Caregiver Substance Abuse and Children’s Exposure to Violence in a Nationally Representative Child Welfare Sample

Abstract

Using data from the National Survey of Child and Adolescent Well-Being II (NSCAW II), this article examines the impact of caregiver substance abuse on children’s exposure to violence in the home in a nationally representative sample of families involved with child protective services (CPS). Logistic regression analyses indicate an increased risk of witnessing mild and severe violence in the home for children whose primary caregiver was abusing alcohol or drugs. However, analyses did not find statistically significant relationships between child report of direct victimization in the home by mild or severe violence and caregiver alcohol or drug abuse.

With over 1.5 million maltreatment reports assessed by child protective services (CPS) in 2010 (US DHHS, 2011) and high rates of substance abuse among caregivers involved with child welfare, deepening our understanding of the impact of caregiver substance abuse on children’s development and well-being is an important step in addressing the needs of the child welfare population. Although prevalence estimates vary greatly depending on sample characteristics, a general consensus is that between 50 and 80 percent of caregivers in the child welfare population are engaged in substance abuse (Besinger, Garland, Litrownik, & Landsverk, 1999; Jones, 2004; US Government Accounting Office [GAO], 1994; GAO 1998). The use of substances by a caregiver has been linked to negative outcomes for children and adolescents including substance use, eating disorders, teenage pregnancy, poor academic attendance and performance and suicidality (Chandy, Blum, & Resnick, 1996; Corvo & Carpenter, 2000). Additionally, emerging evidence supports the existence of a correlation between caregiver substance abuse and children’s exposure to violence (Ondersma, Delaney-Black, Covington, Nordstrom, & Sokol, 2006; Sprang, Clark & Staton-Tindall, 2010). This analysis builds upon prior research by examining the relationship between caregiver substance abuse and children’s exposure to violence in a nationally representative sample of families investigated for child maltreatment.

Exposure to violence describes two types of experiences: direct victimization and witnessing violence (Finkelhor, Turner, Ormrod, & Hamby, 2009). Both types of exposure during childhood place children at greater risk for adverse proximal and distal outcomes related to traumatic stress. Children who directly experience violence through abuse or other situations have a greater likelihood of experiencing traumatic symptomology during childhood or adolescence (Boney-McCoy & Finkelhor, 1995; Fowler, Tompsett, Braciszewski, Jacques-Tiura, & Baltes, 2009). Experiencing abuse or violence in childhood may also increase the likelihood of subsequently experiencing PTSD in adulthood (Brewin, Andrews, & Valentine, 2000; Hetzel & McCanne, 2005; Kulkarni, Graham-Bermann, Rauch, & Seng, 2011; Widom, 1999).

The relationship between witnessing violence as a child and subsequently experiencing PTSD is less straight forward (Feerick & Haugaard, 1999; Fowler et al., 2009; Kulkarni et al., 2011). The majority of the literature on children witnessing violence focuses on witnessing intimate partner violence (IPV). With child physical abuse estimated to co-occur in 45-70% of the families in which IPV is occurring, disentangling the relationship between direct victimization and witnessing violence and the impact each has on developing PTSD is complicated (Edleson, 1999; Holt, Buckley, & Whelan, 2008). While Feerick and Haugaard (1999) found that women who witnessed IPV in childhood were more likely to report adult symptoms of PTSD even after controlling for childhood experiences of abuse, Kulkarni et al. (2011) found that witnessing domestic violence in childhood only predicted adult PTSD among women who also experienced childhood abuse. However, a meta-analysis completed by Fowler et al. (2009) found that PTSD symptoms were equally predicted by experiencing violence, witnessing violence, or hearing about community violence. While many studies focus on children witnessing IPV, emerging research reveals that witnessing community violence is also traumatic and harmful to child development. Findings from the Longitudinal Studies of Child Abuse and Neglect (LONGSCAN) demonstrated that a child’s lifetime prevalence of exposure to community violence or IPV was associated with aggression, depression, and anxiety (Litrownik, Newton, Hunter, English, & Everson, 2003).

Although research indicates that exposure to violence may lead to negative consequences for children, it is still uncertain what impact caregiver substance abuse has on children’s exposure to violence, particularly among child welfare samples. Preliminary studies indicate that children of substance using caregivers may be at an increased risk for experiencing trauma. In a sample of 407 African American mothers and their 6 to 7 year old children, caregiver alcohol abuse, the child’s observation of drug use in the home, and the child’s observation of drug deals were all significantly correlated with the child’s exposure to violence (Ondersma et al., 2006). Reviewing a random sample of case records for 1127 families with open child welfare cases in a Southern state, Sprang et al. (2010) found significant correlations between caregiver drug use (methamphetamine and other drugs) and the child’s exposure to traumatic events. Sprang et al.’s results indicate that children of caregivers who use drugs, particularly those who use methamphetamines, are more likely to witness intimate partner violence, experience child endangerment, and experience child physical abuse than families with no record of substance use or misuse.

Prior research studies have examined caregiver substance abuse using data from NSCAW I (Berger, Slack, Waldfogel, & Bruch, 2010; Carter, 2010; Phillips & Detlaff, 2009; Phillips, Leathers, & Erkanli, 2009). Berger et al. (2010) examined the relationship between caseworker-identified substance abuse and case outcomes. The researchers found that cases with caseworker-identified substance abuse had poorer case outcomes including higher levels of caseworker-reported risk and harm to the child and more frequent substantiation of child maltreatment and removal of the child from the home. Other studies have examined the role caregiver substance abuse plays in specific populations involved with CPS. Examining risk factors for the out-of-home placement of urban American Indian and Alaskan Native children, Carter (2010) found that urban American Indian and Alaskan Native children placed in out-of-home care were more likely to come from homes where caregiver substance abuse or mental health problems were present than White children. Prior research has also examined the prevalence of caregiver substance abuse among who have previously been arrested and among caregivers on probation (Phillips & Detlaff, 2009; Phillips, Leathers, & Erkanli, 2009). Due to the recent release of the National Survey of Child and Adolescent Well-Being II (NSCAW II) data, caregiver substance abuse and children’s exposure to violence have not been explored in the NSCAW II beyond caregiver substance abuse prevalence rates reported in the baseline reports (Ringeisen, Casanueva, Smith, & Dolan, 2011). Using data from NSCAW II, this analysis examines the relationship between caregiver substance abuse and children’s exposure to violence, both witnessing and experiencing, in a nationally representative sample of families involved with child protective services.

The Ecological-Transactional Model of Child Maltreatment (Cicchetti & Lynch, 1993; Cicchetti & Valentino, 2006) proposes that the child, the caregiver, and the environment have a reciprocal influence on each other which impacts child development and long term outcomes. The theory proposes that potentiating and compensatory risk factors related to child maltreatment are present in each ecological system (macrosystem, exosystem, microsystem, and individual). These risk factors impact the transactions between the child, the family, and the environment. Using the Ecological-Transactional Model of Child Maltreatment, it is proposed that caregiver substance abuse serves as a microsystem potentiating risk factor which has reciprocal interactions with the child’s development, family interactions, and factors at the exosystem level. Furthermore, caregiver substance abuse may serve as an additional challenge in vulnerable child welfare families which tips the balance of the potentiating and compensatory factors resulting in children experiencing negative outcomes like exposure to violence, as well as long term consequences such as the development of PTSD. Based on this model, even among a sample of vulnerable children, many of whom have experienced maltreatment, the experience of violence exposure will further increase the likelihood of negative outcomes.

This study examines the correlation between caregiver substance abuse and children’s exposure to violence in a nationally representative sample of families involved with child protective services (CPS). This study tests the following hypotheses:

1. Among families involved with child welfare, children of substance abusing caregivers will self-report higher rates of witnessing violence in the home than children of caregivers who are not abusing substances.

2. Among families involved with child welfare, children of substance abusing caregivers will self-report higher rates of direct victimization in the home than children of caregivers who are not abusing substances.

METHODS

Data from the National Survey of Child and Adolescent Well-Being II (NSCAW II) was analyzed to examine separately the relationships between alcohol use and child-report of violence exposure and drug use and child-report of violence exposure. The NSCAW II is the second in a series of national data collection studies of families investigated by CPS for child maltreatment. The total nationally representative sample consists of 5,872 children between the ages of 0 and 17.5 years and their families, who were investigated for child maltreatment between February 2008 and April 2009. Children were located in 81 primary sampling units PSUs nested within eight sampling strata at the state level. Seven of the 8 strata are the states with the largest child welfare caseloads in the nation. The remaining stratum consists of all remaining states in the sample. A complex weighting strategy accounting for stratification, clustering, weighting, and oversampling of some subgroups was developed to make the PSUs nationally representative. Only one child per family was included in the NSCAW II. Therefore, children are not nested within caregivers. Additional information about the NSCAW II dataset, study design, and data collection methods can be found elsewhere (see National Data Archive on Child Abuse and Neglect [NDACAN], 2010).

This secondary data analysis includes 1,652 children from the original sample of 5,872 who were at least 8 years of age at the time of the report. Age was used as an exclusion criterion because the Violence Exposure Scale for Children was only administered to children 8 to 17 years of age. For this analysis, baseline data collected from caseworkers, the primary caregiver, and the child during in-person interviews were utilized. Caseworkers were interviewed at their agencies and were able to refer to their notes and case records during the interview.

Exposure to Violence

The dependent variable for this analysis, the child’s self-report of witnessing or experiencing violence in the home, was measured with the Violence Exposure Scale for Children (VEX-R; Fox & Leavitt, 1995). The VEX-R measures children’s exposure to violent and criminal acts in the home through the use of questions with cartoon illustrations. In the NSCAW II study, the cartoon illustrations and example questions (e.g., How many times have you watched TV?) were only used with children ages 8 to 10 (NDACAN, 2010). Children 11 to 18 were asked the questions which assess for violence exposure but they were not shown cartoon illustrations or asked the simple example questions in order to make the measure developmentally appropriate for older children.

To provide a sense of privacy to the children and increase the validity of the data, the VEX-R and other measures about sensitive topics were administered through Audio Computer-Assisted Self Interview (ACASI) for children 11 and older. Each child used headphones to listen to the audiotaped VEX-R questions and responded on a computer. Children 8 to 10 listened to the audiotaped VEX-R questions while wearing headphones and then pointed to their answers on cards held by the interviewer. Prior to starting the ACASI or audiotaped questions, the interviewer let each child know that the interviewer would not know his or her answers but responses indicating that the child’s life or health was in danger would be reported to authorities to ensure the child’s safety. Each child listened to sets of questions assessing whether or not he or she had witnessed or experienced the following acts committed by adults toward another person in a home they had lived in: yelling, throwing something at, pushing or shoving hard, slapping hard, beating up, pointing a knife or gun, or the spanking of another child or themselves. Children were also asked if in a home they had lived in they had witnessed but not experienced someone stealing from another person, stab another person with a knife, shoot another person with a gun, someone being arrested, or a person dealing drugs. Factor analyses of the VEX-R with groups of 2^nd and 4^th graders indicate that questions on the measure group into two categories: mild violence and severe violence (Raviv et al., 2001). Severely violent behaviors included threatening with a knife or gun, stabbing, shooting, witnessing an arrest or drug deal, and stealing.

In a sample of minority preschoolers (Shahinfar, Fox, & Leavitt, 2000), the VEX-R was found to have moderate to good levels of inter-item reliability for children’s reports of exposure to violence (Cronbach’s α = .80 – .86). Internal consistency in the NSCAW sample is high for the total sample (Cronbach’s α = .96) and for the subscales (Cronbach’s α = .86 – .92) (NDACAN, 2010). Shahinfar et al. (2000) found modest correlations between children’s distress symptoms and scores on the VEX-R for witnessing mild violence (r = .29, p < .05), experiencing mild violence (r = .22, p < .05), and witnessing severe violence (r = .25, p < .05).

Substance Abuse

The independent variable of caregiver substance abuse was measured with caseworker responses to two separate questions on a risk assessment. Prior to administering the risk assessment questions, interviewers informed the caseworkers that they would be answering questions about the strengths and impairments of the caregivers. Caregiver alcohol abuse was dichotomized (yes or no) based on the CPS caseworker’s response to the question, “At the time of the investigation, was there active alcohol abuse by [the primary caregiver]?” Caregiver drug abuse was dichotomized in the same manner based on the caseworker’s response to the question, “At the time of the investigation, was there active drug abuse by [the primary caregiver]?” Primary caregivers were predetermined based on who was living in the home and providing care to the child. If the mother was one of these caregivers, she was considered the primary caregiver. This individual was identified by name in each of the above questions. Caseworkers were able to refer to all notes, documentation, and records for the case. Caseworkers were interviewed an average of 134 days after the investigation was completed (NDACAN, 2010). Caseworkers in the NSCAW II sample had an average of 7.1 years (median 5.0) of experience working in the Child Welfare System (Dolan, Smith, Casanueva, & Ringeisen, 2011). Almost one-quarter had a master’s degree, 21.9% had a bachelor’s degree in Social Work, and 52.3% had a bachelor’s degree in a field other than social work (Dolan et al., 2011).

Control Variables

Child characteristics

The child characteristics of age, gender, and race/ethnicity were all controlled for in the analysis based on their significance in prior studies. Age was used as a continuous variable indicating the child’s age in years. Prior examinations of the relationship between child age and exposure to violence indicate that age has a significant relationship with rates of exposure to violence but that the direction of this relationship has been inconsistent. Findings of some studies indicate older children report higher rates of exposure (Buka, Stichick, Birdthistle, & Earls, 2001; Richters & Martinez, 1993; Selner-O’Hagan, Kindlon, Buka, Raudenbush, & Earls, 1998) while others indicate that exposure to violence does not always have a positive linear relationship with age as some younger children report higher rates of exposure (Schwab-Stone et al., 1995).

Child gender was a dichotomous categorical variable with children classified as either male or female. Although a number of studies indicate that male children are more likely to witness and experience violence than female children (Buka et al., 2001; Schwab-Stone et al., 1995; Selner-O’Hagan et al., 1998; Singer, Anglin, Song, & Lunghofer, 1995), a study by Attar, Guerra, and Tolan (1994) found that young elementary school girls reported higher rates of violence exposure than boys.

Child race/ethnicity was operationalized as a dichotomous categorical variable with non-Hispanic Caucasians being the comparison for all other races and ethnic groups. Race/ethnicity was also chosen as a control variable based on research indicating that ethnic minority children, both African American and Hispanics, are at higher risk for exposure to violence (Selner-O’Hagan et al., 1998; Singer et al., 1995).

Caregiver age

The caregiver characteristic of age was selected as a control variable. Caregiver age was used as a continuous variable indicating the caregiver’s age in years at time of entry into the study.

Income

Household income was used as a control variable in the analysis based on its consistent association with children’s exposure to violence. Household income was operationalized as primary caregiver report of the total household income combining all adult incomes. This variable was dichotomized into households with a combined annual income of $20,000 or more and households with a combined annual income of less than $20,000.

Data Analysis Strategy

Using logistic regression, baseline data from the NSCAW II was analyzed to examine separately the relationships between alcohol use and child-report of violence exposure and drug use and child-report of violence exposure. First, univariate analyses were conducted to determine sample characteristics. Next, chi-squares and t-tests were conducted to examine the relationships between the exposure to violence variables and child age, child gender, child race/ethnicity, caregiver age, and household income. Chi-squares were also conducted to examine the relationship between the exposure to violence variables and caregiver alcohol and drug use. Finally, a total of eight separate logistic regression models were run for each of the four types of violence exposure (witnessing mild violence, witnessing severe violence, direct victimization with mild violence, direct victimization with severe violence) with each type of substance abuse (alcohol abuse and drug abuse). Data were analyzed in Stata/SE 12.1, and accounted for stratification, clustering and weighting.

RESULTS

Sample Characteristics

In this sample (n = 1652) of children 8 years and above, the mean age was slightly less than 12 years of age (M = 11.99; SD = 0.11) with a range of 8 to 17 years. There were fewer males in the sample than females (45.13% vs. 54.87%). Non-Hispanic Caucasian children made up 42.69% of the sample with 57.31% of the sample identifying as Non-Hispanic African American, Hispanic, Asian, American Indian or Alaskan Native, or Native Hawaiian or Pacific Islander. Characteristics of the analyzed sample can be seen in Table 1.

Table 1.

Univariate Analyses

Variable	Mean or Percentage	SD
Child/Caregiver Variables
Child age	11.99 years	.11
Child gender	45.13% male 54.87% female
Child race	42.69% white 57.31% other
Caregiver age	38.76 years	.35
Household Income
Continuous Income	$27,877	$1533.21
Dichotomous Income	50.58% income $20,000+ 49.42% income less than $20,000
Caseworker report of caregiver alcohol abuse	96.02% no  3.98% yes
Caseworker report of caregiver drug abuse	94.19% no  5.81% yes

Caregivers in the sample had a mean age of 38 years, 9.5 months (M = 38.76, SD = 0.35). Approximately half of the caregivers identified as Non-Hispanic Caucasian (49.81%) with the other 50.19% identifying as Non-Hispanic African American, Hispanic, Asian, American Indian or Alaskan Native, or Native Hawaiian or Pacific Islander. The mean annual household income for the sample was $27,877.00 (SD = $1,533.21) with 50.58% of the sample having a household income greater than or equal to $20,000.

Caseworkers reported that 5.81% of the sample had an active drug abuse problem at the time of the investigation. This was higher than the 3.98% of the sample that caseworkers reported had an active alcohol abuse problem at the time of the investigation.

Violence Exposure

Results (see Table 2) indicate that 91.92% of all children in the sample had experienced some type of violence in the home either directly as a victim or indirectly as a witness.

Table 2.

Child Self-Reported Exposure to Violence in the Home

Variable	Percentage
Witnessed Mild Violence	85.80% Yes 14.20% No
Witnessed Severe Violence	54.73% Yes 45.27% No
Victimized by Mild Violence	83.29% Yes 16.71% No
Victimized by Severe Violence	5.21% Yes 94.79% No
Any Exposure to Mild Violence	90.59% Yes 9.41% No
Any Exposure to Severe Violence	54.78% Yes 45.22% No
Any Exposure to Mild or Severe Violence	91.92% Yes 8.08% No

Of the total sample, 90.59% had been exposed to one or more incident of mild violence by witnessing, being victimized, or both. Of the total sample, 85.80% of children reported witnessing one or more incident of mild violence and 83.29% reported experiencing one or more incident of direct victimization with mild violence. Over half of the children in the total sample (54.78%) reported experiencing one or more incident of direct victimization with severe violence. Over half of the total sample (54.73%) reported witnessing one or more incident of severe violence and 5.21% of the total sample reported experiencing one or more incident of direct victimization with severe violence.

Bivariate Analyses

The chi-square test indicated that active alcohol abuse by a caregiver is associated with the child witnessing mild violence, χ²(1) = 18.49, p < .001. A higher percentage of children of caregivers with active alcohol abuse (96.11%) reported witnessing mild violence in the home compared to children whose caregivers were not actively abusing alcohol (86.17%). A chi-square test indicated that active alcohol abuse by a caregiver is associated with the child witnessing severe violence, χ²(1) = 58.00, p < .01. A higher percentage of children of caregivers with active alcohol abuse (79.69%) reported witnessing severe violence in the home compared to children whose caregivers were not actively abusing alcohol (54.00%). However, chi-square tests indicated that there was no statistically significant association between active alcohol abuse by a caregiver and experiencing direct victimization with either mild violence or severe violence.

The chi-square test indicated that active drug abuse by a caregiver is associated with the child witnessing mild violence, χ²(1) = 18.42, p < .05. A higher percentage of children of caregivers with active drug abuse (94.46%) reported witnessing mild violence in the home compared to children whose caregivers were not actively abusing drugs (86.17%). A chi-square test indicated that active drug abuse by a caregiver is associated with the child witnessing severe violence, χ²(1) = 56.50, p < .01. A higher percentage of children of caregivers with active drug abuse (75.28%) reported witnessing severe violence in the home compared to children whose caregivers were not actively abusing drugs (53.98%). However, chi-square tests indicated that there was no statistically significant association between active drug abuse by a caregiver and experiencing direct victimization with either mild violence or severe violence.

There was a significant effect for child age on experiencing direct victimization by mild violence, F(1, 72) = 5.98, p < .05, with children who had been directly victimized by mild violence being older (M = 12.28 years, SD = 0.11) than children who had not (M = 11.66 years, SD = 0.25). However, there was no significant effect for child age on witnessing mild or severe violence or on experiencing direct victimization with severe violence.

Chi-square tests indicated that there was no statistically significant association between child gender and experiencing direct victimization or witnessing violence or between child race/ethnicity and experiencing direct victimization or witnessing violence.

There was a significant effect for caregiver age on witnessing severe violence, F(1, 72) = 10.63, p < .01, with children who had witnessed severe violence in the home having an older primary caregiver (M = 39.83, SD = 0.46) than children who had not (M = 37.58, SD = 0.54). No statistically significant differences in caregiver age were found for witnessing mild violence, experiencing direct victimization with mild violence, or experiencing direct victimization with severe violence.

There was a significant effect for household income on witnessing mild violence, F(1, 72) = 21.14, p < .001, with children who had witnessed mild violence in the home coming from homes with higher annual household incomes (M = $30,028.45, SD = $1852.29) than children who had not witnessed mild violence in the home (M = $18,724.65, SD = $1565.14). There was also a significant effect for household income on experiencing direct victimization with mild violence, F(1, 72) = 12.16, p < .001, with children who had experiencing direct victimization with mild violence in the home again coming from homes with higher annual household incomes (M = $29,848.56, SD = $1913.24) than children who had not experiencing direct victimization with mild violence in the home (M = $21,384.40, SD = $1437.34). No statistically significant differences in annual household income were found for witnessing severe violence or direct victimization with severe violence.

Regression Analyses

Alcohol abuse and exposure to violence

The model regressing active alcohol abuse by a caregiver on children witnessing mild violence was statistically significant, F(6, 67) = 4.23, p < .01 (Table 3). Controlling for other variables in the model, the odds of a child of an alcohol abusing caregiver witnessing mild violence are 3.22 times the odds of a child whose caregiver was not abusing alcohol (t = 2.54, p < .05). Controlling for other variables in the model, the odds of a child witnessing mild violence increased 2.1% for each $1000 increase in annual household income (t = 4.13, p < .001). Child age, child gender, child race, and caregiver age were not statistically significant in this model.

Table 3.

Logistic Regression Results for Models Regressing Active Alcohol Abuse by Caregiver on Exposure to Violence

Model	Variable	OR	SE	t	CI	Model F
Active Alcohol						4.23**
Abuse on Witnessing Mild	Active Alcohol Abuse	3.22	1.48	2.54*	1.29 – 8.07
Violence	Child Age	1.03	0.05	0.61	0.93 – 1.14
	Child Gender	1.06	0.24	0.24	0.67 – 1.66
(N = 1160)	Child Race	0.61	0.20	−1.54	0.32 – 1.16
	Caregiver Age	1.01	0.01	0.51	0.98 – 1.04
	Income	1.000021	5.02*10-6	4.13***	1.000011 – 1.000031
Active Alcohol						3.97**
Abuse on Witnessing	Active Alcohol Abuse	3.22	1.37	2.75**	1.38 – 7.50
Severe Violence	Child Age	1.04	0.04	0.93	0.96 – 1.11
	Child Gender	0.72	0.13	− 1.83	0.50 – 1.03
(N = 1158)	Child Race	1.12	0.20	0.64	0.78 – 1.61
	Caregiver Age	1.03	0.01	2.74**	1.01 – 1.05
	Income	0.9999964	2.89*10-6	− 1.24	0.9999906 – 1.000002
Active Alcohol						3.00*
Abuse on Mild Violence	Active Alcohol Abuse	2.55	1.36	1.76	0.89 – 7.36
Victimization	Child Age	1.05	0.05	0.92	0.95 – 1.16
	Child Gender	0.81	0.18	− 0.97	0.52 – 1.25
(N = 1159)	Child Race	0.77	0.20	− 0.99	0.45 – 1.31
	Caregiver Age	1.00	0.01	− 0.30	0.97 – 1.02
	Income	1.000015	4.61*10-6	3.33**	1.000006 – 1.000025
Active Alcohol						1.79
Abuse on Severe Violence	Active Alcohol Abuse	2.75	1.58	1.76	0.88 – 8.67
Victimization	Child Age	1.01	0.13	0.10	0.79 – 1.30
	Child Gender	0.53	0.24	− 1.40	0.22 – 1.31
(N = 1146)	Child Race	0.43	0.23	− 1.60	0.15 – 1.23
	Caregiver Age	0.98	0.02	− 1.02	0.95 – 1.02
	Income	1.000002	4.89*10-6	0.34	0.9999919 – 1.000011

^*Note:p < .05

^**p < .01

^***p < .001

The model regressing active alcohol abuse by caregiver on children witnessing severe violence was statistically significant, F(6,67) = 3.97, p < .01 (Table 3). Controlling for other variables in the model, the odds of a child of an alcohol abusing caregiver witnessing severe violence are 3.22 times the odds of a child whose caregiver was not abusing alcohol (t = 2.75, p < .01). Controlling for other variables in the model, the odds of a child witnessing severe violence increased 2.9% for each one year increase in caregiver age (t = 2.74, p<.01). In this model, child age, child gender, child race, and household income were not statistically significant.

The model regressing active alcohol abuse by a caregiver on children experiencing direct victimization with mild violence was also statistically significant, F(6,67) = 3.00, p < .05. However, caregiver alcohol abuse was no longer statistically significant after controlling for the other variables in the model. The only statistically significant variable in the model was household income. After controlling for other variables in the model, the odds of a child experiencing direct victimization with mild violence increased 1.5% for each $1000 increase in annual household income (t = 3.33, p < .01).

The model regressing active alcohol abuse by a caregiver on children experiencing direct victimization with severe violence was not statistically significant (Table 3).

Drug abuse and exposure to violence

The model regressing active drug abuse by caregiver on children witnessing mild violence was statistically significant, F(6,67) = 4.95, p < .001 (Table 4). After controlling for other variables in the model, the odds of a child of a caregiver who was abusing drugs witnessing mild violence were 2.41 times the odds of a child whose caregiver was not abusing drugs (t = 2.03, p < .05). After controlling for other variables in the model, the odds of a child witnessing mild violence increased 2.0% for each $1000 increase in annual household income (t = 3.52, p < .01). In this model, child age, child gender, child race, and caregiver age were not significant.

Table 4.

Logistic Regression Results for Models Regressing Active Drug Abuse by Caregiver on Exposure to Violence

Model	Variable	OR	SE	t	CI	Model F
Active Drug						4.95***
Abuse on Witnessing Mild Violence	Active Drug Abuse	2.41	1.04	2.03*	1.02 – 5.71
	Child Age	1.06	0.05	1.22	0.96 – 1.17
(N = 1173)	Child Gender	1.15	0.25	0.63	0.74 – 1.76
	Child Race	0.66	0.21	− 1.31	0.36 – 1.24
	Caregiver Age	1.00	0.01	0.01**	0.97 – 1.03
	Income	1.00002	5.72*10-6	3.52	1.000009 – 1.000032
Active Drug						3.73**
Abuse on Witnessing Severe Violence	Active Drug Abuse	2.39	0.81	2.56*	1.21 – 4.71
	Child Age	1.05	0.04	1.37	0.98 – 1.12
(N = 1170)	Child Gender	0.73	0.13	− 1.75	0.51 – 1.04
	Child Race	1.16	0.20	0.86	0.83 – 1.62
	Caregiver Age	1.03	0.01	2.74**	1.01 – 1.05
	Income	1.00	2.90*10-6	− 1.32	0.9999904 – 1.000002
Active Drug					2.18
Abuse on Mild Violence Victimization	Active Drug Abuse	1.11	0.38	0.30	0.56 – 2.19
	Child Age	1.07	0.06	1.37	0.97 – 1.19
(N = 1172)	Child Gender	0.85	0.17	− 0.77	0.57 – 1.28
	Child Race	0.83	0.21	− 0.74	0.49 – 1.39
	Caregiver Age	0.99	0.01	− 0.78	0.97 – 1.01
	Income	1.000015	4.88*10-6	3.02**	1.000005 – 1.000024
Active Drug						1.79
Abuse on Severe Violence Victimization	Active Drug Abuse	0.81	0.51	− 0.33	0.23 – 2.85
	Child Age	1.02	0.13	0.18	0.80 – 1.31
(N = 1157)	Child Gender	0.51	0.23	− 1.51	0.21 – 1.24
	Child Race	0.39	0.21	− 1.75	0.14 – 1.14
	Caregiver Age	0.98	0.01	− 1.10	0.95 – 1.01
	Income	1.000002	5.17*10-6	0.33	0.9999914 – 1.000012

^*Note: p < .05

^**p < .01

^***p < .001

The model regressing active drug abuse by a caregiver on children witnessing severe violence was statistically significant, F(6,67) = 3.73, p < .01. Controlling for other variables in the model, the odds of a child of caregiver who was abusing drugs witnessing severe violence are 2.39 times the odds of a child whose caregiver was not abusing drugs (t = 2.56, p < .05). Controlling for other variables in the model, the odds of a child witnessing severe violence increased 2.8% for each one year increase in caregiver age (t = 2.74, p<.01). In this model, child age, child gender, child race, and annual household income were not significant.

The model regressing active drug abuse by a caregiver on children experiencing direct victimization with mild violence and the model regressing active drug abuse by a caregiver on children experiencing direct victimization with severe violence were not statistically significant (Table 4).

DISCUSSION

Children involved with the child welfare system are reporting high levels of exposure to violence in the home. The results of this analysis indicate that the vast majority of children (91.92%) eight and older who are involved with the child welfare system are reporting witnessing or experiencing at least one incident of violence in the home. Children reported high rates of both witnessing mild violence (85.80%) and experiencing victimization with mild violence (83.29%) in their homes. Of extreme concern, over half of the children in the sample (54.73%) reported witnessing at least one incident of severe violence in the home. These behaviors include witnessing an adult threaten another person with a knife or gun, witnessing a stabbing or shooting, witnessing an arrest or drug deal, and witnessing an adult steal from another person. Witnessing episodes of severe violence in childhood may increase the likelihood of experiencing PTSD in adulthood especially among a sample of children in which child maltreatment is prevalent (Feerick & Haugaard, 1999; Fowler et al., 2009; Kulkarni et al., 2011). While only 5.21% of children eight and older reported experiencing victimization in the home by severe violence, these children are highly likely to experience short and long term consequences due to this exposure (Boney-McCoy & Finkelhor, 1995; Brewin et al., 2000; Fowler et al., 2009; Hetzel & McCanne, 2005; Kulkarni et al., 2011; Widom, 1999). The prevalence of violence exposure among children in the child welfare system is concerning and likely contributes to the negative outcomes seen in this population.

Even among a population of vulnerable children at high risk for violence exposure, substance abuse by a caregiver was found to further increase the likelihood of witnessing both mild and severe violence in the home. Both alcohol abuse by a caregiver and drug abuse by a caregiver were associated with a higher likelihood of children witnessing mild and severe violence in the home. These findings were consistent in both bivariate and regression analyses. The hypothesis that among families involved with child welfare, children of caregivers who abuse substances will self-report higher rates of witnessing violence in the home than children of caregivers who are not abusing substances was supported. This finding is consistent with literature citing the high co-occurrence of intimate partner violence and substance abuse (Holt et al., 2008).

Results indicated that neither active alcohol abuse nor active drug abuse significantly increased the likelihood of child self-reports of experiencing mild or severe violence in the home. Therefore, the hypothesis that among families involved with child welfare, children of caregivers who abuse substances will self-report higher rates of direct victimization in the home than children of caregivers who are not abusing substances was rejected. Although prior studies have found a significant relationship between caregiver substance abuse and children experiencing violence (Sprang et al., 2010), caregiver substance abuse was not significantly related to higher rates of violent victimization among children 8 and older in the NSCAW II. This finding was consistent in both bivariate and logistic regression analyses. The sample of CPS-involved families in this study is unique from that of Sprang et al (2010). In the Sprang et al. study, all children (n = 1127) were sampled from child protective service records in the Southern United States. The children were an average age of 5.1 years, predominantly White (71.3%), and the sample had slightly more males (52.6%) than females. Additionally, a large portion of the caregivers engaged in drug use in the Sprang et al. (2010) sample were using methamphetamines. With a mean age of 11.99 years, the sample of our current study (n = 1652) is older and has a higher percentage of females (54.87%) than Sprang et al. The racially diverse sample has a lower percentage of White children (42.69%) than Sprang et al. Although both Sprang et al. (2010) and the current study involve samples of CPS-involved families, the differences in sample characteristics may influence the relationship between caregiver substance abuse and children experiencing violence. Despite the finding that caregiver substance abuse did not significantly increase self-reported victimization rates in families with active caregiver substance abuse, it is important to recognize that children are reporting high levels of victimization in both families with and without caregiver substance abuse.

Although significant relationships between child gender and exposure to violence have been found in samples of urban elementary school-aged African American and Hispanic children (Attar et al., 1994), community samples of middle and high school students (Schwab-Stone et al., 1995; Singer et al., 1995), and children and young adults in Chicago (Selner-O’Hagan et al., 1998), this analysis did not find a statistically significant relationship between child gender and exposure to violence in a nationally representative sample of children 8 to 17 involved with the child welfare system. Prior studies have also found significant associations between child race/ethnicity and exposure to violence in community samples of high school students (Singer et al., 1995) and children and young adults in Chicago (Selner-O’Hagan et al., 1998). However, this analysis did not find a statistically significant relationship between child race/ethnicity and exposure to violence. The difference in findings may be due to differences between the community samples in prior studies and the CPS-involved families in the current study. It is also possible that child gender and child race/ethnicity are associated with exposure to violence among certain subsamples of the child welfare population. The study completed by Attar et al. (1994) found that gender was associated with exposure to violence when interacting with ethnicity and when interacting with grade level. For example, first and second grade girls reported higher exposure to violence than fourth grade boys but there were no significant differences with first and second grade boys or fourth grade girls. Although no interactions were evident in the current study, further analyses could examine the relationship between child gender and exposure to violence among subsets of the sample in the current study.

Bivariate analyses found that children who reported experiencing direct victimization by mild violence were older (M = 12.28 years, SD = 0.11) than children who did not report ever experiencing this type of exposure to violence (M = 11.66 years, SD = 0.25). This finding is not surprising considering that children were asked if they have ever experienced this type of violence. Older children have had more time to acquire negative experiences. Interestingly, age was not found to differentiate children who had witnessed mild or severe violence or experienced severe violence. Future analyses should examine the number of reports of each type of violence exposure and its relationship with child age in the NSCAW II sample.

Bivariate and logistic regression analyses indicated a significant relationship between caregiver age and witnessing severe violence in the home. Children who reported witnessing severe violence in the home had older primary caregivers than children who did not witness severe violence in the home. Although rates of IPV have previously been found to be lower for older cohorts of women than younger cohorts of women (Rennison & Rand, 2003), the sample of primary caregivers in this study are relatively close in age (M = 38.76, SD = .35) and the caregiver age was examined continuously. Future analyses could further examine the relationship between caregiver age and children’s exposure to violence in the NSCAW II dataset by comparing a multinomial categorical variable of caregiver age to child’s exposure to violence.

A surprising finding was the relationship between annual household income and children’s exposure to violence. Results from bivariate analyses indicate children who reported witnessing mild violence or experiencing direct victimization with mild violence had higher annual household incomes than children who did not report these types of exposure to violence. Logistic regression analyses supported the relationship between higher income and witnessing mild violence in the home. This finding may reflect the generally low level of income for child welfare families (M = $27,877, SD = $1533.21). Consistent with the Ecological-Transactional Model of Child Maltreatment (Cicchetti & Lynch, 1993; Cicchetti & Valentino, 2006), families with higher incomes, which serve as a compensatory factor, may need to display much higher levels of distress and need in order to become involved in the CPS system. These findings indicating higher levels of exposure to mild violence in the home may be one potentiating factor initiating CPS involvement. Social workers should consistently assess for intimate partner violence among families involved with CPS regardless of household income.

Limitations

This analysis adds to the literature on the relationship between caregiver substance abuse and children experiencing direct victimization and witnessing violence in the home by examining these concepts in a nationally representative sample of child welfare families with children 8 to 17 years of age. However, it is limited by its inability to draw conclusion on the relationship between caregiver substance abuse and children’s exposure to community violence. As a secondary data analysis, the authors were limited to the data collected by the original research team. This analysis utilized caseworker-report of caregiver alcohol and drug use rather than caregiver self-report. It is possible that caseworkers did not detect all active substance abuse by the primary caregivers. It is also possible that caseworkers reported active substance abuse for some caregivers who were not actively engaged in substance abuse. Although caregiver self-report is available for a sample of families in NSCAW II, caseworker report of alcohol and drug use was chosen for this analysis because it is available for families at all levels of CPS involvement including families with children in the foster care system. Additionally, caseworkers were able to examine all case records when reporting whether or not the caregiver was involved in substance abuse. Although further detail is not available to examine how caseworkers determined that substance abuse was present, data available on the caseworkers indicates most were experienced child welfare professionals.

Although the VEX-R was developed to assess for exposure to violence in samples of children ages 4 to 10, the measure was administered to children 8 to 17 in NSCAW II. The illustrations and example questions were not used when children were over the age of 10. It is possible that the use of this measure in an older group of children may impact the validity of this measure. Children self-reported their exposure to violence. Even with the use of the ACASI technique, some children may have chosen to not report all episodes of witnessed or experienced violence. After being informed by the interviewer that responses indicating the child’s health or safety were endangered would be reported to CPS, some children may have chosen not to report incidents which would meet these criteria.

Additionally, some critiques of the NSCAW II data include its use of eight strata with seven strata representing the states with the largest child welfare caseloads in the nation. The remaining stratum consists of all remaining states in the sample. Some researchers feel that the NSCAW II dataset best represents the seven states with the largest child welfare caseloads in the country than it represents the remaining states. However, the intricate and complex use of weighs is what makes the NSCAW II dataset nationally representative (see NDACAN, 2010).

Implications

Implications for social work practice can be drawn from this analysis. First, when working with children from homes where active substance abuse by a caregiver is present or a history of substance abuse exists, social workers should gather information to assess the child’s current exposure to violence and history of exposure to violence. Workers should assess for current or previous exposure to violence as both a direct victim and a witness. Second, when working with caregivers involved with CPS who are engaged in substance abuse, practitioners should be cognizant of the possibility of intimate partner violence in the home, assess for its presence and impact on children when safe to do so, and consider appropriate resources for the family. Third, when children report witnessing violence, experiencing victimization, or any trauma-related symptoms, interventions to address trauma symptoms must be utilized to prevent immediate and long-term negative outcomes.

CONCLUSION

This study adds to the literature by providing the first examination of the relationship between caregiver substance abuse and children’s exposure to violence in a large, diverse, and nationally representative sample of families reported to CPS. This analysis, utilizing recently collected data, highlights the importance of assessing for exposure to violence among children involved with the child welfare system, particularly those where caregiver substance abuse is present. An additional wave of data is currently being collected for the NSCAW II study. In addition to the future directions already discussed, future research should examine the relationship between caregiver substance abuse and exposure to violence over time.