Conduct Problems Among Children at Battered ...

Journal of Family Violence, Vol. 16, No. 3, 2001

Conduct Problems Among Children at Battered Women's Shelters: Prevalence and Stability of Maternal Reports Holly Shinn Ware/ Ernest N. Jouriles,! "* Laura C. Spiller,' Renee McDonald/ Paul R. Swank,^^ and William D. Norwood^

The present research was designed to (1) replicate prevalence estimates of clinical levels of conduct problems in a large (n — 401) sample of children residing at a shelter for battered women, and (2) assess the stability of mothers' reports of child conduct problems following shelter departure. According to mothers' reports on standardized questionnaires and diagnostic interviews obtained during shelter residence, approximately one third of the children between 4 and 10 years of age exhibited clinical levels of conduct problems. Prior research has demonstrated elevated maternal distress during shelter residence and suggests that such distress may influence mothers' reports of child conduct problems. To examine this issue, a subset of families with children exhibiting clinical levels of conduct problems fn = 68) was reassessed following their shelter departure. Mothers' reports of child conduct problems remained stable despite significant reductions in mothers' distress after shelter exit. KEY WORDS: child conduct problems; externalizing problems; battered women; wife abuse.

Violence toward women by their intimate partners is now recognized as a widespread problem that shatters the lives of millions of Americans each year (see Straus & Gelles, 1990a, for national prevalence estimates). Although the most obvious victims of domestic violence are women, their children suffer as well. These children have been referred to as the "unintended" 'Department of Psychology, University of Houston, Houston, Texas. ^Department of Educational Psychology, University of Houston, Houston, Texas. ^Present address: School of Nursing, University of Texas - Houston. "•TO whom correspondence should be addressed at Department of Psychology, University of Houston, 4800 Calhoun Road, Houston, Texas 77204-5341; e-mail: [email protected]. 291 0885-7482/01/0900-0291 $19.50/0 © 2001 Plenum Publishing Corporation

292

Ware, Jouriles, Spiller, McDonald, Swank, and Norwood

or "invisible" victims (Osofsky, 1998; Rosenbaum & O'Leary, 1981). Many studies now indicate that children of battered women display elevated levels of aggressive, defiant, and violent behavior (collectively referred to as conduct or externalizing problems) compared to children in nonviolent homes (see reviews by Holtzworth-Munroe et al, 1998; Kolbo et al, 1996). In fact, clinical levels of these problems have been reported to occur in as many as a third of children whose mothers have sought refuge at shelters for battered women (e.g., Christopoulous et al, 1987; Hughes & Luke, 1998; O'Keefe, 1994). Thesefindingsare of potentially great importance because persistent conduct problems during childhood predict a variety of serious adjustment problems later in life (Loeber & Hay, 1997; Moffitt, 1993). Identifying children at risk for future problems while they are living in shelters may be particularly advantageous because shelter residence often provides a unique opportunity for clinical intervention (Jouriles et al, 1998; Sullivan & Bybee, 1999). However, several methodological issues require attention before conclusions about the conduct problems of these children can be made with confidence. Most studies that have examined adjustment in children of battered women are based on small samples of violent families (i.e., less than 50 shelter families) and typically include families from only one shelter. Such studies are particularly vulnerable to sampling error, reducing the likelihood of obtaining accurate estimates of the prevalence of children's conduct problems. Most of these studies also have relied solely on the Child Behavior Checklist (CBCL; Achenbach, 1991; Achenbach & Edelbrock, 1983) to determine whether children exhibit clinical levels of problems. Although the CBCL possesses excellent psychometric properties, and is widely used among researchers, it is not universally accepted as a criterion for denoting clinical levels of problems, nor is administration of a questionnaire universally accepted as a sufficient methodology for that purpose. In addition to the issues noted earlier, in most studies of children of battered women, mothers provided data about their children's adjustment at a single point in time: during shelter residence. This is potentially problematic because the very high levels of psychological distress experienced by many mothers during their shelter residence (Christopoulous et al, 1987; Hughes & Luke, 1998; Moore & Pepler, 1998; Wolfe et al, 1986) may infiate their reports of their children's externalizing problems (Hughes & Barad, 1983). This hypothesis is consistent with theory and empirical work suggesting that maternal distress can cause mothers to evaluate their children's behavior more negatively than they otherwise might (e.g., Dix, 1991; Forehand et al, 1987; Jouriles & Thompson, 1993; Richters, 1992). In fact, several studies suggest that within child-clinical samples, mothers' distress substantially influences their reports of children's externalizing problems (e.g., Brody

Conduct Problems Among Children of Battered Women

293

& Forehand, 1986; Schaughency & Lahey, 1985). This potential distortion of mothers' reports is of particular concern given the extent to which researchers rely on mothers for information about their children's adjustment. Mothers are often regarded as the single best source for information concerning young children's conduct problems (Loeber et al, 1991; Richters, 1992), and, in battered women's shelters, obtaining information from other adults who are familiar with a child's behavior is often impossible because of the family's circumstances.^ Thus, to increase confidence in the assertion that the children of sheltered battered women are at increased risk for externalizing problems, it is important to show that mothers' reports of high levels of child externalizing problems persist even after mothers' distress declines. Several investigators have found that women's distress declines significantly following departure from battered women's shelters (Campbell etal, 1995; Holden etal, 1998; Wolfe etal, 1986). It is not clear, however, whether there is a similar decline in mothers' reports of their children's externalizing problems. To our knowledge, only two published studies have examined mothers' reports of children's externalizing problems and maternal distress both during and after shelter residence (Holden et al, 1998; Wolfe et al, 1986). In a cross-sectional study, Wolfe et al (1986) assessed mothers' reports of their own distress and mothers' reports of their children's behavior problems across three groups of families: (1) current residents of a shelter (n = 17), (2) former shelter residents (at least 6 months after shelter departure; n = 23), and (3) a nonviolent community group {n = 23). Mothers residing at shelters reported themselves to have more health problems (including somatic complaints, anxiety, social dysfunction, and depression) than did mothers in the former resident group; however, these two groups did not differ on mothers' reports of child externalizing behavior problems. Thus, the results of this study suggest that reductions in mothers' distress are not associated with concomitant reductions in child externalizing problems as reported by mothers. In a longitudinal study (n = 32), Holden et al (1998) assessed mothers' distress and mothers' reports of their children's externalizing problems during shelter residence and again 6 months after shelter exit. Mean levels of mothers' parenting stress and depressive symptoms, as well as mothers' reports of their children's externalizing behavior ^While women and children are in residence at a women's shelter, collecting data from fathers is often unfeasible and unethical because these women are often escaping or hiding from the fathers. Collecting data from teachers also is not practical because of the need to keep the residence of the mothers and children confidential. Furthermore, children often change schools upon shelter entry, and new teachers usually have had insufficient time to observe the children and report reliably on child problems. Although shelter staff are potentially another source of information, children in shelters often have limited contact with them, reducing the likelihood that such staff can provide credible data on child behavior problems.

294


problems, were lower after shelter departure. Thus, the Holden et al. results are consistent with the hypothesis that maternal distress during shelter residence may account, at least in part, for mothers' reports of children's externalizing problems. In summary, the available data do not converge; therefore, the extent to which the high levels of child externalizing problems reported by battered women during shelter residence reflect a stable perception of child problems versus mothers' contemporaneous distress is unknown. In the present research, we first attempted to replicate findings regarding the prevalence of child externalizing problems within women's shelter samples. In doing so, however, we addressed some of the methodological limitations of past research in this area by recruiting a relatively large sample, sampling from multiple shelters, and gathering evidence of the representativeness of our sample. We also interviewed mothers and assessed for conduct problems using DSM-IV diagnostic criteria (i.e., Oppositional Defiant Disorder (ODD) and Conduct Disorder (CD); American Psychiatric Association (APA), 1994) to supplement mothers' reports on a standard child behavior checkhst, the CBCL (Achenbach, 199t). As noted earlier, children with clinical levels of conduct problems are at risk for continued problems. For children whose mothers report clinical levels of conduct problems during residence at battered women's shelters, it is important to know the extent to which these problems are stable, and to rule out the hypothesis that these reports are a product of mothers' contemporaneous distress. Consequently, we collected longitudinal data on a subsample of families in which the children were judged to be displaying clinical levels of conduct problems. Within this subsample, we evaluated whether similarly high levels of child externalizing problems were reported by mothers after shelter departure, when mothers' level of psychological distress was expected to be lower. We also examined changes in the relation of mothers' distress to mothers' reports of children's externalizing problems during and after shelter residence. Finally, although the focus of this research was on mothers' reports of child externalizing problems, we also explored changes in mothers' reports of children's internalizing problems (e.g., anxiety, depression), another commonly reported dimension of problems among children of battered women.

METHOD Participants Four hundred and one mothers, each with a child between 4 and 10 years of age (M = 5.5, SD = 1.86; 221 boys, 180 girls), were participants in this


295

research. Eligibility criteria for participating in the in-shelter assessment included (1) the woman sought refuge at a shelter because of physical violence by an intimate male partner, and she reported at least one incident of physical violence on the spousal Conflict Tactics Scale (CTS, Straus, 1979) in the past year; (2) the woman had at least one child between the ages of 4 and 10 years; and (3) the woman could communicate sufficiently in English to complete the interview and questionnaires.* In each family, the youngest child between 4 and 10 years of age was selected for participation. Mothers' mean age was 30.0 (SD — 6.01). They had an average of 11.5 (SD = 2.15) years of education and an average of 2.8 (SD — 1.39) children. Mothers had been in relationships with their battering partners an average of 6.6 (SD =4.83) years. The median family income prior to entering shelters was approximately $18,260 (M = 22,889; SD = 25,141). For approximately 65% of the families, the reported income included governmental aid (e.g., AFDC/TANF and food stamps). Approximately 50% of participating children were biologically related to the batterers. The ethnic composition of this sample was 38% Caucasian, 36% African American, 24% Hispanic, and 2% other. Sixty-six percent of the mothers reported that they were beaten up, 38% reported that they were threatened with a knife or gun, and 15% reported that their partners used a knife or gun against them during the year prior to shelter entry. Sixty-eight of the 401 families comprised the longitudinal subsample and were reassessed after shelter departure. Families were eligible for the postshelter assessment if (1) the mother reported that at least one of her children (aged 4-10 years) exhibited behavior problems sufficient to meet D5M-/y criteria for ODD or CD (APA, 1994), (2) the mother did not move out of the greater Houston-Galveston, Texas, metropolitan area, and (3) the mother left the shelter to establish a residence with her children that was independent of her battering partner.^ When more than one child in the family met criteria for ODD or CD, the youngest child in the 4-10 year age range was selected as the target child. The target children were 37 boys and 31 girls, with a mean age of 5.6 (SD = 1.67). Our longitudinal subsample did not differ from the larger sample on the demographic variables reported earlier.

*Less than 5% of the families who sought refuge at one of the participating shelters were excluded from this research because the mother could not communicate sufficiently in English. ^ We included only women who did not return to their battering partners because many sheltered battered women have informed us that their partners would not be receptive to our visiting them in their homes.

296


Procedures

Data for this study were collected as part of two other studies designed to evaluate an intervention program for conduct problems displayed by children of battered women. Families were recruited from seven shelters serving battered women and their children. The participating shelters served women and children from urban, suburban, and rural areas in the HoustonGalveston, Texas, area. To insure the recruitment of a representative sample of shelter residents, a research staff member contacted the shelters several times a week to assess current residency at each of the shelters. When notified that a woman with a child between the ages of 4 and 10 had entered a participating shelter, a research staff member contacted the woman (either in person or by phone) as soon as feasible to introduce herself and describe the research. Women were asked to participate in a comprehensive assessment as part of ongoing attempts to assess the needs of women and children in shelters as well as to determine their eligibility for intervention studies. Assessment appointments were scheduled at mothers' convenience as soon after shelter entry as possible. Most women preferred to schedule appointments after having a few days to "settle in" at the shelter. The mean number of days between shelter entry and completion of the interview was 10.1 {SD = 14.4). Trained staff interviewed mothers concerning their experience of abuse, their children's behavior, and their own levels of distress. The interviews were conducted privately, and children were supervised in play in another room of the shelter while their mothers were being interviewed. Mothers were given the option of completing the instruments on their own or having the items of these instruments read aloud to them. When our assessments revealed the existence of problems warranting clinical attention, research staff offered families referrals for appropriate services. Ninety-seven percent of the women who met our eligibility criteria and who remained in shelter for at least 3 days were asked to participate in the present research; 89% of these women participated. Participation rates at individual shelters ranged from 80 to 95%. Of the 11% who did not participate, 6% exited the shelter before data collection could be arranged, and 5% declined participation. In short, we believe that our sample is representative of English-speaking families with a child between 4 and 10 years of age who seek refuge at a shelter because of battering and stay at the shelter for at least a few days. Mothers in the longitudinal subsample participated in a second assessment of their children's behavior and their own psychological distress after leaving shelters. The average period of time between shelter and postshelter interviews was 56.1 days {SD = 33.5). Postshelter assessments were conducted in families' homes by the same research staff who conducted the


297

in-shelter assessment. These interviews were conducted as soon as convenient for the famihes after shelter departure (mean of 27.9 days, SD = 20.5). No formal intervention was initiated between shelter exit and the postshelter assessment; however, during shelter residence, many of the mothers received assistance from shelter staff (typically emotional support and referrals for community services). While mothers were interviewed in their homes, undergraduate research assistants supervised and played with their children. Mothers were compensated for participating in the postshelter assessment at a rate of approximately $10 per hour. Eighty-four percent of the families who were eligible for the longitudinal study completed the postshelter assessment. Those who were eligible but did not participate either declined to participate (n — 6), or we were unable to locate them after their shelter departure (n = 7). Some of the families in this latter group of seven may have been ineligible for participation (e.g., because they moved out of the area), but we were unable to confirm their eligibility status. Measures

Conflict Tactics Scale (CTS; Straus, 1979) The physical violence subscale of the CTS was completed by mothers during shelter residence for purposes of sample selection and sample description. The CTS includes eight items describing physically violent acts, and mothers indicated how often they were the victims of each of these violent acts at the hands of their intimate partner during the previous 12 months. The physical violence subscale of the CTS is commonly used for sample selection in domestic violence research, and the incidence and frequency of physical spousal violence, as indexed by this subscale, is associated with many variables that are theoretically linked to domestic violence (Straus & Gelles, 1990b). Symptom ChecJclist—90—Revised (SCL-90-R; Derogatis, 1977) The General Severity Index of the SCL-90-R was used as an index of mothers' psychological distress, and it was administered to mothers during both the in-shelter and postshelter assessments. This measure assesses how much respondents were bothered by 90 different symptoms of psychological distress (including fear, anxiety, depression, and somatic complaints) in the previous week. There is substantial evidence for the validity of this

298


instrument for psychiatric and nonpsychiatric populations (Derogatis et al, 1976). The coefficient alpha for the General Severity Index in our shelter sample was .97.

Child Behavior Checklist (CBCL; Achenbach, 1991) Mothers' perceptions of their children's externalizing (EXT) and internalizing (INT) behavior problems were measured using the CBCL. Mothers completed this measure during the in-shelter and postshelter assessments. Mothers were asked to report on their children's behavior over the past 6 months. The CBCL is a widely used and well-validated measure of child behavior problems. As recommended by Achenbach (1991), a cutoff score of 63 was used to denote clinical levels of child externalizing problems. Of particular relevance to the present research, the 1-week test-retest rehability coefficients for the externalizing and internalizing scales are .93 and .89, respectively (Achenbach, 1991). Also relevant to the present research is the practice effect discussed in the CBCL manual (Achenbach, 1991). Scores tend to decline over brief test-retest intervals, although the dechne observed has been small (10.7% mean decrease in problem scores on average). Diagnostic Interview (McDonald & Stephens, 1996) A structured diagnostic interview was administered to assess whether children met DSM-IV (APA, 1994) criteria for ODD and CD at the time of the interview. The interview was designed to obtain information about the frequency of occurrence of specific child behaviors outlined in the DSM-IV as characteristic of ODD and CD. Sample questions include the following: "Does your child just seem to feel angry or resentful and irritated a lot of the time?" "Does your child often just refuse to do things that you, the teachers, or other adults have asked him/her to do?" Mothers were asked to describe incidents in which each reported behavior occurred and to describe her child's behaviors during those incidents; responses were probed until sufficient information was obtained for the interviewer to determine the presence or absence and frequency of occurrence of each criterion behavior. In accord with DSA/-/V criteria, the interviewer assessed the time period during which the behaviors had been occurring and whether the behaviors resulted in problems for the child across multiple domains or settings. The interview was administered to 381 mothers during the in-shelter assessment, including 46 of the 68 mothers who participated in the postshelter assessments. Data from the shelter diagnostic interview were used to examine the


299

prevalence of clinical levels of child conduct problems and to select children for the longitudinal portion of this research. A subset of interviews {n = 203) were audiotaped and coded by a second rater. Kappas for interrater agreement on diagnosis were .95 for ODD, .99 for CD, and .96 for the presence of either ODD or CD.

RESULTS Prevalence of Clinical Levels of Conduct Problems and Their Association With Mothers' Distress During Shelter Residence

For descriptive purposes, we examined the relation between demographic variables and child problems. Child age was not associated with diagnosis or CBCL externalizing T scores. There were no significant gender differences in CBCL externalizing T scores; however, proportionally more boys than girls met DSM-IV criteria for clinical levels of conduct problems (23%boys,14%girls;x^(l,n = 381) = 4.3,p < .05), which is consistent with national prevalence data indicating that childhood CD and ODD are more common in males (APA, 1994). The mean CBCL-EXT T score was 56.8 {SD = 12.02). Thirty percent (95% CI, 25.4-34.6%) of children had CBCL-EXT Tscores above the cUnical cutoff of 63. Thirty-seven percent (95% CI, 32.2-41.8%) of children met DSM diagnostic criteria for either ODD or CD; 15 % met criteria for CD, and 22% met criteria for ODD. The McNemar Chi-square test for correlated proportions indicated that the prevalence estimates derived from the CBCL and diagnostic interviews differed significantly, x^(l. '^ = 381) = 9.12, p < .01. As can be seen in Table I, there was agreement between the diagnostic data and the CBCL on the presence of clinical levels of problems in 77% of the cases. The kappa coefficient was .50. The probability of mothers reporting clinical levels of externalizing problems given that their child met diagnostic Table I. Clinical Levels on CBCL-EXT by Diagnosis of ODD or CD > 63 on CBCL EXT Diagnosis Yes No Total

Yes

No

Total

85 (22.3) 29 (7.6) 114(29.9)

57 (14.9) 210 (55.1) 267(70.1)

142 (37.3) 239 (62.7) 381" (100)

Note. CBCL-EXT = Child Behavior Checklist - Externalizing Subscale 7"score; ODD = oppositional defiant disorder; CD = conduct disorder. Values given in parentheses are percentages. "Diagnostic data were not available for 20 participants.

300


criteria was .60. The probability of the child meeting DSM diagnostic criteria given that their mothers reported clinical levels of externalizing problems on the CBCL was .75. We also examined the relation between mothers' distress and child externalizing problems. The mean SCL-90-R Tscore for mothers was 65.7 {SD = 9.70). The CBCL-EXT and SCL-90-R were positively correlated, r = .39, p < .05. In addition, mothers of children who met diagnostic criteria for ODD or CD had significantly higher scores on the SCL-90-R (M = 68.8, SD - 9.04) than did mothers of children who did not meet the diagnostic criteria (M = 64.1, SD = 9.04), f(399) = 4.67, p < .05. Longitudinal Results

Mean scores for the CBCL-EXT and SCL-90-R at Time 1 (in-shelter) and Time 2 (postshelter) for the longitudinal subsample of families (n = 68) appear in Table II. As expected, mothers' SCL-90-R scores decreased significantly from Time 1 to Time 2, F{1, 67) = 20.89, p < .05. A repeatedmeasures ANOVA indicated no significant difference between Time 1 and Time 2 CBCL externalizing scores . Another repeated-measures ANOVA revealed no significant interaction of time and diagnostic category, indicating that CBCL scores were stable for both CD and ODD diagnoses. Percentages calculated for descriptive purposes indicated that 45% of mothers reported no or very little change on the CBCL-EXT (0-4-point increase or decrease, i.e., less than one half of one standard deviation), and 30% reported their children's externalizing problems to be substantially worse (5-point increase or greater) after leaving shelters. The probability of a mother reporting clinical levels of externalizing problems at Time 2 given that she reported clinical levels of such problems at Time 1 was .82. All of the families participating in the postshelter assessments originally met DSM-IV diagnostic criteria for either ODD or CD during shelter residence; 93% (43/46) did so at the postshelter assessment as well. Table II. Means of Measures and Results of Repeated Measures ANOVA for Longitudinal Sample At shelter (Time 1) Measure CBCL-EXT SCL-90-R

Postshelter (Time 2)

M

SD

M

SD

F(l, 67)

P

64.2 66.1

9.40 11.08

66.2 60.2

8.64 11.65

3.93 20.89

.052 .0001

Note, n = 68. CBCL-EXT = Child Behavior Checklist, Externalizing Subscale T score; SCL90-R = Symptom Checklist-90-Revised, General Severity Index T score. Higher scores on measures indicate more problems.


301

Table III. Correlations Between Maternal Distress and Children's Externalizing Problems During Shelter and After Shelter Exit Timel CBCL-EXT Timel SCL-90-R Time 2 CBCL-EXT SCL-90-R

SCL-90-R

Time 2 CBCL-EXT

.43' .56' .21

.36' .57'

.26'

Note, n = 68. CBCL-EXT = Child Behavior Checklist-Externalizing Subscale T score; SCL90-R = Symptom Checklist-90-Revised, General Severity Index 7 score, "p < .05.

Correlations between the CBCL and the SCL-90-R at Time t and Time 2 are presented in Table III. Even among the Time 1 subgroup of children selected on the basis of the diagnostic interview, maternal distress was associated with concurrent reports of children's externalizing problems during shelter residence. There was also a significant concurrent relation between maternal distress and child externalizing problems after shelter exit. Using Fisher's r to z transformation, a comparison of Time 1 and Time 2 correlations between maternal distress and child externalizing problems indicated that the magnitude of the relation did not differ significantly across the two time points. Although mothers' distress declined after shelter departure, their reports of child conduct problems remained stable. To rule out the plausible explanation that mothers' reports of child problems after shelter departure were attributable, in part, to some residual effects of mothers' level of distress while in the shelter, we conducted a path analysis examining the relation between mothers' distress during shelter residence and their reports of child externalizing problems after shelter departure. The path analysis provided information that differed from the simple bivariate correlations because it accounted for other relations in the model (depicted in Fig. t). In this model, mothers' reports of child problems and mothers' distress were allowed to correlate at Time t, and the disturbance terms of these variables were allowed to correlate at Time 2. The nonsignificant coefficient for the path between SCL-90-R at Time t and CBCL-EXT at Time 2 (also shown in Fig. t) indicated that mothers' distress while in shelters did not relate to their reports of child problems after shelter exit. Exploratory Analyses With Internalizing Problems

Although the principal hypotheses of this study involved child externalizing problems and children in the longitudinal study were selected on the

302


CBCL- EXT

SCL-90-R

.46* (.60)

.62* (.59)

CBCL- EXT

SCL-90-R

Fig. 1. Path analysis with externalizing problems. Covariance between exogenous variables (standardized covariance coefficient in parentheses) is indicated by superscript a; Covariance between error terms/zeta (standardized covariance coefficient in parentheses) is indicated by superscript b; *p