationship between reason for placement and medical ndings among children in foster care. Pediatrics, 101, 201207. Tjaden, P., & Thoennes, N. (2000). Full report of the prevalence, incidence, and consequences of intimate partner violence against women: Findings from the National Violence Against Women Survey. Washington, DC: U.S. Department of Justice. U.S. Bureau of the Census (2000). Census of the Population 2000. Washington, DC: Author. U.S. Department of Justice (1994). Violence Against Women Act of 1994. Washington, DC: Author. U.S. Department of Justice (2006). Criminal victimization in the United States. Retrieved June 14, 2006 from http://www.ojp.usdoj.gov/bjs/cvict.htm U.S. Department of Justice, Bureau of Justice Statistics (1998). Criminal victimization 1998: Changes 19971998 with trends from 19931998. Retrieved February 8, 2005 from http://www.ojp.gov/bjs/pub/pdf/cv98.pdf Ware, H. S., Jouriles, E. N., Spiller, L. C., McDonald, R., Swank, P. R., & Norwood, W. D. (2002). Conduct problems among children at battered women's shelters: Prevalence and stability of maternal reports. Journal of Family Violence, 16, 291308. Werkerle, C., & Wall, A. (2002). The violence and addition equation: Theoretical and clinical issues in substance abuse and relationship violence. New York: Brunner-Routledge. Wright, L. E., & Seymour, C. B. (2000). Working with children and families separated by incarceration: A handbook for child welfare agencies. Washington, DC: Child Welfare League of America.

Osofsky, J. D., Wewers, S., Hann, D., & Fick, A. (1993). Chronic community violence: What is happening to our children? Psychiatry, 56, 3645. Parke, R.D., & ClarkeStewart, K.A. (2002). Effects of parental incarceration on young children. Working paper prepared for the From Prison to Home Conference (January 3031, 2002). Retrieved on March 29, 2006 from http://www.ncjrs.gov/App/ Publications/abstract.aspx?ID=202994 Patterson, G. T. (2004). Police-social work crisis teams: Practice and research implications. Stress, Trauma, and Crisis: An International Journal, 7, 93104. Peled, E. (1998). The experience of living with violence for preadolescent witnesses of woman abuse. Youth & Society, 29, 395430. Pynoos, R. S., & Nader, K. (1993). Issues in the treatment of posttraumatic stress in children and adolescents. In J. P. Wilson, & B. Raphael (Eds.), The international handbook of traumatic stress syndromes (pp. 535549). New York: Plenum. Rodgers, A. Y. (1998). Multiple sources of stress and parenting behavior. Children and Youth Services Review, 20, 525546. Rossman, B. B. R., Hughes, H., & Rosenberg, M. S. (1998). Children and interparental violence: The impact of exposure. Philadelphia: Brunner/Mazel. Sheu, H. B., & Sedlack, W. E. (2004). An exploratory study of help-seeking attitudes and coping strategies among college students by race and gender. Measurement and Evaluation in Counseling and Development, 37, 130143. Slep, A. M. S., & O'Leary, S. G. (2005). Parent and partner violence in families with young children: Rates, patterns, and connections. Journal of Consulting and Clinical Psychology, 73, 435444. Straus, M. (1992). Children as witness to marital violence: A risk factor to lifelong problems among a nationally representative sample of American men and women. Report of the Twenty-Third Ross Roundtable Columbus, OH: Ross Laboratories.