The goal of reunifying children in foster care with their families of origin was made a clear priority in child welfare systems in the Adoption Assistance and Child Welfare Act of 1980 (P.L.96-272). The law acknowledged the importance of family and the rights of parents to care for their children, while recognizing situations when supervening parental authority is in the child's best interest. States were required to make reasonable efforts to prevent the need for removal of children from their families, and to make it possible to return them home when removal had occurred. The presumptive course of action, then, was to promote reunification of children in foster care with their biological parents before considering an alternate permanent placement. This policy has largely been driven by philosophical beliefs that children are vulnerable to poor developmental outcomes when they have less contact with their biological families (Gelles, 1993; Maluccio, Abramczyk, & Thomlison, 1996). Furthermore, reunification e fforts were hoped to diminish the threat of a generation of children growing up in foster care, experiencing instability and repeated losses in the form of multiple placement changes (Newton, Litrownik, & Landsverk, 2000).
However, concerns about risks inherent in the universal application of reunification efforts were intimated in the Adoption and Safe Families Act of 1997, which amended the previous law to refocus the priority from the inviolability of the family to the well-being of the child. Section 101 of the Act states that, "Efforts to preserve and reunify the family shall not include certain parents if they pose a serious risk to the child's health and safety." Investigators have documented failures of Child Protective Service (CPS) reunification efforts largely in terms of rates of re-entry to foster care (e.g., Jones, 1998; Terling, 1999; Turner, 1986). However, less is known about the implications of reunifying children in foster care on developmental psychopathology. Studying trajectories of child psychosocial adjustment while in foster care and following reunification has been identified as an important research priority (Maluccio, Fein, & Davis, 1992).
There is evidence that a child's emotional and behavioral profile influences the likelihood of reunification from out-of-home care in the first place. Studies have demonstrated that children in foster care who are selected to be reunified with their birth families initially demonstrate fewer behavior problems compared to children who remain in substitute care (Landsverk, Davis, Ganger, Newton, & Johnson, 1996; Lawder, Poulin, & Andrews, 1986). Landsverk et al. (1996) found that children in foster care with externalizing behavior problems (foster parent reports above borderline severity on the Child Behavior Checklist) were only one half as likely to be reunified within 18 months of removal than children without these problems. The authors speculate that children with poorer, psychosocial adjustment are less likely to be reunified with their birth parents because the child's functioning may be perceived as a reflection of the levels of maltreatment or parenting competence in the home from which the child was r emoved. Another possibility is that children with behavioral disturbance (particularly externalizing problems) may be perceived as presenting more difficulties for the prospective reunified family.
Unfortunately, while children selected for reunification are better adjusted initially, these differences may not be maintained over time relative to children who remain in out-of-home care. Studies have reported that children who are not reunified experience better long-term outcomes, including gains in performance on intelligence tests (Fanshel & Shinn, 1978) and lower rates of criminal recidivism (Jonson-Reid & Barth, in press). Overtime, reunified children appear to develop more behavioral problems than their counterparts who remain in substitute care (Taussig, Clymen, & Landsverk, 2001). In their 6-year follow-up study of 149 children aged 7-12 years entering foster care, Taussig et al. (2001) found that youth evidenced significantly more behavior problems when reunified with their birth families than when not reunified.
Clearly, further investigation is necessary to explore determinants of these observed negative outcomes of reunification on child psychopathology. Our legal and social service systems uphold the value of preserving and reunifying families whenever possible. It is imperative to identify the mechanisms by which reunification is associated with child and family outcomes. Previous work conducted by Litrownik, Newton, Mitchell, Richardson, and Landsverk (in review) compared children who continued out-of-home placement to those who had been reunified with their biological families. The authors report that the reunified children were exposed to a variety of risk factors for maladjustment, including heightened exposure to violence, more maternal health problems, poorer family functioning, and lower levels of parental social support.
This study was based on data derived from the same larger study of children in foster care as the Taussig et al. (2001) and Litrownik et al. (in review) studies described above. Taussig et al. illustrated that reunification of children from foster care with their biological parents is associated with increased externalizing problems. Litrownik et al.'s cross-sectional study, suggested possible environmental characteristics that may be responsible for this exacerbation of problems in reunified households. The current study expands on these previous works in three distinct ways. First, in the current study we sought to elucidate how and when reunification can fail these vulnerable children by building a mediational model of the effects of reunification on later child symptomatology, thereby bridging previous findings noting differences in eventual symptomatology with findings regarding differences in environmental characteristics associated with reunified homes. Second, we focused our attention on a child behav ioral syndrome that has yet to be examined in the literature on reunification outcomes, namely, internalizing child behavior problems. Third, expanding upon previous findings we examined mediators of child outcomes that may increase emotional and behavioral problems in reunified children, as well as processes that may lower risk for later problems. Hence, in the current longitudinal study we broadened our investigation to include potentially harmful and salutary consequences of reunification.
Just as the removal of a child from their biological family is associated with a host of changes in a child's life, so too is the subsequent reunification to that family of origin. This cascade of changes is not unitarily positive or negative, rather it is possible to enumerate a variety of risks and benefits associated with reunification. In this paper, we review some possible effects of reunification on children's psychosocial functioning, attending to candidate consequences that may be either salutary or harmful. We then test an integrative model of the longitudinal effects of reunification on trajectories of child psychosocial functioning. The concepts of multifinality and equifinality can be invoked here to suggest that multiple pathways can lead to both resilience and maladaptation in the context of the reunification process.
Children's Social Isolation and Reunification
From the child's perspective, reunification with the birth family may represent fulfillment of the child's most heartfelt hopes. Although there is little research examining preferences for reunification among children in foster care, there is some evidence that children often hold strong attachments to their biological parents even when in foster care for long periods (Poulin, 1985). Moreover, the emotional quality of the attachment of children in foster care to their biological parents appears to confer resilience in terms of their subsequent ability to form attachments to foster parents and the development of fewer internalizing symptoms (Milan & Pinderhughes, 2000).
Folman (1998) has begun to investigate the child's experience of removal from their biological parents using qualitative interviews of 90 inner-city children in foster care. Youths' narratives depicted a progression of successively traumatic events beginning with their initial removal and continuing through the placement process. While reunification may not repair damage from these traumas, it may alleviate hurtful experiences associated with out-of-home placement, including fearfulness of further placement changes and the pain of separation from biological family. It is possible that one salutary effect of reunifying children with their families of origin may be an attenuation of children's feelings of social isolation and loneliness that can be associated with out-of-home placement. Although parental perceived social support may be lower in reunified households (Litrownik et al., in review), it is possible nonetheless that children may feel greater levels of interpersonal connectedness when returning to the ir families of origin than when remaining in substitute care. This hypothesis has not been examined in previous studies of reunification.
Family Stress and Reunification
A familiar and intuitive explanation for child maltreatment is that the parent was under stress and as a result mistreated his/her children. Ecological/transactional models describe how environmental conditions, caregiver attributes and child characteristics operate reciprocally to contribute to risk of maltreatment (e.g., Belsky, 1993; Bronfenbrenner, 1979; Cicchetti & Rizley, 1981). Abuse and neglect are thought to occur when risk factors, such as stress, overwhelm available compensating factors, such as social support (Cicchetti & Rizley, 1981). Research evidence has generally supported this explanation. Early prospective studies found that the number of stressful life events differentiated families reported for child abuse (Egeland, Breitenbucher, & Rosenberg, 1980) and that the number of experienced stresses in a parent's life is associated with the number of maltreatment incidents reported (Herrenkohl & Herrenkohl, 1981). These investigators characterized abusive families as disorganized, chaotic, less able to remove themselves from problematic circumstances, and less skilled at avoiding further difficulties. These findings have received further support from more recent studies linking stressful life events, daily hassles and maltreatment risk (e.g., Kolko, Kazdin, Thomas, & Day, 1993; Kotch, Browne, Ringwalt, Dufort & Ruina, 1997; Spicer & Franklin, 1994).
One risk inherent in reunification is the potential danger of returning children to environments burdened by stress and ill equipped to effectively cope. Family reunification programs often provide services to biological parents that focus on relationship building, parenting skills training, behavioral interventions, and assistance meeting families' concrete needs (Fein & Staff, 1993; Fraser, Walton, Lewis, & Pecora, 1996; Gillespie, Byrne, Workman, 1995; Landy & Munro, 1998). There is mounting data questioning the effectiveness of these programs in preventing further abuse and reentry into out-of-home care even with short-term follow-up periods (Frame, Berrick, Brodowski, 2000; Landy & Munro, 1998; Terling, 1999). Existing programs may not be sufficient to alleviate the stressful circumstances in these homes. There is clear research evidence indicating that elevated levels of family stress (independent of maltreatment) negatively impact child psychosocial functioning (Costello, 1989; Jensen, Richters, Ussery , & Bloedau, 1991; Mathijssen, Koot, & Verhulst, 1999). In so far as families continue to lack the skills necessary to control the disorganization, stress and chaos that initially contributed to abuse and neglect, it is unlikely that return to this environment will be beneficial for the child. These types of concerns have led some advocates to doubt that family reunification policies are in the best interest of children (Gelles, 1993). Whether children are actually better off when returned to a family previously identified as abusive remains in question (Berliner, 1993).
Mental Health Service Use and Reunification
There is compelling evidence that rates of psychopathology among children in foster care are very high. Estimates of the proportion of children meeting clinical criteria for behavioral problems or psychiatric diagnosis range from 54 to 80% (Clausen, Landsverk, Ganger, Chadwick, & Litrownik, 1998; Halfon, Mendonca, & Berkowitz, 1995; Urquiza, Wirtz, Peterson, & Singer, 1994). These rates are not surprising given the many risk factors associated with foster placement, most notably a history of maltreatment. Blumberg, Landsverk, Ellis-MacLeod, Ganger, & Culver, (1996) demonstrate that many of these children first begin receiving psychological treatment after recognition of their needs upon entry to foster care. In a cohort of 662 children placed in out-of-home care, 56% used mental health services in the first 5-8 months following entry into out-of-home care, while less than 10% had received mental health services prior to entry (Blumberg et al., 1996).
Landsverk and Garland (1999) articulate the need to study the impact of exits from foster care on the continuity of psychological treatment for these high-risk children. It is possible that children are more likely to receive mental health services when they are in the foster care system. Involvement in the foster care system may ensure closer monitoring of children's functioning and facilitation of access to care. In fact, there is evidence that children in less closely monitored kinship placements are less likely to receive treatment compared to children in nonrelative foster care (Leslie, et al., 2000). It is conceivable that exit from foster care and return to the biological parent(s) would similarly result in fewer assurances that children receive needed mental health treatment, which in turn may result in poorer psychosocial outcomes for reunified children.
Hypothesized Model: Direct and Mediated Pathways Between Reunification and Child Adjustment
In this study, we examined four candidate child functioning outcomes that may be impacted by the experience of reunification from out-of-home care. Specifically, we explored the relationship between reunification (prior to age 4) and the outcomes of internalizing behavior problems, social isolation, stressful life events (all at age 6), and lifetime receipt of mental health treatment. Furthermore, we propose that the impact of reunification on child behavior problems may be mediated by its impact on stress, social isolation, and access to treatment.
First, we expected reunification to be related to decreases in the child's feelings of social isolation. A child's return to their family of origin may be associated with an alleviation of loneliness experiences during separation. Second, because children may be returning to environments with unresolved difficulties, we hypothesized that reunification may be related to increased exposure to stressors related to instability, physical harm, and family dysfunction. Compared to the child remaining in out-of-home (or substitute) care, the reunified child may be subject to stresses in their living environment, such as greater mobility, more frequent school transitions, changes in family composition , greater familial health and injury risks, exposure to family conflict, and the like. Third, we posit that children reuniting with their biological families may access mental health treatment at a reduced rate relative to children who remain in foster care. Decreases in monitoring and service associated with reunificati on may result in situations where needed treatment is not received.
In turn, the direct effects of reunification on social isolation, stress, and mental health service utilization are then thought to impact child adjustment in terms of internalizing problems. Internalizing behavior problems were selected as the primary symptom outcome indicator because these problems do not appear to be associated with CPS decisions to reunify children with their biological families. Evidence indicates that externalizing or disruptive behavior problems are related to reunification decisions, while internalizing problems may be more evenly distributed among children remaining in placement and reunified children (Landsverk et al., 1996). By studying the effects of reunification on subsequent internalizing problems, we hope to study outcomes that are less confounded with initial group differences in adjustment.
In summary, the effects of reunification on child psychosocial functioning were conceived of as multifaceted and complex, with both salutary and harmful effects involved. We hypothesized that the impact of reunification on child internalizing symptoms is mediated by its more proximal effects on social isolation, familial stressors and mental health services utilization. This hypothesized model is depicted in Fig. 1. Hypothesized direct and indirect effects of reunification on later child internalizing problems are depicted in bold arrows.
The sample for this study was drawn from a larger sample of children aged 0-16 years who were recruited between May 1990 and October 1991 in a study of the mental health needs of children in foster care in San Diego County (Landsverk, Litrownik, Newton, Ganger, & Remmer, 1996). The cohort of children eligible for the Foster Care Mental Health (FCMH) study were the approximately 5000 children experiencing their first episode of out-of-home care entering through San Diego's primary emergency sheltered care facility in the 18-month study referral period. One thousand two hundred and twenty-one (1221) of these children were legally placed in out-ofhome care for at least 5 months and were included in the original FCMH study. Of these, 532 children who entered the child welfare system before the age of 3.5 years were recruited for this study. A total of 319 children were successfully recruited and completed baseline assessments at age 4. These participants were representative of the larger cohort of 532 eligible in terms of gender, race/ethnicity, and type of placement, including reunification outcome. Of these 319 children, follow-up assessments at age 6 were completed for 287 children. Inclusion in this study required that children were in the same placement from age 4-6 years. This selection criterion was in place for two reasons: (a) to ensure the same primary caregiver was reporting on outcomes and (b) to control for the possibility of placement changes during the study period contributing to changes in child functioning over time. The final sample under study included 218 children who remained in the same placement from age 4-6 years.
Sample Demo graphics
There was a fairly even gender distribution in the sample, with 103 (47.2%) boys and 115 (52.8%) girls. The study sample reflected the ethnic diversity of the larger community of San Diego County, and the additional overrepresentation of ethnic minority representation common in the child welfare populations. The sample was comprised of 64 (29.4%) Caucasian, 82 (37.6%) African American, 34 (15.6%) Hispanic, 3 (1.4%) Asian American, 34 (15.6%) multiracial children, and 1 (0.5%) child in the Other race/ethnicity category. By the time of the baseline interview at age 4, children were in one of four types of placement outcomes: 78 (35.8%) had been reunified with their biological parents, 56 (25.7%) had been adopted, 39(17.9%) were in kinship foster care, and 45 (20.6%) were living in nonrelative foster care. Of the 140 children who were not reunified at the time of the age 4 interview, 9 had experienced a previously unsuccessful reunification attempt.
Table I displays characteristics of the total sample and of the reunified and non-reunified subsamples. The reunified and non-reunified children did not differ in terms of gender distribution and reasons for entry into the foster care system. The most common reasons for removal from the biological parents included neglect (22%), caretaker absence or incapacity (20%), and a positive toxicology screen of the child at birth indicating prenatal substance exposure (29%). By the age 4 interview, reunified children had experienced more placements than non-reunified children (t = 3.45, p < .01). Reunified children tended to have been older at the time they were originally removed and placed into foster care (t = 4.62, p < .01). Within the reunification group, children had been reunified with their biological parents an average of 22.9 months by the time of the baseline age 4 interview.
Two latent constructs were present in the proposed model including social isolation and stress. The latent construct of social isolation was formed by an index of the number of supportive figures named by the child and a scale of loneliness and social dissatisfaction. The latent construct of stress was composed of indices of the number of stressful life events in each of three domains. In addition, the model included the measured variables of reunification status, internalizing behavior problems (age 4-6) and mental health treatment utilization.
Two indicators were used conjointly to measure the child's feelings of social isolation, including a measure of loneliness and an indicator of perceived social support. We combined these indicators because theoretically and empirically, the relationship between low perceived social support and feelings of loneliness is well established across child, adolescent and adult populations (Gaudin, Polansky, Kilpatrick, & Shilton, 1993; Ginter, Lufi & Dwinell, 1996; Henwood & Solano, 1994; Pierce, Samson, & Sarason, 1991). More specifically, social network size and density have been found to be associated with perceived loneliness (Bell, 1991; Berg & McQuinn, 1989). The number of supportive figures and loneliness were significantly correlated (r = .34, p < .001) and the loadings of these measured variables on the common latent construct of social isolation were well within the range noted to be acceptable for measurement models in confirmatory factor analysis (Fabrigar, Wegener, MacCallum, & Strahan, 1999).
Number of Supportive Figures. Children were administered a project developed measure, the Inventory of Supportive Figures (1SF), which is based on the Purdue Social Support Scale (Burge & Figley, 1987). Respondents are asked to list up to three individuals who have been helpful to them, and rate the quality of their support on four dimensions. As an index of perceived support, we used the number of supportive figures named by the child (range 0-3). This item was reverse coded to yield an indicator of social isolation. In the current sample there is support for the concurrent and predictive validity of this indicator of perceived network size. The number of supportive figures reported by the child was correlated with parent-reported problems with social functioning on the CBCL (r = .32, p = .001), and with child reported anxiety on the Trauma Symptom Checklist for Children administered 2 years later (r = .28, p = .01).
Loneliness and Social Dissatisfaction. The second indicator was derived from the Loneliness and Social Dissatisfaction Scale (LSDS, Cassidy & Asher, 1992). This 24-item questionnaire includes 16 questions assessing the child's loneliness and social dissatisfaction primarily in peer relations (e.g., Are you lonely at school?), the remaining eight items are fillers focusing on children's hobbies and activities. Children are asked to rate the extent to which each statement is true by saying "yes," "no," or "sometimes." Ten items are reverse scored, filler items are omitted and a total score is obtained by summing all loneliness and social dissatisfaction items, yielding a range of 16-48. This scale has satisfactory internal consistency reliability ([alpha] = .79), and validity with significant correlations between total LSDS score and sociometric measures of peer status as well as teacher reports of child social behavior (Cassidy & Asher, 1992).
Three measured variables were used to comprise the latent construct of stress. Caregivers completed the project-developed measure, Child Life Events Scale. This instrument contains 31 items covering events including modifications to the family structure, upheavals in living arrangements, sickness, injury or death of people close to the child, sickness or injury experienced by the child, school changes, legal problems in the family, and the child's exposure to conflict and violence, and family accomplishments. Parents or caregivers were asked to indicate whether the event in question occurred in the last year. In terms of validity, this measure was found to be correlated (r = .25, p < .0001) with the Everyday Stressors Index (Hall, Williams, & Greenberg, 1985).
We constructed composite scores tapping negative life events in the domains of family dysfunction, instability and, experiences of harm or injury to the child or family members. These domains were derived from a principal components factor analysis with an orthogonal rotation. Each event was compiled into the factor upon which it had the highest factor loading. Composite variables were computed by summing the endorsed events pertinent to each domain. Internal consistency of the composite scores was low with alphas ranging from .46 to .51. This was due to highly restricted variance on a number of these dichotomous variables, some of which were endorsed by fewer than 5% of respondents. Bivariate correlations between composite scores ranged from r = .37 (p < .001) to r = .42 (p <.001).
Family Dysfunction. The life event items that loaded onto this domain included separations, divorces, family member arrested, family member incarcerated, child witnessing loud long arguments between family members.
Harm to Self or Others. This domain comprised of items measuring the occurrence of events such as someone close to the child dying or suffering a serious accident or illness, the child suffering a serious accident or illness, the family being a victim of a property crime, the child witnessing someone being threatened with a weapon, shot or stabbed, killed, sexually assaulted, or otherwise physically harmed. A number of the violence items were endorsed very rarely and thus were not included in the principal components analysis, but were conceptually linked to this domain and were thus included.
Instability. Instability was indicated by the occurrence of events including having new children in the home, someone moving into or Out of the home, the family moving to a new place, the family being evicted or the child being homeless, the child changing schools, and household finances getting worse.
Placement information and reunification status were obtained from Child Protective Services (CPS) chart abstraction and administrative databases, and confirmed by the child and caregiver at subsequent interviews. The main variable of interest in this model was the child's reunification status by age 4 years.
Internalizing Behavior Problems
Caregivers reported on child behavior problems on the Child Behavior Checklist (CBCL; Achenbach, & Edelbrock, 1991). The CBCL consists of 118 child problem behaviors rated on a 3-point scale (0 - not true, 2 - somewhat or sometimes true, 3 - very true or often true). The CBCL yields broad band factor scores for (1) Internalizing problems (including narrow-band syndromes of withdrawn behaviors, somatic complaints, and anxious/depressed behaviors) and (2) Externalizing problems (including aggressive and delinquent behaviors), as well as a Total problems score. The CBCL is a reliable measure with alphas for the Internalizing and Externalizing factors and Total problems score are 0.89, 0.93, and 0.96, respectively (Achenbach & Edelbrock, 1991). One week test-retest reliability obtained from a community sample of mothers ranged from 0.89 for Internalizing problems score to 0.93 for the Externalizing and Total problems scores (Achenbach & Edelbrock, 1991). The criterion validity of the CBCL has been demonstrated by its ability to discriminate between referred and nonreferred children (Achenbach & Edelbrock, 1991).
Broad-band scores on the Internalizing scale at age 4 and age 6 were utilized as the indicators of child psychosocial functioning in this study. By including this measure at age 4 and again at 2-year follow-up, we could control for the child's baseline level of behavior problems and meaningfully determine longitudinal change. Further, this design allowed us to examine relations between earlier functioning and receipt of mental health services and later outcomes of stress and social isolation at age 6.
Mental Health Treatment
Children's mental health service utilization was assessed using the project-developed measure of Service Utilization administered at the baseline caregiver/parent interview. This instrument was designed to assess reasons for seeking services, types of service provider seen, number of visits and satisfaction with services received. Adults are asked about services that have ever been sought and received to address the child's emotional, behavior, and school problems. In the current study, a lifetime index of specialty outpatient mental health service utilization was computed by summing the number of different mental health provider types seen by the child up to the age 4 interview. These provider types included mental health professional, school counselor, speech and language therapist, and developmental evaluation specialist. The internal consistency of the items in this composite score was fair ([alpha] = .70). The range of corrected item total correlations (CITC) was from .24 to .60 (mean CITC = .44).
Construction of the Structural Equation Model
As previously described, the goal of this study was to examine an array of possible effects of reunification on child psychosocial functioning outcomes. We posit that reunification exerts a number of influences on children, some of which may be positive and some negative. It is proposed that the relationship between reunification and children's later internalizing behavior problems is mediated through more proximal effects on children's feelings of social isolation, experience of family stress, and utilization of mental health services. The proposed relationships are depicted in Fig. 1. Each rectangle represents a measured variable (constructed as described in the methods section). The single headed arrows represent hypothesized direct effects. This model was subjected to structural equation modeling, developed by Bentler and Weeks (1980), using the EQS program (Bentler, 1995), version 5.7b. Chi square was used to evaluate the goodness of fit of the model. Tests of statistical significance of the estimated pa rameters were set at the .05 level, one-tailed tests were used where hypotheses were directional, two-tailed when there was no directional hypothesis.
There are nine measured variables in the model with 218 observations per variable. The hypothesized model contained 26 parameters to be estimated, including 18 regression coefficients (paths and loadings) and 8 variances. The ratio of cases to observable variables is 24:1; the ratio of cases to parameters is 8:1. Since some of the measured variables were dichotomous or ordinal category and not normally distributed, robust estimation procedures were used to estimate parameters in the model. The robust comparative fit index (CFI), root mean square error of approximation (RMSEA), and a root mean squared residual (RMSR) were examined to evaluate model fit (Hu & Bentler, 1999).
After running the hypothesized model, post hoc model modifications were performed in an attempt to develop a more parsimonious better fitting model, with greater degrees of freedom. The multivariate Lagrange Multiplier test was run to determine if improvements in model fit could be made by adding pathways in the model. The Wald test was executed to determine whether elimination of hypothesized pathways and parameters would result in better model fit. Modifications were made to the hypothesized model based on the results of this post hoc test. Fit indices and tests of significance were reported for the final model that best fit the data.
Table II displays descriptive statistics and bivariate analyses for the reunified and non-reunified groups on each measure entered in the structural model. Group means, standard deviations, t statistics and Cohen's d effect sizes were calculated. Results indicated that reunification was significantly associated with greater parent reported stressful life events in the instability, t(216) = -3.01, p = .003, harm, t(216) = -3.13, p = .002, and family dysfunction areas, t(2 16) = -5.99, p < .001. The associated effect sizes were in the moderate range for harm (d = .54) and instability (d = .41), and in the large range for family dysfunction (d = .78; Cohen, 1988). Reunification was also significantly associated with lower likelihood of receipt of mental health services, t(216) = 2.07, p = 04), this effect was in the small range (d = .30). Finally, reunification was associated with greater perceived social support, t(216) = 2.05, p = 04, this effect was in the small range (d = .30). Reunification was not signific antly associated with internalizing problems at baseline or follow-up, or with child reported loneliness, the effect sizes associated with these dependent variables were in the small range (d = .16-.21).
The measurement model for stressful life events demonstrated adequate fit, Robust CFI = .99, [chi square](1) = 2.16, p = .14. The measurement model for social isolation could not be independently evaluated because a measurement model with two manifest indicators without structural paths to other factors and variables is not by itself identified. However, within the larger structural model the path coefficients associated with each indicator of the social isolation factor are adequate according to criteria defined by Fabrigar et al. (1999). The complete measurement model was tested in one analysis correlating stressful life events and social isolation, loading the five variables onto their respective constructs. The fit of this model was adequate, Robust CFI = .96, [chi square](4) = 7.52, p = .11. Turning now to the structural model, the independence model that tests the hypothesis that the variables are uncorrelated with one another was easily rejected [chi square](36) = 334.82, p < .0001. Next, the hypothesi zed model was tested. A chi-square test indicated that the hypothesized model could not be rejected [chi square](18) = 34.30, p = .01, suggesting that the model may not adequately fit the data. However, some support was obtained for the hypothesized model in terms of a Robust CFI of .95. To improve fit with the data, the following model modifications were employed.
After running the hypothesized model, several post hoc model modifications were performed in an attempt to develop a more parsimonious better fitting model, with greater degrees of freedom. Results of the Wald test indicated that several direct paths were not supported by the data and should be dropped from the model. Pathways dropped from the model included the direct effects of reunification on internalizing behavior problems at age 6; mental health treatment on stress, social isolation and internalizing behavior problems at age 6; and social isolation on internalizing problems at age 6. The multivariate Lagrange Multiplier test indicated that improvements in model fit could be made by adding pathways in the model.
Pathways were added between the stress indicator of instability and social isolation, and between reunification and the stress indicator of family dysfunction. For the final model, a Robust CFI of .99, RMSEA of 0.03, and a standardized RMSR of .04 were obtained. This combination of fit indices reveals that the model fit the data well (Hu & Bentler, 1999). The chi-square test of the model was nonsignificant, scaled [chi square](19) = 20.22, p = .38. Thus, the model cannot be rejected, and it can be stated that the data support the final model. (7) Examination of the Lagrange Multiplier modification indices for this final model suggested that no further paths need be included. The standardized beta coefficients for the pathways in the final model are presented in Fig. 2.
As would be expected, the path between internalizing problems at age 4 and 6 was strong (path coefficient = .56), parent reports of these symptoms remained stable across the 2-year follow-up period. Baseline internalizing symptoms were related to children's lifetime receipt of treatment at baseline, and feelings of social isolation and experiences of stressful life events at follow-up.
Examination of direct effects in the resultant model shows that some of the original model predictions were borne out. First, reunification was associated with increases in stressful life events experienced by the child and family (path coefficient = .30). This increase in stress was in turn related to greater internalizing behavior problems (path coefficient .14). Thus, reunification impacted later internalizing through its effects on familial stress. In order to determine whether this indirect path from reunification to familial stress to internalizing symptoms actually improved the fit of the model, we compared the results of this model to one that omitted this indirect path. We found that the model that omitted the indirect path from reunification to familial stress to internalizing symptoms did not appear to fit the data as well, scaled [chi square](21) = 32.63, p = .05. A scaled difference chi-square test appropriate for nested models (Satorra & Bentler, 1999) indicated that the final model that include d this indirect path resulted in a significant increase in fit, [[chi square].sup.2.sub.diff](2) = 10.60, p = .005. Second, reunification was negatively related to the child's feelings of social isolation (path coefficient = -.21), such that reunified children reported less social isolation in terms of the number of supportive figures and their loneliness and social dissatisfaction. Third, reunification was associated with lower mental health service utilization (path coefficient = -.17).
However, contrary to our hypotheses social isolation did not mediate the relationship between reunification and later internalizing problems as there was no association between social isolation and parent reports of internalizing behaviors at age 6. Similarly, lifetime mental health treatment prior to age 4 was not associated with internalizing symptoms at age 6. As such there was no indirect path of reunification impacting symptoms through treatment.
A few findings were unexpected but not inconsistent with the study hypotheses, and therefore two model modifications were performed. Reunification was specifically associated with increases in family dysfunction related events (path coefficient = .18) above and beyond its association with the variance captured by the latent construct of stressful life events. In other words, reunification was related to familial stress in general, and was also specifically related to heightened problems with family dysfunction (including family conflict, deviant parental behavior, and violence in the home). These problems harken the pathogenic family processes that typically trigger the placement of children into foster care. Second, parent reported stressful events related to family instability were associated with children's experiences of social isolation (path coefficient = .27). This empirical finding makes conceptual sense because stressors including frequent changes in school settings and neighborhoods, and changes in the family composition likely give rise to decreased stability of peer, community, and familial support.
Reunification of children in foster care with their biological parents has long been considered a most favorable disposition for children in placement. Common wisdom places a great deal of value on children being raised by their birth families. Child advocates recoil at the idea of thousands of children being raised in long term substitute care. However, recent studies have depicted categorically negative outcomes for reunified children. Although, children in foster care selected for reunification initially demonstrate fewer behavior problems (Landsverk et al., 1996), over time they appear to deteriorate and develop more behavioral problems than children who remain in out of home placements (Taussig et al., 2001). However, reunification undoubtedly carries with it a host of changes in the child's life that may influence psychosocial functioning in a variety of ways. We examined the possibility that the outcomes of reunification are likely complex and multifaceted.
This study employed structural equation modeling to test a model whereby reunification exerts multiple influences on child outcomes through its effects on stressful life events and perceived social isolation. In the case of internalizing behavior problems, the data suggest that reunification does not directly contribute to more symptomatology. Rather reunification prior to age 4 is associated with children being exposed to more adverse life events by age 6, including exposure to elevated family dysfunction, instability and harm. In turn, stressful life events are associated with greater internalizing problems reported at age 6.
On the other hand, reunification appeared to be associated with lower child perceptions of social isolation at age 6. Our results indicated that reunified children felt they had supportive adults in their lives and reported feeling less lonely at school. However, this effect was smaller in magnitude compared to the associated increases in stressful life events following reunification. And while decreased perceptions of social isolation may be a positive outcome in and of itself, comparatively low levels of social isolation did not lead to decreases in caregiver reported internalizing problems among reunified children. Contrary to our predictions, resolution of feelings of isolation among reunified children did not result in a positive indirect effect on child internalizing symptoms at follow-up. Satisfaction in the social relationship domain did not appear to act as a buffer between reunification and resulting symptomatology.
It is also important to note that the levels of internalizing problems reported in this study fall largely within the normative range. Only 19% of the reunified group and 13% of the non-reunified group were in the borderline clinical range. With regard to internalizing spectrum behavioral problems, many of these children with a history of foster care placement demonstrate resilience. This finding can be contrasted with the robust finding of highly elevated levels of externalizing problems among children in foster care and among children reunified with their biological families (e.g., Clausen et al., 1998; Halfon et al., 1995; Taussig et al., 2001; Urquiza et al., 1994). Of course, this rate of clinical level internalizing problems is still much higher than would be expected in the general population (Achenbach & Edelbrock, 1991).
In terms of the performance of the mental health service system under study, some positive conclusions can be drawn. Baseline internalizing problems predicted previous mental health utilization, but internalizing problems at follow-up were not related to service utilization. This finding can be interpreted in a positive light, in that children with more symptoms were more likely to receive services. Further, after mental health service use occurred, the low use and high use groups were not distinguishable in terms of internalizing behaviors at follow-up.
However, additional concern about reunification comes from the evidence that children who are reunified are less likely to receive mental health services. This effect holds while simultaneously controlling for levels of internalizing symptomatology. This finding is consistent with concerns that reunification is associated with a decrease in monitoring of children's mental health needs and decreased access to care. We can also argue that children who are reunified are on average in greater need of continued services due to the higher levels of family dysfunction, instability and harm to which they are exposed.
The results of this investigation have policy implications for child welfare systems and family reunification programs. The findings presented here echo previous findings of the negative effects associated with the reunification of children in foster care with their birth families. Upon reunification, children appear to be re-exposed to an array of stressful family circumstances and in turn are at risk of greater internalizing symptomatology. Contrary to predictions, there was relatively less support for the assertion that reunification has beneficial effects on child functioning. There was a small but significant ameliorative effect of reunification on social isolation, however, this did not appear to result in resilience from internalizing symptoms.
However, the pattern of results suggests that reunification need not directly result in greater internalizing dysfunction. Children in reunified families are subject to maladjustment to the extent that they are exposed to stressful living environments. These environments seem prone to higher levels of family conflict, upheavals in family composition, legal problems, mobility, injury, illness, and exposure to violence, as these types of events comprised our measure of stressful life events. Moreover, these children are not receiving treatment at levels commensurate with their counterparts remaining in substitute care. Indeed, these children appear lost to the system. It seems that child welfare and mental health service systems are in a prime position to intervene with the specific risks of reunification identified here. With appropriate service allocation and effective prevention efforts, these reunified children could likely maintain the relatively high psychosocial functioning with which they enter the reun ification process. In light of our examination of potential mediators of outcomes, we conclude that the types of risk presented with reunification may be amenable to supportive interventions provided during and after reunification. Unfortunately, the data also suggest that reunified children seem not to have equal access to potentially protective interventions.
However, careful attention to these data raises the possibility of alternate interpretations of the findings. One limitation of this study is that we cannot rule out the possibility that differences in the experiences of the reunified children prior to returning to their biological families may have contributed to the observed outcomes. For example, we noted that children who were later reunified were older at the time they were removed from their birth parents. It is possible that children who experience the trauma of removal at an early age are more resilient because they had not yet developed cognitive representations of their attachments to their biological caregivers, and they may have experienced maltreatment during preverbal stages and do not retain vivid memories of the abuse or neglect. In contrast, reunified children may have been at greater risk by virtue of being older at the time of these multiple traumas. However, these suppositions are not supported by the empirical literature. Earlier age of onset of maltreatment has been found to be associated with greater psychopathology (e.g., Famularo, Fenton, Kinscherff, Ayoub, & Barnum, 1994), and earlier age at entry to foster care has not been found to be related to enhanced attachment quality with foster caregivers (Dozier, Stovall, Albus & Bates, 2001).
Another potential pre-reunification difference between the groups that may account for the observed outcomes involves exposure to stress within the foster care system. In this sample, children in the reunified group experienced a greater number of placements and placement changes than did the children in the non-reunified group. Newton et al. (2000) reported that multiple changes in placement predicted internalizing and externalizing behavior problems among foster children. In this study, children who were eventually reunified were subjected to more placement disruptions within the system that may have contributed to problems in the reunified group. Future studies are needed to disentangle the contributions of prefoster care history, stressors during foster care, and postreunification experiences to better understand the implications of reunification policies and practices.
[FIGURE 1 OMITTED]
This research was supported by a grant (#90CA1566) to the Consortium for Longitudinal Studies on Child Abuse and Neglect (LONGSCAN) from the Children's Bureau, Office on Child Abuse and Neglect, Administration for Children, Youth and Families.
Received September 21, 2001; revision received October 26, 2002; accepted January 13, 2003
(7.) The structural model was also run using all manifest variables because of concerns about possible disattenuation of effects of reunification on internalizing symptoms stemming from the use of manifest variables for internalizing problems and latent constructs for the proposed mediators. A model using composite variables for social isolation and stressful life events (composites were the sum of standardized scores on the respective indicators) revealed very similar results. The path model fit the data well (Robust CFI = .95, scaled [chi square] = 14 t p = .06), with similar path coefficients to the full model presented.
Achenbach, T. M., & Edelbrock, C. (1991). Manual for the Child Behavior Checklist and 1991 Profile. Burlington, VT: University of Vermont Press, Department of Psychiatry.
Bell, R. A. (1991). Gender, friendship network density, and loneliness. Journal of Social Behavior and Personality, 6, 45-56.
Belsky, J. (1993). Etiology of child maltreatment: A developmental ecological analysis. Psychological Bulletin, 114, 413-434.
Bentler, P. M. (1995). EQS Structural Equations Program Manual. Encino, CA: Multivariate Software.
Bentler, P. M., & Weeks, D.G. (1980). Linear structural equations with latent variables. Psychoinetrika, 45, 289-308.
Berg, J. H., & McQuinn, R. D. (1989). Loneliness and aspects of social support networks. Journal of Social and Personal Relationships, 6, 359-372.
Berliner, L. (1993). Is family preservation in the best interest of the children? Journal of interpersonal Violence, 8, 556-557.
Blumberg, E., Landsverk, J., Ellis-MacLeod, E., Ganger, W., & Culver, S. (1996). Use of the public mental health system by children in foster care: Client characteristics and service use patterns. Journal of Mental Health Administration, 23, 389-405.
Bronfenbrenner, U. (1979). The ecology of human behavior. Cambridge, MA: Harvard University Press.
Burge, S. K., & Figley, C. R. (1987). The Social Support Scale: Development and initial estimates of reliability and validity. Victimnology, An International Journal, 11(4), 231-243.
Cassidy, J., & Asher, S. (1992). Loneliness and peer relations in young children. Child Development, 63, 350-365.
Ciccehetti, D., & Rizley, R. (1981). Developmental perspectives on etiology, intergenerational transmission and sequelae of child maltreatment. New Dire ctions for Child Development: Developmental Perspectives on Child Maltreatment, 11, 31-56.
Clausen, J., Landsverk, J., Ganger, W., Chadwick, D., & Litrownik, A. J. (1998). Mental health problems of children in foster care. Journal of Child and Family Studies,7, 283-296.
Cohen, J. (1988), Statistical power analysis for the behavioral sciences. Hillsdale, NJ: Erlbaum.
Costello, E. J. (1989). Child psychiatric disorders and their correlates: A primary care pediatric sample. Journal of the American Academy of Child and Adolescent Psychiatry, 28, 851-855.
Dozier, M., Stovall, K. C., Albus, K. E., & Bates, B. (2001). Attachment for infants in foster care: The role of caregiver state of mind. Child Development, 72, 1467-1477.
Egeland, B., Breitenbucher, M., & Rosenberg, D. (1980). Prospective study of the significance of life stress in the etiology of child abuse. Journal of consulting and Clinical Psychology, 48(2): 195-205.
Fabrigar, L. R., Wegener, D. T., MacCallum, R. C., & Strahan, E. C. (1999). Evaluating the use of exploratory factor analysis in psychological research. Psychological Methods, 4, 272-299.
Famularo, R., Fenton, T., Kinscherff, R., Ayoub, C., & Barnum, R. (1994). Maternal and child posttraumatic stress disorder in cases of child maltreatment, Child Abuse and Neglect, 18, 27-36.
Fanshel, D., & Shinn, E. B. (1978). Children in foster care: A longitudinal investigation. New York: Columbia University Press.
Fein, E., & Staff, I. (1993). Last best chance: Findings from a reunification services program. Child Welfare, 72, 25-40.
Folman, R. D. (1998). "I was tooken": How children experience removal from their parents preliminary to placement into foster care. Adoption Quarterly, 2, 7-35.
Frame, L., Berrick, J. D., & Brodowski, M. L., (2000). Understanding reentry into out-of-home care for reunified infants. Child Welfare, 79, 339-369.
Fraser, M. W, Walton, E., Lewis, R. E., & Pecora, P. J. (1996). An experiment in family reunification: Correlates of outcomes at one-year follow-up. Children and Youth Services Review, 18, 335-361.
Gaudin, J. M., Polansky, N. A., Kilpatrick, A. C., & Shilton, P. (1993). Loneliness, depression, stress, and social supports in neglectful families. American Journal of Orthopsychiatry, 63, 597-605.
Gelles, R. J. (1993). Family reunification/family preservation: Are children really being protected? Journal of interpersonal Violence, 8, 557-562.
Gillespie, J. M., Byrne, B., & Workman, L. J. (1995). An intensive reunification program for children in foster care. Child and Adolescent Social Work Journal, 12, 2 13-228.
Ginter, E. J., Lufi, D., & Dwinell, P. L. (1996). Loneliness, perceived social support, and anxiety among Israeli adolescents. Psychological Reports, 79, 335-341.
Halfon, N., Mendonca, A., & Berkowitz, G. (1995). Health status of children in foster care: The experience of the Center for the Vulnerable Child. Archives of Pediatric and Adolescent Medicine, 149, 386-392.
Hall, L., Williams, C. A., Greenberg, R. 5. (1985). Supports, stressors and depressive symptoms in mothers of young children. American Journal of Public Health, 75, 518-521.
Henwood, R G., & Solano, C. H. (1994). Loneliness in young children and their parents, Journal of Genetic Psychology, 155, 35-45.
Herrenkohl, E. C., Herrenkohl, R. C. (1981), Some antecedents and developmental consequences of child maltreatment. New Directions for Child Development, 11, 77-92.
Hu, L., & Bentler, P. M. (1999). Cutoff criteria for fit indexes in covariance structure analysis: Conventional criteria versus new alternatives. Structural Equation Modeling, 6, 1-55.
Jensen, P. S., Richters, J., Ussery, T., Bloedau, L. (1991). Child psychopathology and environmental influences: Discrete life events versus ongoing adversity. Journal of the American Academy of Child and Adolescent Psychuiatry, 30, 303-309.
Jones, L. (1998). The social and family correlates of successful reunification of children in foster care. Children and Youth Services Review, 20, 305-323.
Jonson-Reid, M., & Barth, R. P. (in press). From placement to prison: The path to adolescent incarceration from child welfare supervised foster or group care. Children and Youth Services Review.
Kolko, D.J., Kazdin, A. E.,Thomas, A. M., & Day, B. (1993). Heightened child physical abuse potential: Child, parent, and family dysfunction. Journal of interpersonal Violence, 8, 169-192.
Kotch, J. B., Browne, D. C. Ringwalt, C. L., Dufort, V., & Ruina, E. (1997). Stress, social support, and substantiated maltreatment in the second and third years of life. Child Abuse and Neg/ect, 21, 1025-1037.
Landsverk, J., Davis, I., Ganger, W., Newton, R., & Johnson, I. (1996). Impact of psychosocial functioning on reunification from out-of-home placement. Children and Youth Services Review, 18, 447-462.
Landsverk, J., & Garland, A. F. (1999). Foster care and pathways to mental health services. In P. A. Curtis, G. Dale, & J. C. Kendall (Eds.), The foster care crisis: Translating research into policy and practice (pp. 193-210) Lincoln, NE: University of Nebraska Press.
Landsverk, J., Litrownik, A. J., Newton, R., Ganger, W., & Remmer, J. (1996). Psychological impact of child maltreatment. Washington, DC: National Center on Child Abuse and Neglect.
Landy, S., & Munro, S. (1998). Shared parenting: Assessing the success of a foster parent program aimed at family reunification. Chi/d Abuse and Neglect, 22, 305-318.
Lawder, E. A., Poulin, i.E., Andrews, R. G. (1986). A study of 185 children in foster care 5 years after placement. Child Welfare, 65, 241-251.
Leslie, L. K., Landsverk, J., Ezzet-Lofstrom, R., Tschann, J. M., Slymen, D. & Garland, A.F. (2000). Children in foster care: Factors influencing outpatient mental health service use. Child Abuse and Neglect, 24, 465-476.
Litrownik, A. J., Newton, R. R., Mitchell, B. E., Richardson, K. K., & Landsverk, J. (in review). Long-term follow-up of young children placed in foster care: Subsequent placements, child behavior problems and characteristics of these placements at 4-years of age. Manuscript submitted for publication.
Maluccio, A. N., Abramczyk, L. W., & Thomlison, B. (1996). Family reunification of children in out-of-home care: Research perspectives. Children and Youth Services Review, 18, 287-305.
Maluccio, A. N., Fein, E., & Davis, I. (1992). Family reunification: Research findings, issues, and directions. Child Welfare, 73, 489-504.
Mathijssen, J. J. P., Koot, H. M., & Verhulst, F. C. (1999). Predicting change in problem behavior from child and family characteristics and stress in referred children and adolescents. Development and Psychopathology, 11, 305-320.
Milan, S. E., & Pinderhughes, E. E. (2000). Factors influencing maltreated children's early adjustment in foster care. Development and Psychopathology, 12, 63-81.
Newton, R. R., Litrownik, A. J., & Landsverk, J. A. (2000). Children and youth in foster care: Disentangling the relationship between problem behaviors and number of placements. Child Abuse and Neglect, 24, 1363-1374.
Pierce, G. R., Sarason, I. G., & Sarason, B. R. (1991). General and relationship-based perceptions of social support: Are two constructs better than one? Journal of Personality and Social Psychology, 61, 1028-1039.
Poulin, J. E. (1985), Long term foster care, natural family attachment and loyalty conflict. Journal of Social Service Research, 9, 17-29.
Satorra, A., & Bentler, P. M. (1999). A scaled difference chi-square rest statistic for moment structure analysis. UCLA Statistics Series Preprint #260. Retrieved September 3, 2002 from http://www.stat.ucla.edulpaperslpreprints/2601
Spicer, S. J., & Franklin, C. (1994). Exploratory effects of social support, stress and locus of control on the conflict tactics of parents at-risk for child maltreatment. Journal of Social Service Research, 19, 1-22.
Taussig, H. N., Clymen, R., & Landsverk, J. (2001). Children in foster care who return home: A six-year prospective study of behavioral and emotional outcomes in adolescence. Pediatrics, 108, e10.
Terling, T. (1999). The efficacy of family reunification practices: Reentry rates and correlates of reentry for abused and neglected children reunited with their families. Child Abuse and Neglect, 23, 1359-1370.
Turner, J. (1986). Successful reunification of foster care children with their biological parents: Characteristics of parents and children. Child Care Quarterly, 15, 50-54.
Urquiza, A. J., Wirtz, S. J., Peterson, M. S., & Singer, V. A. (1994). Screening and evaluating abused and neglected children entering protective custody. Child Welfare, 123, 155-171.
Anna S. Lau, (1, 2, 6) Alan J. Litrownik, (2, 3) Rae R. Newton, (2, 4) and John Landsverk, (2, 5)
(1.) Department of Psychology, University of California, Los Angeles, California.
(2.) Child and Adolescent Services Research Center, Children's Hospital and Health Center, San Diego, California.
(3.) Department of Psychology, San Diego State University, San Diego, California.
(4.) Department of Sociology, California State University, Fullerton, Fullerton, California.
(5.) School of Social Work, San Diego State University, San Diego, California.
(6.) Address all correspondence to Anna Lau, PhD, Department of Psychology, University of California at Los Angeles, 1285 Franz Hall, Box 951563, Los Angeles, California 90095-1563: e-mail: email@example.com.
COPYRIGHT 2003 Plenum Publishing Corporation
COPYRIGHT 2003 Gale Group