Maltreated children in out-of-home care: The relation between attachment quality and internalizing symptoms

Abstract

Maltreated children in out-of-home care are at high risk for poor relationships with caregivers (i.e., biological parents and substitute caregivers) and high levels of internalizing symptoms. It is unclear if these poor relationships are related to, and account for a large portion of the variance in maltreated children’s internalizing symptoms, above and beyond maltreatment type and out-of-home care factors. This study examined the relation between attachment quality with both biological parents and substitute caregivers and children’s internalizing symptoms within a sample of 493 maltreated children (aged 9-11; 51.0% male) recently placed in out-of-home care. A series of hierarchical regression models indicated that greater child-reported attachment quality with both biological parents and substitute caregivers was associated with fewer child-reported anxiety (β = −.15, p < .01; β = −.29, p < .001, respectively) and depression symptoms (β = −.14, p < .01; β = −.28, p < .001, respectively) as well as fewer child internalizing symptoms (β = −.12, p < .05; β = −.14, p < .01, respectively). Attachment quality with the biological parent and substitute caregiver each explained a significant proportion of the variance in children’s internalizing symptoms, above and beyond child demographics, maltreatment type, and out-of-home care variables. The study also examined whether children’s attachment with substitute caregivers moderated the relationship between children’s attachment with biological parents and children’s internalizing symptoms. No statistically significant moderation effects were found. Future clinical work should focus on enhancing attachment quality between children and both biological parents and substitute caregivers, as these relationships appear to individually relate to the children’s internalizing symptomology.

Introduction

Exposure to maltreatment is a significant risk factor for the development of internalizing and externalizing symptomology among children (Cicchetti & Valentino, 2006; Manly, Kim, Rogosch, & Cicchetti, 2001; Toth, Manly, & Cicchetti, 1992). Children who experience maltreatment severe enough to warrant their removal from their homes may be at an even greater risk for mental health and behavioral problems due to their cumulative exposure to additional adverse events (e.g., caregiver disruptions, exposure to community violence; Raviv, Taussig, Culhane, & Garrido, 2010). Although internalizing and externalizing symptoms often co-occur (Fanti & Henrich, 2010), externalizing behaviors arguably receive the most clinical attention due to their outward expression. Internalizing symptomatology in childhood, however, is equally concerning as it is associated with a host of mental health problems across the lifespan (Kovacs & Devlin, 1998; Moffit et al., 2007), and may place maltreated children at risk for poor adult outcomes such as substance use disorders and homelessness (Lo & Cheng, 2007; Roos et al., 2013). Yet, not all children who experience abuse and/or neglect and enter out-of-home care develop significant internalizing disturbances. In order to design programs that prevent and/or attenuate these poor mental health outcomes among maltreated children in out-of-home care, it is imperative to identify factors associated with variations in children’s internalizing symptoms.

Attachment, which refers to the close, emotional bond that develops between a child and his or her primary caregiver (Bowlby, 1973) may be one factor related to maltreated children’s internalizing symptomatology. Attachment theory states that children construct “internal working models” in which children internalize their interactions and the responsiveness they receive from their caregivers, leading them to develop a self-image through the eyes of their primary caregiver (Bowlby, 1969, 1973). In a secure attachment, a child perceives his or her primary caregiver as sensitive, warm, and responsive, which builds trust and security within the relationship (Ainsworth, 1979; Ainsworth, Blehar, Waters, & Wall, 1978). Not all parent-child relationships, however, manifest in secure attachments. Rather, some relationships may be disrupted or severed resulting in an insecure attachment. Maltreated children in particular are at increased risk for insecure attachments with studies showing up to 75 percent of maltreated children exhibiting insecure relationships with biological parents (Baer & Martinez, 2006; Cicchetti & Barnett, 1991; Morton & Browne, 1998). Through the “internal working model,” a child who develops an insecure attachment due to inconsistent and/or inadequate caregiving may begin to perceive him or herself as a failure, potentially fueling internalizing symptomatology.

Because maltreated children in out-of-home care have several caregiving relationships, one with biological parent(s) and one with substitute caregiver(s) (e.g., non-relative foster parent, kin), it is important to attend to how both relationships may be associated with children’s internalizing symptoms. Several studies have found associations between child-reported “relatedness” to biological mothers (i.e., a closely related concept to attachment) and children’s internalizing symptoms among school-aged, maltreated children (ranging from 7 to 14 years-old). The relatedness scale measures children’s perceptions of closeness to the caregiver and parental support, rather than the type of attachment formed (i.e., insecure or secure). Children’s ratings on the relatedness scale form patterns that are consistent with attachment theory, such that poor relatedness is similar to an insecure attachment and positive relatedness is similar to a secure attachment (Toth & Cicchetti, 1996). In a cross-sectional study, Toth and Cicchetti (1996) found that maltreated children who reported positive relatedness to their mothers evidenced fewer depressive symptoms than did their maltreated counterparts with poor relatedness. Additionally, Alink et al. (2009) found that maltreated children who reported positive relatedness with their mothers were concurrently at lower risk for emotion dysregulation and internalizing symptoms. Lastly, in a longitudinal study, Kim and Cicchetti (2004) found that children’s positive relatedness with their mothers was directly related to fewer internalizing symptoms one year later, as well as indirectly related through children’s self-esteem while controlling for baseline internalizing symptoms. Combined, these findings suggest that maltreated children’s positive relatedness with their biological mothers may serve as a protective factor for children’s internalizing symptomology. However, studies have not examined this association among children recently placed in out-of-home care.

To our knowledge, only two studies have examined maltreated children’s relationships with substitute caregivers in relation to child adjustment outcomes. Although the study’s findings should be interpreted with caution, as statistical power was limited due to a small sample size (N = 32), Milan and Pinderhughes’ (2000) longitudinal study found that children who self-reported positive relatedness with their foster mothers tended to exhibit fewer internalizing symptoms relative to children who self-reported poor relatedness to their foster mothers one month later (Milan & Pinderhughes, 2000). The second study did not examine internalizing symptoms but did find an association between the quality of maltreated children’s attachment with foster mothers (assessed using the Child Attachment Interview) and parent-reported conduct behaviors (Joseph, O’Connor, Briskman, Maughan, & Scott, 2014). While the aforementioned studies lack an experimental design, the pattern of findings suggests that maltreated children’s perceptions of the quality of their relationship with substitute caregivers may be associated with child adjustment including internalizing symptoms.

Because quality attachments are known to be protective against the development of poor child adjustment outcomes (Brennan, Le Brocque, & Hammen, 2003; El-Sheikh & Buckhalt, 2003), quality attachment relationships with substitute caregivers may moderate the negative effects of poor attachment histories with biological parents. Although relationships with substitute caregivers have not been previously explored as potential moderators, studies have shown that social support from other family members and peers buffers the negative impact of maltreatment on adult’s trauma and mental health symptoms (Evans, Steel, & DiLillo, 2013; Sperry & Widom, 2013). Similarly, parenting histories that include maltreatment may have larger or weaker effects on children’s internalizing symptoms depending on whether or not children develop positive attachment experiences with their substitute caregivers

Although maltreated children in out-of-home care are more likely than non-maltreated children to experience poor attachment quality with their biological parents and potentially with substitute caregivers, the presence of a poor relationship alone may not be the only factor contributing to an increased risk of internalizing symptoms. Children in out-of-home care who experience abuse and/or neglect also experience numerous risk factors related to their placement experiences. First, maltreated children who are removed from their biological parents often face frequent placement disruptions, which may be related to poor child adjustment and poor attachment to subsequent substitute caregivers. For example, a prospective study of school-aged children and adolescents found that the number of placement changes predicted internalizing symptoms at 17 months, after accounting for these same emotional symptoms at 5-months post-initial placement (Newton, Litrownik, & Landsverk, 2000). Second, placement type is an important consideration, as there is some evidence that kinship caregivers (e.g., relatives, friends of family) tend to report fewer internalizing problems among children in their care in comparison to children in other types of placements (i.e. non-relative foster care; Hegar & Rosenthal, 2009). However, other studies have found that placement type is not significantly associated with child outcomes when other out-of-home care factors are considered (Fechter-Leggett & O’Brien, 2010). Third, frequent contact with biological mothers may be related to lower levels of emotional problems among maltreated children and adolescents in out-of-home care (McWey, Acock, & Porter, 2010; McWey & Mullis, 2004). In contrast, Leathers (2003) proposes that frequent visitation with biological parent(s) may result in loyalty conflicts (i.e., conflict of feelings toward foster family versus biological family), leading to symptoms of anxiety and depression (Leathers, 2003). Overall, these findings indicate that contextual factors related to out-of-home care may influence children’s internalizing symptoms. Unfortunately, few studies have investigated the relationship between children’s attachment quality with caregivers and child mental health, while also accounting for multiple placement factors (e.g., length of current placement, frequency of maternal contact). Accounting for these factors may help researchers and clinicians tease out the association between children’s attachment quality with caregivers and children’s internalizing symptoms.

The current study examines the following two research questions: (1) What is the relationship between child-reported attachment quality (with the biological parent and substitute caregiver) and child internalizing symptoms among a sample of maltreated children recently placed in out-of-home care, controlling for multiple placement factors? (2) Does attachment to substitute caregivers moderate the relationship between children’s attachment to biological parents and children’s internalizing symptoms? We hypothesize that children’s attachment to biological parents and substitute caregivers will each be associated with children’s internalizing symptoms, such that children who have greater attachment quality with their biological parents and/or substitute caregivers will exhibit fewer internalizing symptoms. Further, we hypothesize that children’s attachment to substitute caregivers will act as a moderator, such that a quality relationship with the substitute caregiver will buffer the negative association between children’s attachment quality with biological parents and children’s internalizing symptoms. If findings do indicate that children’s attachment quality with both caregivers is related to children’s internalizing symptoms, it may support practices that are aimed at improving relationships and interactions between children and their caregivers.

Method

Participants

The sample included children, aged 9–11, and their substitute caregivers living in a large, urban Western city. Children and their caregivers were recruited for a randomized controlled trial of an intervention for preadolescents in out-of-home care called Fostering Healthy Futures (Taussig, Culhane, & Hettleman, 2007). Children were eligible for participation in the study if the following criteria were met: (1) they were placed in out-of-home care due to maltreatment by court order within the preceding year; (2) they still resided in out-of-home care at the time of the baseline interview; (3) they were within 35-minutes driving distance from the program location; and (4) cognitive functioning was sufficient to comprehend the interview questions. Ninety-one percent of children meeting eligibility requirements were enrolled in the study. Because we are interested in examining children’s relationships with their substitute caregivers, children from congregate care (n = 23; 4.5% of the total sample) were not included in the analysis. Thus, baseline data were analyzed for a total of 493 children and their current caregivers.

Of the 493 children in the study sample, 51.1% were male with a mean age of 10.4 years (SD = .90). Approximately half of the children reported their ethnicity as Hispanic (47.5%), 47.1% as White, 26.8% as African American, 13.4% as Native American, 2.6% as Asian American, 1.6% as Pacific Islander, and 6.3% as other ethnicity. Racial and ethnic identity is reported as non-exclusive categories, such that youth may identify as more than one racial/ethnic background (e.g., as both White and African American).

Procedure

University Institutional Review Board approval and informed consent/assent were obtained prior to children and their current caregivers (i.e., substitute caregivers) being interviewed. Children and their caregivers were interviewed separately, either at their residence or at another other community location, by trained graduate student research assistants. Both children and their current caregivers received $40.00 for their participation in the study.

Measures

Demographic and maltreatment variables.

Child age and sex (1 = male; 0 = female) were determined from child welfare records. Race and/or ethnicity was self-reported and dichotomously coded (1 = minority; 0 = non-minority). Participants who identified as White and another race and/or ethnicity were assigned a value of 1. Child Protection Services’ intake reports and Dependency and Neglect petitions (i.e., legal petitions filed on the family in order to remove child from their custody) were used to code maltreatment types. Maltreatment types were coded using the Maltreatment Classification System (Barnett, Manly, & Cicchetti, 1993) by at least two research assistants who resolved discrepancies through consultation with one of the senior investigators. Types of maltreatment were coded as present or absent for each child and included: physical abuse, sexual abuse, physical neglect, supervisory neglect, and emotional maltreatment. In addition, all positively endorsed maltreatment subtype variables were summed together to create a continuous, cumulative variable of number of types of maltreatment.

Placement experience variables.

Variables that measured placement experiences included: (1) placement type; (2) number of caregiver transitions from birth to baseline assessment; (3) frequency of contact with biological mother; (4) length of time in out-of-home care; and (5) length of time in current placement. Type of placement was examined utilizing child protective service reports and was dichotomously coded as 1 = foster care and 0 = kinship care. Number of caregiver transitions was examined by having interviewers assist children in creating a chronological log of all the caregivers with whom the child had lived with since birth. When children lived with the same caregiver for noncontiguous periods of time, each occasion was included in the summed number of caregiver transitions. Frequency of contact with the biological parent (i.e. mother) was assessed by asking children, “How often do you see or talk to your [biological mom]?” Responses ranged from 1 (more than 2 times per week) to 8 (never). Response categories were recoded into a 4-point scale, such that 0 = never, 1 = a little, 2 = sometimes, and 3 = often. Child welfare records were used to determine length (in months) in out-of-home care. Finally, length of current placement was examined by subtracting the date of the interview from the date the current placement began (as obtained from child welfare records). Length was coded in terms of months.

Independent variables.

Attachment with children’s biological parents and substitute caregivers was assessed using the Inventory of Parent and Peer Attachment – Short Form (IPPA-Short Form; Gifford-Smith, 2000). The IPPA was derived from an adolescent version of the measure (Armsden & Greenberg, 1987) and was modified for use with fourth and fifth graders by the FAST Track Project (Gifford-Smith, 2000). The scale contains 15 items. Sample IPPA items include questions asking children to rate whether their caregivers, “help me with my problems” and “care about me.” Scores range from 1 (not true) to 3 (often true), with higher scores indicating greater perceived attachment. In the present study, attachment was assessed between the child and his or her (a) biological parent and (b) substitute caregiver. Cronbach’s alphas for attachment quality with substitute caregiver and attachment quality with biological parent for the current sample were .83 and .92, respectively.

Dependent variables.

Substitute caregivers completed the Child Behavior Checklist (CBCL; Achenbach & Rescorla, 2001) and children completed the Trauma Symptom Checklist for Children (TSCC; Briere, 1996), as well as the Revised Children’s Manifest Anxiety Scale (RCMAS; Reynolds & Richmond, 2000). The CBCL is a widely used standardized measure of child behavior problems with acceptable levels of internal consistency (Cronbach’s alphas ranging from .78 to .97; Achenbach & Rescorla, 2001). The 32-item Internalizing Problems subscale was used in analyses. Substitute caregivers were asked to indicate how true each item was for their child “now or within the past 6 months,” with response options ranging from 0 (not true) to 2 (very true or often true). Higher t-scores on the Internalizing Problems subscale indicated greater internalizing problems. The TSCC is a 54-item measure of trauma-based symptoms with acceptable levels of reliability (Cronbach’s alphas ranging from .67 to .89; Briere, 1996). The TSCC is comprised of six clinical subscales, which include dissociation, posttraumatic stress, anxiety, depression, anger, and sexual concerns. In the current study, we examined the depression subscale only. For each item, children were asked to indicate “how often each thing happens to you,” with responses scored on a 4-point scale ranging from 0 (never) to 3 (almost all of the time). Lastly, the Revised Children’s Manifest Anxiety Scale, a 37-item measure of anxiety was used to yield a total anxiety score, with higher scores indicating greater anxiety symptoms. For each item, children were asked to indicate, “whether you ever have thoughts and/or feeling like these,” with responses scored as 1 (yes) or 0 (no).

Data Analytic Plan

Correlations were used to estimate the bivariate relationships between parent-child attachment, child demographics, maltreatment and placement experience variables, and the three dependent variables of child internalizing symptoms. Next, a series of hierarchical regression models was specified to assess the relationship between parent-child attachment quality and child internalizing symptoms (i.e., youth-reported depression and anxiety, and caregiver-report of child internalizing symptoms). For each model assessing child internalizing symptoms, the demographic variables of sex and age were entered into step one. Minority status was not included because it was not significantly correlated with any of the outcome variables or parent-child attachment quality variables. At step two, the maltreatment variables and placement experience variables were entered. We chose to include variables that were marginally significant in addition to variables that were significantly related to any of the three outcomes or the parent-child attachment quality variables at baseline. Although not all of these variables were significantly correlated with the dependent variables, prior research suggests that many of these variables are significant predictors of child internalizing behaviors (see James, Landsverk, Slymen, & Leslie, 2004; Tarren-Sweeney, 2008). At step three, the parent-child attachment variables were entered. Lastly, at step four, the interaction variable, child attachment to biological parents x child attachment to substitute caregivers, was entered. The interaction variable was created by multiplying the child attachment to biological parent’s variable by the child attachment to substitute caregiver variable. Prior to multiplying these variables, the continuous variables were centered at the mean. Interaction effects were examined by probing at one standard deviation above and below the mean. Missing data was handled using listwise deletion. To ensure there were not issues of multicollinearity in the models, we examined variance inflation factor (VIF) values for each predictor. All VIF values were below the recommended threshold value of five (Belsley, Kuh, & Welsch, 1980).

Results

Table 1 presents the descriptive statistics for the children’s internalizing symptoms, attachment quality, maltreatment, and placement experience variables. The means for the depression and anxiety subscales were 46.5 (SD = 9.9) and 51.4 (SD = 10.6), respectively. The mean for the total sample on the CBCL internalizing subscale was 59.4 (SD = 11.6). On average, youth reported relatively high mean scores on the attachment quality scale with both biological (M = 38.6; SD = 6.8; R = 15.0 – 45.0) and substitute caregivers (M = 38.6; SD = 5.6; R = 16.0 – 45.0). The majority of children (84.6%) experienced supervisory neglect, 62.3% experienced emotional maltreatment, 48.9% experienced physical neglect, 26.0% experienced physical abuse, 11.4% experienced sexual abuse, and 78.5% experienced multiple types of maltreatment. With regard to placement experiences, 54.8% of children were placed in kinship care and 45.2% were placed with a nonrelative foster caregiver. On average, children had 2.7 (SD = 2.0) caregiver transitions since birth and reported having some contact with their biological mothers (M = 2.2; SD = 1.2). Children, on average, spent 7.1 (SD = 3.4) months in out-of-home care and 6.1 (SD = 3.8) months in their current placement.

Table 1.

Descriptive Statistics for Study Variables

	M or %	SD	Range
Child internalizing symptoms
Anxiety	51.4	10.6	25.0 – 79.0
Depression	46.5	9.9	32.0 – 84.0
Caregiver-reported child internalizing	59.4	11.6	33.0 – 86.0
Attachment Quality
Substitute caregiver	38.6	5.0	16.0 – 45.0
Biological parent	38.6	6.8	15.0 – 45.0
Maltreatment type a
Physical abuse, %	26.0	–	–
Sexual abuse, %	11.4	–	–
Physical neglect, %	48.9	–	–
Supervisory neglect, %	84.6	–	–
Emotional abuse, %	62.3	–	–
Multiple maltreatment	2.3	1.0	0.00 – 5.0
Placement experience variables
Nonrelative foster care, %	45.2	–	–
Kinship care, %	54.8	–	–
Caregiver transitions	2.7	2.0	1.0 – 11.0
Frequency of contact	2.2	1.2	0.0 – 3.0
Length in out-of-home care	7.1	3.4	1.0 – 15.0
Length of current placement	6.1	3.8	1.0 – 24.0

^aThese categories were nonexclusively coded; that is, a child could be coded in more than one category.

Table 2 presents the bivariate correlations between study variables. Children’s greater attachment quality with both the substitute caregiver and biological parent were related to child-report of fewer anxiety and depression symptoms, as well as to caregiver-report of fewer internalizing symptoms. Of the child demographic variables, only age and gender were related to children’s internalizing symptoms. Older children tended to report fewer anxiety and depression symptoms than younger children and females tended to report greater anxiety symptoms relative to males. Of the maltreatment types, sexual abuse was positively correlated with children’s reports of depressive symptoms and supervisory neglect was positively related to caregiver-report of child internalizing symptoms. Of the placement experience variables, placement in foster care (versus kinship care) was positively correlated with caregiver-report of internalizing symptoms and lower attachment quality with substitute caregivers. A greater number of caregiver transitions since the child’s birth was also correlated with lower attachment quality with substitute caregivers. Last, less contact with the biological mother, greater length of time in out-of-home care, and longer time in current placement were related to lower attachment quality with the biological parent.

Table 2.

Bivariate Correlations between Independent Variables and Child Internalizing Symptoms

	Internalizing Symptoms			Attachment Quality
	Anxiety	Depression	Internalizing	Substitute Caregiver	Biological Parent
Attachment quality (AQ)
AQ substitute caregiver	−.311***	−.326***	−.157**
AQ biological parent	−.164***	−.164***	−.118*	.095*
Demographics
Age	−.141**	−.145**	.010	.035	−.081†
Sex (1 = male)	−.124**	−.075	.032	−.050	.016
Minority status	−.013	.046	.029	−.060	−.011
Maltreatment type a
Physical abuse	.065	.080†	−.006	−.041	−.002
Sexual abuse	.052	.101*	−.011	−.073	−.001
Physical neglect	.024	−.010	.073	.018	.032
Supervisory neglect	−.005	.013	−.100*	.029	−.018
Emotional abuse	.052	.033	.013	−.008	−.089†
Multiple maltreatment	.082†	.085†	.001	−.025	−.035
Placement experience variables
Placement type (1 = foster care)	.079	.095†	.189***	−.115*	.073
Caregiver transitions	.074	.085†	.088†	−.097*	−.027
Frequency of contact	−.004	.019	−.016	−.063	.200***
Length in out-of-home care	.007	.055	−.042	−.040	−.159***
Length of current placement	.009	.022	−.024	.033	−.105*

^aThese categories were nonexclusively coded; that is, a child could be coded in more than one category.

^†p<.10

^*p<.05

^**p<.01

^***p<.001

Table 3 presents hierarchical multiple regression models predicting child and caregiver reports of child internalizing symptoms. In the first step of the hierarchical regression models predicting youth-reported anxiety, two variables were entered: sex and age. This model was statistically significant F(2, 411) = 5.46; p < .01 and explained 2.1% of the variance in anxiety. After entry of maltreatment and placement experience variables at step two, the model became marginally significant F(12, 401) = 1.75; p < .10 and still explained 2.1% of the variance in anxiety. Adding attachment variables in step three increased the amount of the variance explained by 11.1%, explaining a total of 13.2% of the variance in anxiety F(14, 399) = 5.49; p < .001. Finally, in step four, with the addition of the interaction variable (i.e., children’s attachment quality with the substitute caregiver X children’s attachment quality with the biological parent), the model was statistically significant F(15, 398) = 5.18; p < .001 and still explained 13.2% of the total variance in anxiety.

Table 3.

Hierarchical Regression Models Predicting Child Internalizing Symptoms

	Anxiety				Depression				Internalizing symptoms
	β	95% CI	R2	R2Δ	β	95% CI	R2	R2Δ	β	95% CI	R2	R2Δ
Step 1:			.021				.017				−.004
Age	−.15**	[2.84, −0.69]			−.14**	[−2.55, −0.56]			.00	[−.1.20, 1.28]
Sex	−.09†	[−3.89, 0.03]			−.07	[−3.14, 0.50]			.02	[−.1.83, 2.75]
Step 2:			.021	.024			.034	.040†			.027	.056*
Sexual abuse	−.01	[−3.98, 3.23]			.07	[−.1.16, 0.66]			−.06	[−6.49, 2.30]
Physical abuse	.08	[−.1.11, 4.76]			.09	[−0.80, 4.65]			−.10	[−6.18, 0.81]
Supervisory neglect	.01	[−3.26, 3.83]			.04	[−2.42, 4.16]			−.10	[−7.73, 0.81]
Emotional abuse	−.01	[−2.92, 2.57]			−.02	[−2.99, 2.11]			−.05	[−4.30, 2.08]
Multiple maltreatment	.00	[−.1.89, 1.97]			−.02	[−.1.92, 1.68]			.13	[−0.70, 3.81]
Placement type	.06	[−0.89, 3.19]			.06	[−0.63, 3.17]			.18**	[1.82, 6.64]
Caregiver transitions	.07	[−0.13, 0.86]			.09†	[−0.05, 0.87]			.05	[−0.28, 0.87]
Frequency of contact	−.01	[−0.94, 0.74]			.01	[−0.68, 0.88]			.03	[−0.70, 1.25]
Length in out-of-home care	−.09	[−0.62, 0.08]			.00	[−0.32, 0.33]			−.09	[−0.69, 0.11]
Length of current placement	.07	[−0.12, 0.52]			.06	[−0.16, 0.43]			.07	[−0.14, 0.56]
Step 3:			.132	.112***			.136	.103***			.057	.034**
AQ with substitute caregiver	−.29***	[−0.80, −0.41]			−.28***	[−0.73, −0.37]			−.14**	[−0.53, −0.09]
AQ with biological parent	−.15**	[−0.38, −0.09]			−.14**	[−0.35, −0.07]			−.12*	[−0.36, −0.03]
Step 4:			.132	.002			.140	.006†			.055	.001
AQ with substitute caregiver X AQ with biological parent	.05	[−0.01, 0.04]			.08†	[−0.004, 0.04]			−.03	[−0.04, 0.02]

Note. AQ = Attachment quality. Standardized betas and adjusted R² are reported.

^†p<.10

^*p<.05

^**p<.01

^***p<.001

The first step of hierarchical regression models predicting youth-reported depression included sex and age. This model was statistically significant F(2, 413) = 4.66; p < .05 and accounted for 1.7% of the variance in depression. After the entry of maltreatment and placement experience variables at step two, the model remained statistically significant F(12, 403) = 2.23; p < .05 and explained 3.4% of the variance in depression. Adding the attachment quality variables in step three increased the variance by 10.2%, explaining a total of 13.6% of the total variance in depression F(14, 401) = 5.66; p < .001. Finally, in step four, with the addition of the interaction variable, the model remained statistically significant F(15, 400) = 5.49; p < .001 and explained 14.0% of the variance in depression.

The first step of the hierarchical regression models for substitute caregiver-report of child internalizing symptoms included sex and age. This model was not statistically significant F(2, 391) = .215; p = ns and explained 0.04% of the variance in child internalizing symptoms. After the entry of maltreatment and placement experience variables at step two, the model became significant F(12, 381) = 1.912; p < .05 and explained 2.7% of the variance in child internalizing symptoms. After the entry of the attachment variables in step three, the model remained statistically significant F(14, 379) = 2.69; p < .01 and explained 5.7% of the variance in child internalizing symptoms. Finally, in step four, with the entry of the interaction variable, the model was statistically significant F(15, 378) = 2.53; p < .01 and explained 5.5% of the variance in child internalizing symptoms.

Discussion

The present study tested the relationship between children’s attachment quality with caregivers (i.e., biological parents and substitute caregivers) and children’s internalizing symptoms among maltreated preadolescents recently placed in out-of-home care. Children’s attachment quality with caregivers was significantly associated with child-report and substitute caregiver-report of children’s internalizing symptoms, such that children who reported greater attachment quality with caregivers tended to exhibit fewer internalizing symptoms. Above and beyond multiple maltreatment and placement experience variables, children’s attachment quality with their parents and caregivers accounted for a significant proportion of variance in children’s internalizing symptoms – roughly 10% of the variance in child–reported anxiety and depression symptoms, and 3% of the variance in caregiver-reported internalizing symptoms. These findings support our first hypothesis and indicate that attachment quality with both caregivers is associated with child internalizing symptoms over and above one another. These findings are also consistent with theoretical and empirical studies that suggest that a high quality attachment relationship with a caregiver is a key determinant of positive child adjustment (Bowlby, 1969; Masten & Coatsworth, 1998).

Although both relationships (i.e., attachment quality with biological parent and substitute caregiver) appear to influence children’s internalizing symptoms independently, children’s attachment quality with substitute caregivers was more strongly associated with children’s internalizing symptoms than children’s attachment with biological parents. Because all children were living with their substitute caregivers at the time of the study, it may be that attachment quality with the current primary caregiver (i.e., substitute caregivers) is more strongly associated with children’s current mental health functioning. While both relationships are important to attend to, children’s relationships with the caregiver with whom they are currently residing with may be especially important for maltreated children’s current internalizing symptomology.

We hypothesized that quality attachments with substitute caregivers would moderate the association between attachment quality with biological parents and children’s internalizing symptoms. No significant moderation effects were found. It appears children’s relationships with biological parents and substitute caregivers are unique and both independently contribute to children’s internalizing symptoms. This is further supported by our bivariate analyses, which indicated that children’s attachment quality with biological parents was only modestly correlated with children’s attachment quality with substitute caregivers, suggesting that relationship histories with the biological parent may not be strongly associated with the quality of subsequent attachment experiences with the substitute caregiver. This finding is also consistent with Joseph et al. (2014) study, which found that some children report secure attachments with their substitute caregivers despite also reporting insecure attachments with their biological parents (Joseph et al., 2014). There was, however, a marginally significant interaction between children’s attachment quality with substitute caregivers and children’s attachment quality with biological parents in predicting children’s depression symptoms (p = .096), such that a high quality attachment with a substitute caregiver buffered the negative impact of a poor attachment with the biological parent on depressive symptoms. Although no inferences can be made due to the marginally significant association and the cross-sectional design of the study, future research should continue to examine whether children’s attachment quality with substitute caregivers may buffer the relationship between children’s poor attachment quality with biological parents and children’s increased depressive symptoms. Perhaps by enhancing substitute caregiver-child attachment for maltreated children in out-of-home care, the impact of poor attachment with biological parents on children’s internalizing symptoms will be attenuated.

Regarding child demographic and placement factors, age was significantly associated with child internalizing symptoms, such that older youth tended to report fewer anxiety and depression symptoms (both bivariately and in the regression models). Few of the maltreatment types were significantly related to children’s internalizing symptoms. Most notably, sexual abuse was associated with depression. Children’s internalizing symptoms differed by placement type, such that placement in foster care was associated with more internalizing problems based on caregiver report. Some studies indicate that children in kinship evidence fewer emotional problems relative to children in foster care (Hegar & Rosenthal, 2009; Keller et al., 2001). Yet, other studies find placement type is unrelated to mental health outcomes when other out-of-home care factors are taken into account (Fechter-Leggett & O’Brien, 2010). It is difficult to tease apart whether kinship care, per se, leads to fewer internalizing symptoms, as placement in kinship care is not random and these children likely differ from children placed in foster care in important ways (e.g., greater contact with biological parent, placement stability). Regardless, children’s attachment quality with caregivers is positively associated with children’s fewer internalizing symptoms over and above placement type and other out-of-home care factors.

Several limitations of the current study should be noted. First, because of the cross-sectional and non-experimental design, we are unable to attribute causation to attachment quality; internalizing symptoms may have preceded attachment quality. Indeed, children who exhibit greater internalizing symptoms may have a more difficult time forming positive attachments with their substitute caregivers. Second, children were living with their substitute caregivers for an average of six months; therefore, the results are not generalizable to maltreated children who have been placed in out-of-home care for a longer period of time. To assess whether results differ based on the length of time children were living with their current caregiver, several post-hoc analyses were conducted. First we conducted a subgroup analysis by examining whether results differed among youth living with their substitute caregiver for six months or longer relative to youth living with their substitute caregiver for fewer than six months. Findings revealed no significant differences in the relationship between children’s attachment quality with both the biological parent and substitute caregiver and children’s internalizing symptoms. Although not statistically different, youth living with their substitute caregivers for six months or longer, on average, reported greater attachment quality (M = 38.8; SD = 4.5) relative to youth living with their substitute caregiver for fewer than six months (M = 38.3; SD = 5.5). Additionally, substitute caregivers who have been living with the child for a shorter duration may not know the child well enough to accurately report and/or have a strong sense of the child’s internalizing symptoms. Post-hoc analyses, however, did not reveal a significant difference in the mean number of internalizing symptoms reported by caregivers living with the child for six months or longer versus caregivers living with youth fewer than six months, t (444) = −0.37, p = ns. Third, a large proportion of the sample identified a Hispanic/Latino, which may not be generalizable to other child welfare populations. Given the increasing size of the Latino population in the United States, it is important to attend to potential cultural differences and how this may be related to attachment quality and maltreatment. Fourth, attachment quality was assessed through child-report only. An observational measure or multi-informant measure would have provided a more comprehensive picture of the attachment relationship. Yet, children appear to be making distinctions between attachment quality with biological parents and substitute caregivers, as children’s reports were not highly correlated with one another.

The current study does, however, overcome several limitations of previous research. We utilized both child and caregiver report of child internalizing symptoms. This helps to confirm our hypothesis that child-reported attachment quality is associated with child internalizing symptoms, and is not simply due to shared method variance. Strengths also include the use of a relatively large sample size for a high-risk population and a high recruitment rate for a hard-to-reach population. Past studies examining attachment quality among foster children had relatively small sample sizes. Additionally, we included the child’s assessment of attachment to both the biological parent and to the substitute caregiver, which we examined simultaneously, allowing us to see if one relationship over the other appeared to be a stronger indicator of child internalizing symptoms. Finally, we accounted for multiple maltreatment and placement variables (e.g., type of maltreatment, length in current placement, frequency of visitation, number of caregiver transitions) that were related to both child internalizing symptoms and attachment quality with caregivers, which have often been neglected in previous studies.

The current study’s findings suggest several directions for future research and clinical work. Future studies should examine the relationship between attachment quality and child internalizing symptoms in a longitudinal study, controlling for baseline mental health functioning and placement variables in order to establish a temporal relationship between children’s attachment quality with caregivers (i.e., biological parents and substitute caregivers) and children’s internalizing symptoms. If our findings are replicated in a longitudinal design, it may suggest that clinical interventions should focus on promoting positive attachment experiences between maltreated children and both their biological parents and substitute caregivers. It would be especially valuable to assess the association between children’s attachment and internalizing symptoms over time in order to determine if one relationship (i.e., biological parents or substitute caregivers) is more predictive of children’s current and later mental health functioning. It may be that maltreated children’s early attachment experiences with their biological parents will be more predictive of children’s internalizing symptoms over time.

If future studies find that a positive relationship with a substitute caregiver ameliorates the negative impact of poor early parental care, it may suggest that a positive relationship with a substitute caregiver serves as a protective factor for maltreated children recently removed from their homes. Intervention programs that attempt to form attachment relationships between substitute caregivers and children may be one avenue to prevent increased internalizing problems among maltreated children and potentially disrupt the negative impact of poor attachment quality with biological parents on children’s mental health outcomes.

Attachment and Biobehavioral Catch-up (ABC; Dozier, Lindhiem, & Ackerman, 2005) may be one intervention that is particularly well suited to enhance attachment quality between substitute caregivers and maltreated children. Stovall-McClough and Dozier (2004) show that children placed with autonomous caregivers (i.e., available and responsive to the child’s needs) as opposed to non-autonomous caregivers (i.e., reject and/or inconsistently respond to child’s needs), are more likely to form quality attachments to their substitute caregivers (Stovall-McClough & Dozier, 2004). In order to help substitute caregivers exhibit autonomous behaviors, the ABC intervention trains substitute caregivers on how to provide nurturing care even when children exhibit distressing behaviors, while also providing a responsive and predictable environment (Dozier et al., 2005). Results from the study show that children whose caregivers received the intervention showed significantly fewer signs of insecure attachments than children whose caregivers received the comparison, educational intervention (Dozier, Lindhiem, Lewis, Bick, Bernard, & Peloso, 2009). The program has also been effective for high-risk biological parents who have been referred to child protective services, such that children living at home who were randomly assigned to the ABC intervention had lower rates of disorganized attachment to biological parents compared to children in the control group (Bernard et al., 2012). Thus, this program may also be effective for improving attachment quality between children living at home and biological parents. Unfortunately, ABC was designed for younger children (i.e., infants,) and may not be applicable to older children (i.e., preadolescents). Further, although the program has shown enhancements in attachment quality and other domains (e.g., cortisol regulation, emotion expression), it is unclear if the intervention is associated with fewer internalizing symptoms among maltreated children.

Although interventions targeting attachments have primarily focused on infants and young children, our findings suggest that older children such as preadolescents are able to develop positive relationships with caregivers and that these relationships are related to fewer internalizing symptoms. The Attachment, Self-Regulation and Competency (ARC) intervention may be more suited for this age range. ARC focuses on three domains including self-regulation, competency (e.g., executive functions, self-development), and attachment for children exposed to complex psychological trauma such as maltreatment (Blaustein & Kinniburgh, 2010; Kinniburgh, Blaustein, Spinazzola, Van der Kolk, 2005). Evaluations of the treatment have found it successful in reducing PTSD symptoms, externalizing and internalizing symptoms among adolescents placed in residential care (Hodgdon, Kinniburgh, Gabowitz, Blaustein, & Spinazzola, 2013) and young children involved in child protective services due to maltreatment (Arvidson et al., 2011). Although findings of ARC are promising, it is still unclear how the treatment affects children’s attachment relationships with caregivers, and whether it is through this mechanism that fewer internalizing symptoms emerge.

Further examination of what key factors may foster a quality attachment relationship with biological and substitute caregivers within the first year of out-of-home care is needed. A current review of foster parent training programs notes that there are major weaknesses in the current literature, as the mechanisms by which these programs may influence quality relationships between substitute caregivers and foster children are unknown (Festinger & Baker, 2013). The current study takes the first step in demonstrating an association, above and beyond maltreatment type and out-of-home care factors, between children’s attachment quality with caregivers (i.e., biological and substitute caregivers) and children’s internalizing symptoms.