You are on page 1of 16

Children and Youth Services Review

27 (2005) 1167 1182


www.elsevier.com/locate/childyouth

Domestic violence and pathways into child welfare


services: Findings from the National Survey of
Child and Adolescent Well-Being
Patricia L. Kohla,*, Jeffrey L. Edlesonb, Diana J. Englishc,
Richard P. Bartha
a

University of North Carolina, Chapel Hill, School of Social Work, NC 27599 3550, United States
b
University of Minnesota, United States
c
Department of Social and Health Services, State of Washington, United States
Available online 1 July 2005

Abstract
Domestic violence (DV) and child maltreatment co-occur in many families and this may lead to
involvement with child welfare services (CWS). This study explores the role of domestic violence in
CWS using data from the National Survey of Child and Adolescent Well-Being, a large, national
probability sample of families investigated for child maltreatment. It relies on child welfare worker
report of active DV or a history of DV to examine the association of DV with child maltreatment
type, substantiation, and placement of children into out of home care. Maltreatment type
classification was similar for children with and without exposure to DV. Families with active DV
were substantiated for child maltreatment at higher rates than other groups, but DV was not a
powerful contributor to the CWS decision-making process. Families with co-occurring DV and
maltreatment often had high levels of cumulative risks, and children in families with the highest level
of cumulative risk were 10 times more likely to be placed into foster care than children in families
assessed with low levels of risk. The policy and practice implications of these findings are discussed.
D 2005 Elsevier Ltd. All rights reserved.
Keywords: Domestic violence; Child welfare services; National Survey of Child and Adolescent Well-Being

* Corresponding author.
E-mail address: pkohl@email.unc.edu (P.L. Kohl).
0190-7409/$ - see front matter D 2005 Elsevier Ltd. All rights reserved.
doi:10.1016/j.childyouth.2005.04.003

1168

P.L. Kohl et al. / Children and Youth Services Review 27 (2005) 11671182

1. Introduction
Knowledge regarding the co-occurrence of domestic violence and child maltreatment has grown over the past several years. Strong evidence indicates that children
exposed to adult domestic violence (DV) are at greater risk of also being maltreated
with reviews of the empirical literature indicating that upwards to half of children of
battered women are also physically maltreated (Appel & Holden, 1998; Edleson,
1999; OLeary, Slep, & OLeary, 2000). Co-occurring exposure to domestic violence
and child maltreatment increasingly brings families to the attention of child welfare
services (CWS). Yet, the involvement of CWS with caregivers exposed to DV is little
understood.
Available research indicates that CWS has contact with many children from homes
where DV is also occurring or has occurred. Nearly a third of 74 randomly selected families
with open CWS in one small Midwestern city were assessed by child welfare workers as
also experiencing DV (Shepard & Raschick, 1999). Child welfare workers identified DV as
a risk factor in 36% of 407 families referred for an investigation of maltreatment in New
Hampshire (Kantor & Little, 2003) and again in 36% of 383 cases investigated in a large
urban center in Minnesota (Edleson & Beeman, 1999). Finally, in a statewide study of child
protection referrals, English, Edleson, and Herrick (2005-this issue) recently found that DV
was present in almost half (47%) of cases accepted for investigation and assigned a
moderate to high level of risk.
The high level of DV in the CWS caseload is clear, but what is not so clear is how
the CWS responds to these children. In a recent landmark class action suit, the City of
New York was found to be violating mothers constitutional rights as a result of their
interventions in cases of children exposed to domestic violence. The City CWS
substantiated mothers maltreatment for bengaging in domestic violenceQ when they were
in fact victims of others violence. In some cases children were removed from their
mothers custody after such a finding by CWS. A federal court judge has subsequently
ruled that removal of children from their mothers custody solely due to the mothers
victimization was unconstitutional (Nicholson v. Williams, 203 F. Supp.2d 153, E.D.
N.Y. 2002). This suit has gained the attention of many concerned about how children
and their battered caregivers are responded to but does not provide large-scale data from
which to judge the practices of CWS.
A few studies do exist that shed light on CWS intervention practices with families in
which DV is present, including assessment of risk and differential classification of cases.
In one urban Minnesota county, Beeman, Hagemeister, and Edleson (2001) found that
child protection workers assessed cases in which domestic violence was also occurring
to be significantly more often at higher risk: 45% for such cases vs. 26% for child
maltreatment only cases. Families in which domestic violence was occurring were also
more likely to have their case opened for services: 22.6% for such cases vs. 10.4% for
child maltreatment only cases. Although they found no statistically significant
differences in overall classifications of maltreatment type (e.g., physical abuse vs.
neglect) for families, with and without DV, families with DV were significantly more
likely to be found for disregard for safety, a subtype of neglect analyzed separately due
to worker indication that this category was frequently used when DV was present in the

P.L. Kohl et al. / Children and Youth Services Review 27 (2005) 11671182

1169

home (46.4% vs. 24.1%). English et al. (this issue) found that children in families where
DV was indicated were significantly more likely to be placed out of home than were
children in non-DV families (80.7% vs. 62.3%). Somewhat contrary to these findings,
Kantor and Little (2003) found that only 8% of the families they studied with cooccurring DV and child maltreatment were classified under the maltreatment type of
neglect, failure to protect.
Domestic violence may be only one of many risk factors that influence the child welfare
workers decision-making. For example, English, Brummel, Graham, and Coghlan (2002)
found in a random sample of 2000 Washington State CWS cases that worker rankings of
DVs importance in their decision-making ranged from only 16th to 19th out of a total of 37
possible risk factors. More recently, English et al. (2005-this issue) reanalysis of these data
found that the level of DV risk was not one of the factors predicting the placement of DV
exposed children in out of home care. In another investigation of family violence, including
families with co-occurring DV and maltreatment, Shipman, Rossman, and West (1999)
found that DV occurs within ba broader context of family dysfunctionQ (p. 99), which may
account for English et al.s findings. Although no linkage to risk assessment was made,
Shipman and her colleagues add to the evidence that families with DV and child
maltreatment have high cumulative risk.
Some research findings are also available on the services delivered to CWS involved
families in which DV is present. Childrens exposure to domestic violence has been
found to be associated with a higher likelihood of a having a prior report of maltreatment
(Hazen, Connelly, Kelleher, Landsverk, & Barth, 2004) and of re-referral to CWS for
new allegations of child maltreatment (English Marshall, Brummel, & Orme, 1999; Hess,
Folaron, & Jefferson, 1992). Although recurrent maltreatment is more likely in families
with DV, the presence of DV often does not result in higher levels of intervention. Among
open cases in Beeman et al.s (2001) study, families with evidence of domestic violence
received fewer services and were more likely to be referred to the county attorney, but
their children were no more likely to be placed out of the custody of their parent(s) than
families with no known domestic violence present. English et al. (2002) in Washington
State found that DV influenced the child welfare workers offer of services; however, as
pointed out above it was not a highly ranked factor in the decision regarding case
opening.
These findings provide an initial indication that the presence of domestic violence in
child maltreatment cases is widespread but handled inconsistently across child welfare
systems. Little is known beyond the few available court cases and empirical studies
about the pathway of these families into CWS. The purpose of this study is to explore,
using a national probability sample, the role of DV in determining the pathways of
families through CWS. We were particularly interested in whether or not families in
which children who are exposed to DV and investigated for child maltreatment will be
more likely to have their children placed into out of home care or remain at home.
Overall, this study seeks to answer the following five research questions: (1) what
maltreatment classification is given to children with co-occurring child maltreatment and
exposure to domestic violence; (2) are families with child maltreatment and DV more
likely to have substantiated cases of maltreatment; (3) is DV a critical factor influencing
decisions made by the child welfare worker; (4) what is the relationship of DV to other

1170

P.L. Kohl et al. / Children and Youth Services Review 27 (2005) 11671182

family and environmental risk factors; and (5) what family characteristics, including DV,
predict placement into out of home care?

2. Methods
2.1. Description of data set and sample
This study used data from the National Survey of Child and Adolescent Well-Being
(NSCAW), a national probability sample of children and families referred to and
investigated by CPS systems. Families who underwent a complete investigation for child
maltreatment were included in NSCAW. Further, inclusion was not limited to those
families who were substantiated for maltreatment or received CWS following the index
report, but extended to families with no ongoing CWS following the investigation. The
inclusion of all families investigated for maltreatment makes it possible to compare
families who did and did not receive subsequent child welfare intervention. The sample
design involved a stratified two-stage sample, with the Primary Sampling Units (PSUs)
being county Child Protection Services agencies. The Secondary Sampling Units were
children chosen from a list of completed investigations at the sampled agencies. The entire
NSCAW cohort includes 5504 children, ages birth to 15 selected from 92 PSUs
proportionate to caseload size in 97 counties, located in 36 states. For more details about
the complex sampling design, see NSCAW Research Group (2002).
The sample for this study is 3931 caregivers of children who underwent an
investigation of child maltreatment between October 1999 and December 2000. These
children were in one of several placements following the investigation: in home without
CWS, in home with CWS, foster care, or kin care. Caregivers of children in group care or
other out of home placement were excluded from the analyses due to the small sample size
in these placement types. Caregivers with missing data on worker identification of DV
were also omitted from the analyses.
Overall, 91% of the responding caregivers are female, with 85% being the biological
mother. While the gender of the caregiver can be determined for in home cases, this same
information cannot be ascertained when the child is in an out of home placement. For out
of home cases where the respondent is a non-permanent caregiver, such as a foster parent,
only the gender of the respondent is known.
2.2. Measures
Face to face interviews were conducted with the child welfare worker. Prior to the
interview with the child welfare worker, the researchers made telephone contact with the
worker and requested that the case record be available for reference during the interview.
The interview with the worker at baseline focused on the case investigation, alleged
maltreatment, and risk assessment. In the risk assessment portion of the interview, the
worker responded to multiple questions regarding risk factors present in the home at the
time of the investigation. These crossed multiple domains, including the child, primary
caregiver, secondary caregiver, parenting behaviors, and other life stressors, such as

P.L. Kohl et al. / Children and Youth Services Review 27 (2005) 11671182

1171

difficulty paying for basic necessities. As part of the risk assessment, the worker indicated
whether active domestic violence toward the caregiver was present at the time of the
investigation. The worker also assessed for a history of domestic violence against the
caregiver. Some overlap between these two variables was present, in that in some instances
the worker indicated both active DV and a history of DV. To create distinct categories, a
derived variable was created: 1 = history of DV only, 2 = active DV only or both active and a
history of DV, and 3 = no DV. Differentiation was made between active DV and a history of
DV because prior studies have not clearly defined DV in terms of current or prior DV.
Inclusion of these distinct categories might help determine how risk assessment items
should be structured. Because data was not collected from the primary caregiver for
children in out of home placement, a caregiver report of domestic violence is unavailable
for children in out of home care; therefore reports of DV for purposes of this paper are
limited to that by the child welfare worker.
The worker also made a determination of whether or not caregivers in the home had
problems with alcohol and drugs. The child welfare worker was asked to respond yes or no
to whether the primary caregiver had a problem with active alcohol abuse. A second
question was asked about active drug abuse. These same inquiries were made about the
secondary caregiver, if one resided in the home. The two substances were combined into
one variable for purposes of this paper: active alcohol or drug abuse by primary caregiver,
and active alcohol or drug abuse by secondary caregiver. Other risk assessment items,
specific to the primary caregiver, included the presence of any serious mental health or
emotional problems, a history of recent arrests or detention in jail or prison, and a
childhood history abuse or neglect. Each of these questions was dichotomous with b1Q
indicating the presence of the risk factor and b2Q being the absence of the risk.
Following the investigation, classification of the case determination included
substantiation, indication, and neither substantiation nor indication. Substantiation is
the disposition when the allegation of maltreatment is supported by state law or state
policy and indication is the determination that there is reason to suspect maltreatment, but
it cannot be substantiated under state law (The U.S. Department of Health and Human
Services, Administration on Children, Youth, and Families, 2003). Only a small number of
states, although these include populous states such as California and Illinois, use the third
category of case determination: indicated. In addition, a few agencies in NSCAW opt for a
completely different coding system and instead use high, medium, or low risk for their
case determination following the investigation. In this study, briskQ designated cases were
included to take advantage of the additional statistical power gained over leaving them as
missing data. These codes were included in the current analysis by recoding high risk to
substantiation, medium risk to indicated, and low risk to neither substantiated nor
indicated. To ensure that this was a conservative approach, analysis excluding the briskQ
items was completed. Across all groups substantiation rates were slightly higher, and the
proportions with the case determinations of indicated or neither were slightly lower than
when these risk designations were included. Inclusion of these items does not, therefore,
result in an overestimation of substantiation.
Child welfare workers responded to questions regarding the circumstances surrounding
the investigation of the child maltreatment allegations that lead to inclusion in NSCAW.
They first indicated all maltreatment types included in the report from 10 possible

1172

P.L. Kohl et al. / Children and Youth Services Review 27 (2005) 11671182

categories: physical abuse, sexual abuse, emotional abuse, neglect: failure to provide,
neglect: failure to supervise, abandonment, moral/legal maltreatment, educational
maltreatment, exploitation, and other. Next, they deduced the most serious maltreatment
type. The workers also assessed the level of harm to the child. Child welfare workers were
asked, bregardless of the outcome of the investigation, how would you describe the level
of harm to the child. Would you say (1) none, (2) mild, (3) moderate, or (4) severe.Q The
mild and moderate categories were collapsed into one middle category of bmild or
moderate harm.Q
Families entering CWS often have multiple risks. In an effort to develop an estimate of
the cumulative risk assessed by CWS, a cumulative score was derived from 22 items on
the risk assessment: (1) prior reports of maltreatment, (2) childs poor ability to self
protect, (3) alcohol abuse by primary caregiver, (4) drug abuse by primary caregiver, (5)
alcohol abuse by secondary caregiver, (6) drug abuse by secondary caregiver, (7) primary
caregivers serious mental health problems, (8) caregiver recent arrest or detention, (9)
caregiver cognitive impairment, (10) poor parenting, (11) unrealistic expectations of
caregiver, (12) history of child abuse or neglect of the primary caregiver, (13) history of
child abuse or neglect of the secondary caregiver, (14) use of excessive discipline by the
primary caregiver, (15) use of excessive discipline by the secondary caregiver, (16)
caregiver recognition of problem and motivation to change, (17) cooperativeness of
caregiver, (18) high stress on family, (19) low social support, (20) trouble paying for basic
necessities, (21) active domestic violence, and (22) history of domestic violence. The
calculation of the cumulative risk took into account whether or not a secondary caregiver
was present in the home. The number of risks identified by the child welfare worker was
summed and then divided by the total number of items applicable to each family. This
score was categorized into three levels: low (b 20% of the total applicable risks present),
moderate (20% to b 40% of the total applicable risks present) and high (z 40% of the total
applicable risks present).
2.3. Data analysis
All analysis was done using SUDAAN to correctly adjust standard errors and account
for clustering and stratification in the complex NSCAW sampling design. The adjustments
made through weighting allow for generalizations to all families investigated for child
maltreatment nationally. A series of bivariate analyses were first carried out to determine
the association between domestic violence and several factors, including most serious
maltreatment type, substantiation rates, and other family characteristics. Additional
analyses were run to ascertain information regarding the presence of domestic violence
and the subsequent receipt of child welfare services.
Multivariate analyses were conducted to determine factors influencing placement
decisions. Multinomial logistic regression allows for the use of a dependant variable with
multiple categories and was the first multivariate method used. The impact of the
independent variables can differ for each level of the outcome (Long, 1997). The
categories of the dependent variable were in home without CWS, in home with CWS, foster
care, and kinship foster care. In home with CWS was selected as the reference group for
the outcome variable because it created a bridge between in home without CWS and the

P.L. Kohl et al. / Children and Youth Services Review 27 (2005) 11671182

1173

two categories of out of home placements. Following the preliminary analysis using
multinomial logistic regression, a binary logistic regression analysis was also performed.
For these purposes, the two in home placement settings were combined into in home
placement and foster care and kinship foster care were combined into out of home
placement. To check for multicollinearity, the variance inflation factors were obtained. For
each variable, the variance inflation factor fell below the critical cut-off of 10, ranging
from 1.1 for child age to 2.3 for the cumulative risk score. This is an indication that the
predictor variables in the model are independent of each other and appropriate for
inclusion in the model.

3. Results
3.1. Identification of domestic violence by the child welfare worker and child placement
type
Active DV was identified by the child welfare worker in 14% of the families
investigated for child maltreatment and a history of DV was indicated in 19% of the
families (not shown in table). Overall, children remained in home and did not receive
ongoing CWS in the majority of families (67%). This was slightly lower for families with
active DV identified by the worker (55%) and a history of DV (60%). Children were
placed in foster care in the smallest proportion of families; however the proportion was
higher when there was active DV (7%) and a history of DV (6%) than when the worker did
not identify domestic violence (3%). Placement with a relative was more frequent among
children exposed to active DV (11%) than children with a history of exposure to DV (5%)
or children with the absence of DV in the home (3%). Looked at from the other
perspective, among all children at home, 28% of caregivers experienced active DV
compared to 60% of caregivers of children placed in either foster family care or kinship
family care.
3.2. Classification of child maltreatment cases with co-existing domestic violence
How domestic violence is captured in maltreatment typologies is an important
consideration. One hypothesis is that families with co-existing domestic violence and
child maltreatment are being categorized as failure to provide or failure to supervise.
While a significant relationship exits between most serious maltreatment type and recent
domestic violence (v 2 = 77.17, p b 0.001, see Table 1), the proportion of cases with
failure to supervise as the most serious maltreatment type was nearly identical for
families with active DV, history of DV, and no DV. In addition, a substantially higher
proportion of children from homes without DV were classified with the most serious
maltreatment type of failure to provide (21%) than those children living in homes with
active DV (12%). Emotional maltreatment was a more frequent finding for children with
exposure to active DV than non-exposed children (23% vs. 4%, respectively), and
children in homes with a history of DV were classified with emotional abuse as the most
serious maltreatment type more frequently (8%) than those children with no DV in the

1174

P.L. Kohl et al. / Children and Youth Services Review 27 (2005) 11671182

Table 1
Comparison of most serious maltreatment type by groupa
Most serious maltreatment type

Total

Active DV

History of DV

No DV

Physical abuse
Sexual abuse
Emotional abuse
Failure to provide
Failure to supervise
Abandonment
Moral/legal maltreatment
Educational maltreatment
Exploitation
Other
Total

26.9
11.3
7.4
19.2
27.3
1.5
0.6
1.5
0.1
4.3
100.1

22.7
3.0
23.1
12.1
27.9
1.4
1.2
1.5
0.2
7.1
100.2

25.1
15.9
8.0
17.2
29.6
1.3
0.3
0.1
0
2.6
100.1

28.2
11.8
3.9
21.3
26.4
1.6
0.5
1.9
0.2
4.3
100.1

Totals do not equal to 100 due to rounding.


a
v 2 = 77.17, p b 0.001; most serious maltreatment type and status of DV are significantly related.

home (4%). Active DV was distributed fairly evenly across the three most serious
maltreatment types: failure to supervise (28%), physical abuse (23%), and emotional
maltreatment (23%).
Analyses were then conducted to examine whether emotional maltreatment was being
coupled with failure to provide and failure to supervise cases. Emotional maltreatment was
identified in 12% of all cases and 29% of cases with DV (not shown in table). When the
child welfare worker reported active DV, emotional maltreatment co-existed with failure to
provide in 9% of the cases and 5% of the failure to supervise cases. It appears that these
cases are not being classified as failure to provide or supervise either alone or in
conjunction with emotional maltreatment at a substantial rate.
3.3. Substantiation rates for families with co-occurring DV and child maltreatment
Substantiation rates by domestic violence indicators were determined and are presented
in Table 2. The differences among the three groups: active DV, history of DV, and no DV
are significant (v 2 = 25.76, p b 0.001). While just over a quarter (28%) of all cases are
substantiated, more than half (52%) of the maltreatment investigations of families with
active DV were assigned the case determination of substantiation compared to 29% of
those with a history of DV and 22% with no domestic violence identified by the child
welfare worker.
Table 2
Comparison of case determination of child maltreatment investigations by domestic violencea

Substantiated
Indicated
Neither
Total

Total

Active DV

History of DV

No DV

27.8
9.3
63.0
100.1

52.3
10.7
37.0
100.0

29.2
8.6
62.3
100.1

22.1
9.2
68.7
100.0

Totals may not equal to 100 due to rounding.


a
v 2 = 25.76, p b 0.001; case determination and status of DV are significantly related.

P.L. Kohl et al. / Children and Youth Services Review 27 (2005) 11671182

1175

3.4. Contribution of domestic violence in child welfare worker decision-making process


The growing recognition of the linkage between domestic violence and child
maltreatment brings new import to understanding the place of information about DV in
the decision-making of child welfare workers. In NSCAW, the workers were asked to
identify the two most critical factors that influenced decisions about how to proceed with
the case. Active domestic violence was one of 28 factors the worker could consider. As
shown in Table 3, according to the workers, DV contributed to the their decision in a
small percentage of cases. It was rarely determined to be a critical factor in decisionmaking even when active or a history of DV had been identified by the child welfare
worker on the risk assessment. When examining this by child setting, both active DV and
a history of DV were most frequently used in the decision-making process when children
remained in the home with CWS or were placed into kin care. It appears that domestic
violence is used in conjunction with other risk factors to determine that some ongoing
services are needed.
3.5. Relationship of domestic violence to other risk factors
The association between recent domestic violence and other family characteristics is
presented in Table 4. While no significant difference was found among the groups for
child age, the groups differed across many of the family characteristics. DV was more
often associated with prior reports of maltreatment. Families with a history of DV had the
highest rate of prior reports (67%) compared to active DV (57%), and no DV (46%).
Fewer children exposed to DV were assessed at a level of no harm (21%); while more of
these children are rated at mild/moderate harm (66%) and severe harm (13%). In
comparison, more children without exposure to domestic violence were assessed at a level
of no harm (51%), and fewer were rated at mild/moderate harm (44%) and severe harm
(5%).
Consistent with the literature on domestic violence and other risk factors, a larger
proportion of homes with DV in this sample included primary and secondary caregivers
with substance abuse problems. In addition, larger proportions of primary caregivers with
mental health issues, history of recent arrest, and a history of childhood abuse or neglect
were found in families with active DV and a history of DV than those families without DV.
More of these families also have difficulty paying for basic necessities. Overall, as
indicated by the cumulative risk count, families in which the caregiver has experienced
recent DV had more risk factors present than other families.
Table 3
Proportion of cases with domestic violence identified as a critical factor in CWW decision-making
DV reported by CWW

Total

Child setting at baseline


In home, no CWS

In home, with CWS

Foster Care

Kin Care

Active DV
History of DV
No DV
Total

2.4
5.1
2.7
3.1

2.2
2.4
2.4
2.4

2.1
11.2
3.2
4.9

1.6
2.7
2.0
2.1

5.0
2.1
6.3
5.1

1176

P.L. Kohl et al. / Children and Youth Services Review 27 (2005) 11671182

Table 4
Association between domestic violence and other case characteristics

Child age
02
35
610
11+
Most serious maltreatment type
Physical abuse
Sexual abuse
Neglect: Failure to provide
Neglect: Failure to supervise
Other
Prior reportsa
Level of harm
No harm
Mild/Moderate
Harm
Severe harm
Caregiver education
None
High School Equiv.
High School Diploma
Voc Tech Cert/Diploma
Associate Degree
RN Diploma
Bachelors Degree
Masters Degree
M.D., Ph.D., Law, Dental
Other
Substance abusePrimary CG
Substance abuseSecondary CG
Mental health problemPrimary CG
History of arrestsPrimary CG
History of CANPrimary CG
History of CANSecondary CG
Physically impairedPrimary CG
Poverty
b 50%
5099%
100149%
150199%
z 200%
Pay basics
Cumulative risk count
Low
Medium
High

Active DV, % Yes

History of DV, % Yes

No DV

23.7
23.4
35.5
17.4

17.3
20.6
33.4
28.7

18.5
20.2
38.5
22.9

24.4
3.2
13.0
31.6
27.8
57.4

25.7
16.3
17.6
31.7
8.7
67.3

29.4
12.3
22.2
29.3
6.7
46.0

21.1

37.3

50.8

65.7
13.2

50.6
12.2

44.4
4.7

26.4
22.4
20.3
16.1
6.9
1.5
4.9
0.5
0.3
0.7
30.9
51.2
25.2
24.6
30.5
20.0
9.2

27.2
21.6
25.6
12.4
5.7
0.1
4.0
0.5
0.2
2.8
17.9
20.6
19.6
19.2
45.8
23.5
4.3

30.7
15.1
27.3
14.6
4.6
0.2
4.5
0.4
0.3
2.4
8.5
7.5
11.7
8.0
13.0
8.6
4.1

14.8
32.7
22.3
11.4
18.8
36.1

22.2
34.2
17.8
7.5
18.4
33.9

22.7
26.6
17.6
13.1
20.1
18.2

13.1
39.8
47.1

19.8
45.2
35.0

65.4
27.3
7.4

v2
10.6

38.5***

26.3***
38.8***

21.8

35.9***
50.7***
19.0**
36.0***
53.4***
15.0**
3.6
13.9

36.3***
150.8***

P.L. Kohl et al. / Children and Youth Services Review 27 (2005) 11671182

1177

3.6. Domestic violence and other factors as predictors of child setting


Preliminary multivariate analyses were then conducted for the four categories of child
settings (in home no CWS, in home with CWS, foster care, and kinship foster care) as a
dependent variable in the model. The results of the multinomial logistic regression did not
show significant differences between in home with CWS and in home without CWS.
Kinship care and foster care also had similar findings in comparison to the reference
group. To simplify the presentation of the data, logistic regression was conducted with the
outcome variable coded as in home vs. out of home placement. The predictors in the
model included indication of DV and the significant variables presented in Table 4.
Although child age was not significant in the bivariate analysis, it was included for
conceptual reasons. While significant group differences were found for substance abuse by
the secondary caregiver, this item could not be included in the model due to singularity
problems. This problem arose due to the fact that not all families have secondary
caregivers in the home creating too many empty cells in the model.
Overall, DV alone did not appear to be strongly associated with whether the child was
placed or remained at home. While children of caregivers with a history only of DV were
less likely to be removed from the home than those children without exposure to DV, active
DV was non-significant in the model (see Table 5). Other characteristics of children and
families undergoing investigations of maltreatment were more influential in placement
decisions than DV. The preschool age children (age 35) were less likely to be placed in an
out of home placement (foster care or kinship care) compared to the oldest children (11+).
The level of harm assessed by the child welfare worker was a significant predictor of child
setting. Children with severe levels of harm were nearly nine times more likely to be
removed from their home than those with no harm. Those children experiencing the most
severe level of harm were receiving higher levels of intervention (i.e., placement in foster or
kinship care). Substance abuse by primary caregivers also increased the likelihood of a
placement. Children of primary caregivers with a substance abuse problem were three and a
half times more likely to be in out of home care than children of non-substance abusing
parents. Finally, the cumulative risk count was a significant predictor of where children
resided. Children with medium and high risk were much more likely to be in out of home
care than to remain in the home with their caregiver. Children placed with either a relative or
non-relative in comparison to remaining in home with services were four times more likely
to have medium risk than low risk and 10 times more likely to have high risk than low risk.
Children with the highest levels of risk were being provided with higher levels of protection.
While active DV was not a significant predictor of placement setting when controlling
for the other case characteristics, the cumulative risk count was a strong predictor. As
noted in Table 4, 47% of all families with active DV reported by the child welfare worker
Notes to Table 4:
CAN = Childhood abuse and neglect. CG = caregiver.
*p b 0.05.
a
For dichotomous variables, the proportion of families with an affirmative response is reported. The chi-square
tests the association between responses of yes and no.
*** p b 0.001.
** p b 0.01.

1178

P.L. Kohl et al. / Children and Youth Services Review 27 (2005) 11671182

Table 5
Variables predicting child setting: in home vs. out of homea
Predictor variables
Domestic violence
Active DV
History only**
No DV
Child age
02
35*
610
11+
Most serious maltreatment typeb
Physical abuse
Neglect: Failure to provide
Neglect: Failure to supervise
Other
Prior reports of maltreatmentc
Level of harm
No harm
Mild/moderate harm
Severe harm***
Substance abuse by PCG***
PCG with serious mental illness
PCG with history of CAN
PCG with physical impairment
Trouble paying for basis necessities
Cumulative risk count
Low risk
Medium risk**
High risk***

Odds ratio

OR (95% CI)

0.94
0.54
Reference Group

0.50, 1.76
0.34, 0.84

1.4
0.42
0.78
Reference group

0.72, 2.73
0.19, 0.90
0.46, 1.34

Reference group
1.4
1.7
1.5
0.79

0.73,
0.91,
0.82,
0.52,

2.66
3.03
2.62
1.19

Reference group
1.4
8.9
3.6
1.1
0.99
1.23
1.49

0.83,
4.51,
2.33,
0.69,
0.66,
0.61,
0.83,

2.40
17.73
5.50
1.59
1.50
2.49
2.66

Reference group
4.0
9.8

1.66, 9.42
3.78, 25.23

Group care and other out of home care excluded from the analysis due to small sample size.
Sexual abuse, abandonment, moral/legal, educational maltreatment, exploitation and bOtherQ included in
this category.
c
Unless otherwise specified reference group is absence of the risk.
* p b 0.05.
** p b 0.01.
*** p b 0.001.
b

had high cumulative risk scores in comparison to 37% of families with a history of DV
only and 7% of families with no domestic violence. This analysis confirms English et als
(2005-this issue) state-level findings that families with domestic violence entering the
child welfare system seemed to be burdened with multiple problems that contributed to the
CWS response, independent of their DV experience.

4. Discussion
The aim of this study was to explore the role of domestic violence in investigations of
child maltreatment and important information was learned about families in child welfare

P.L. Kohl et al. / Children and Youth Services Review 27 (2005) 11671182

1179

services that have domestic violence identified by the child welfare worker. First, it is
important to point out that some limitations could not be overcome due to the nature of the
data. Because we did not have caregiver reports of DV for families in which children went
into out of home care, we almost certainly under identified the level of DV in CWS (Kohl,
Barth, Hazen, & Landverk, 2005-this issue). When considering caregiver report, DV is
more prevalent than the 14% identified by the workers in this study. In addition, male
caregivers represented 9% of the sample. The authors acknowledge that the DV
experiences of males and females may be very different; however, the data for children
in out of home care could not be limited to female caregivers. Limited information is
available about secondary caregivers who may very well be the perpetrator of both the DV
and child maltreatment.
Identification of DV by child welfare workers in NSCAW is substantially lower than in
studies cited in the literature, which suggest that child welfare workers indicate the
presence of DV in at least a third of families investigated for child maltreatment. Previous
study of this area has typically been limited to one geographic area. A higher rate of
identification may have been in response to voices collectively raised in concern over the
problem in the targeted area. In addition, the selection criteria used by CWS in the risk
assessment process vary by states. The difference in rates found between this national
probability sample and samples from limited geographic regions is that our study included
data from multiple states, some of which may have lower identification rates than states
which have focused on the specific problem of DV in families with child maltreatment.
Despite these limitations, the findings from this study provide several important
implications for policy and practice. Some discussion in the literature (e.g., Kantor &
Little, 2003) has focused on the role of failure to protect in families with co-occurring DV
and child maltreatment and this can be a point of contention between CWS and domestic
violence advocates. One central concern is that blame is being placed on battered mothers
through the classification of maltreatment based on neglect (even though efforts to protect
the child could be at substantial risk to the welfare of the mother). Yet, when DV is
present, the most serious maltreatment type was evenly distributed in this national sample
across three of the 10 codes: failure to supervise, physical abuse, and emotional
maltreatment, with a smaller proportion being coded as sexual abuse or neglect: failure to
provide. These results indicate that DV cases are not being classified primarily as failure to
provide or failure to supervise. Further, DV cases being distributed in similar proportions
across three maltreatment types suggest that, although children living in homes with active
DV are more likely to be classified under emotional abuse, this category does not appear to
be a proxy for cases with DV. Classifications are similar to cases in which DV is not
indicated.
Families with a history of DV have a greater likelihood of a prior report than both
families with active DV and those with no such risk. One possible explanation for this is
that prior maltreatment occurred during periods when active DV was present in the home.
In addition, families with both active DV and a history DV are more often assessed at
higher levels of harm than those children without this exposure. Evidently, child welfare
workers perceive that children in homes where DV occurs or has occurred in the past
experience higher levels of harm. These findings suggest that risk assessment items should
differentiate between active and past DV exposure.

1180

P.L. Kohl et al. / Children and Youth Services Review 27 (2005) 11671182

Although a higher proportion of families with DV had their maltreatment allegations


substantiated and were more likely to have prior reports of maltreatment, DV is rarely
mentioned as a primary influence on worker decision-making. Workers considered
information on numerous risk factors as they made decisions about families during the
investigation process. In addition, families with DV tended to experience multiple risks
and other risks may have taken a more prominent role in the decision-making process.
Given the additional risk that exposure to DV may represent for the child, factoring DV
into the decision-making process when it has been identified by the worker seems
warranted. Consideration should be given to the particular risk DV poses to families in
CWS, which may differ across families. Our study found that families experiencing DV
were more often substantiated and assessed at higher levels of harm than families with no
DV. Therefore, workers identification of and decision-making about families in which DV
has occurred should be more carefully examined. A more thorough assessment of the risks
presented by DV should in turn provide better-informed decision-making by child welfare
workers.
The practice implications extend beyond the information provided about substantiation
and placement decisions. Case planning processes can also be informed from these
findings. In examining the association between DV and other risk factors, differences
clearly exist among the three groups: active DV, history of DV, and no DV. Substance
abuse and mental health problems of the primary caregiver are more prevalent in the group
with active DV identified by the worker. Most importantly, a larger proportion of this
group has the highest cumulative risk count indicating that domestic violence is one of
many problems experienced by these families. These findings can help child welfare
workers better understand the dynamics of DV, including the possible interaction among
DV, mental health issues, substance abuse problems, and childhood history of maltreatment. Because the findings suggest that it is the cumulative nature of the risks in these
families that is related to level of CWS involvement, our findings suggest that case
planning should include services that address multiple issues in these families. While not
intending to minimize the importance of parenting interventions, the narrowly constructed
and brief versions now used by CWS agencies (Barth et al., in press) are unlikely to
sufficiently address the general risk patterns of families with DV.
Future research should focus on how to better understand the co-existing problems and
risk in families where DV and child maltreatment co-occur. If families experiencing DV in
CWS can be combined into multivariate group profiles we may be able to better inform
child welfare practice decisions. For instance, we may be able to determine how the needs
of families with one mix of problems and risk factors compare to those of families with
another mix and subsequently design more tailored interventions to suit these families
needs. The re-report experiences of children with co-occurring DV and child maltreatment
in this national, probability sample should also be explored. Families with DV are more
likely to have had a report prior to the allegations which led to the inclusion in NSCAW;
however whether or not they are also more likely to experience new allegations of
maltreatment needs to be further examined as this has additional implications for the
pathway of families with co-occurring DV and child maltreatment through CWS.
In summary, this study found that DV is often one of many risk factors in families
entering CWS and appears to have only a minor role in influencing the decisions of child

P.L. Kohl et al. / Children and Youth Services Review 27 (2005) 11671182

1181

welfare workers; yet, children exposed to DV often have multiple contacts with CWS due
to the higher number of repeat allegations of maltreatment. Children with co-occurring DV
and child maltreatment were not being systematically classified as neglect and had similar
patterns of maltreatment classification as children without exposure to DV. Further, these
children did not have a higher rate of placement into out of home care. Ironically, children
in families experiencing DV were more often substantiated for maltreatment, assessed at
higher levels of harm and had higher levels of cumulative risks in their lives. These
findings suggest that the growing attention among policy makers about the important
intersection of domestic violence and child welfare has not resulted in the widespread
placement of children who are exposed to domestic violence. The absence of a strong
relationship between the presence of DV and the delivery of CWS is surprising given our
findings about the association between DV and substantiation, harm and cumulative risk.
Clearly, these data indicate the need for careful attention to the development of practice
protocols for how to effectively intervene when DV is identified in families entering CWS.

Acknowledgements
NSCAW is funded by the Administration for Children and Families of the U.S.
Department of Health and Human Services; conclusions are those of the authors and do
not necessarily represent the official position of DHHS. Support for this paper was
provided by a Childrens Bureau pre-doctoral fellowship (Kohl) and the National Institute
on Justice funded CAD-V project (Edleson, English, and Barth).

References
Appel, A. E., & Holden, G. W. (1998). The co-occurrence of spouse and physical child abuse: a review and
appraisal. Journal of Family Psychology, 12, 578 599.
Barth, R. P., Landsverk, J., Chamberlain, P., Reid, J., Rolls, J., Hurburt, M., et al. (in press). Parent training in
child welfare services: Planning for a more evidence based approach to serving biological parents. Research
on Social Work Practice.
Beeman, S. K., Hagemeister, A. K., & Edleson, J. L. (2001). Case assessment and service receipt in families
experiencing both child maltreatment and woman battering. Journal of Interpersonal Violence, 16, 437 458.
Edleson, J. L. (1999). The overlap between child maltreatment and woman battering. Violence Against Women, 5,
134 154.
Edleson, J. L., & Beeman, S. K. (1999). Responding to the co-occurrence of child maltreatment and adult
domestic violence in Hennepin County. Minneapolis, MN7 University of Minnesota available online at http://
www.mincava.umn.edu/link
English, D. J., Brummel, S. C., Graham, J. C., & Coghlan, L. K. (2002). Factors that influence the decision not to
substantiate a CPS referral: Phase II. Mail and telephone surveys of child protective service social workers.
Olympia, WA7 Department of Social Services, Childrens Administration, Management Services Division,
Office of Childrens Administration Research.
English, D. J., Edleson, J. L., & Herrick, M. E. (2005). Domestic violence in one states child protective caseload:
A study of differential case dispositions and outcomes. Children and Youth Services Review, 27, 1183 1201
(this issue).
English, D. J., Marshall, D. B., Brummel, S., & Orme, M. (1999). Characteristics of repeated referrals to child
protective services in Washington State. Child Maltreatment, 4(4), 297 307.

1182

P.L. Kohl et al. / Children and Youth Services Review 27 (2005) 11671182

Hazen, A., Connelly, C. D., Kelleher, K., Landsverk, J., & Barth, R. (2004). Intimate partner violence among
female caregivers of children reported for child maltreatment. Child Abuse and Negect, 28, 301 319.
Hess, P. M., Folaron, G., & Jefferson, A. B. (1992). Effectiveness of family reunification services: an innovative
evaluation model. Social Work, 31, 304 311.
Kantor, G. K., & Little, L. (2003). Defining the boundaries of child neglect: when does domestic violence equate
with parental failure to protect? Journal of Interpersonal Violence, 18, 338 355.
Kohl, P. L., Barth, R. P., Hazen, A., & Landsverk, J. (2005). Child welfare as a gateway to Domestic Violence
Services. Children and Youth Services Review, 27, 1203 1221 (this issue).
Long, J. S. (1997). Regression models for categorical and limited dependent variables. Thousand Oaks, CA7
Sage.
Nicholson v. Williams, 203 F. Supp.2d 153 (E.D. N.Y. 2002).
NSCAW Research Group. (2002). Methodological lessons from the National survey of child and adolescent WellBeing: The first three years of the USAs First National probability study of children and families investigated
for abuse and neglect. Children and Youth Services Review, 24, 513 541.
OLeary, K. D., Slep, A. M. S., & OLeary, S. G. (2000). Co-occurrence of partner and parent aggression:
research and treatment implications. Behavior Therapy, 31, 631 648.
Shepard, M., & Raschick, M. (1999). How child welfare workers assess and intervene around issues of domestic
violence. Child Maltreatment, 4, 148 156.
Shipman, K. L., Rossman, B. B. R., & West, J. C. (1999). Co-occurrence of spousal violence and child abuse:
conceptual implications. Child Maltreatment, 4, 93 102.
U.S. Department of Health and Human Services. (2003). Administration on children, youth, and families. Child
Maltreatment 2001. Washington, DC7 U.S. Government Printing Office.