Mentoring children in foster care: Examining relationship histories as moderators of intervention impact on children’s mental health and trauma symptoms

Abstract

Mentoring-based interventions show promise among children in foster care, but previous research suggests that some benefit more than others. Because children in foster care experience relationship disruptions that could affect mentoring effectiveness, we examined whether children’s relational histories at baseline (i.e., relationship quality with birth parents, relationship quality with foster parents, caregiver instability, and previous mentoring experience) moderated the impact of a mentoring intervention on children’s mental health, trauma symptoms, and quality of life. Participants included 426 racially and ethnically diverse children (ages 9-11; 52% male) who participated in a randomized controlled trial of the Fostering Healthy Futures program (FHF), a 9-month one-to-one mentoring and skills group intervention. Results showed that relationship quality with foster parents and prior mentoring experience did not moderate intervention impact. Relationship quality with birth parents and caregiver instability pre-program, however, moderated the effect on some outcomes. The impact on quality of life was stronger for children with weaker birth parent relationships and fewer caregiver changes. Likewise, the impact on trauma symptoms was stronger for those with fewer caregiver changes. Overall, FHF seems to positively impact children with varied relational histories, yet some may derive more benefits – particularly those with fewer caregiver changes pre-program.

Introduction

Children’s exposure to multiple adverse childhood experiences (e.g., abuse, chronic instability, community violence) has a detrimental effect on children’s mental health (Raviv et al., 2010). Rates of mental health, mood, and anxiety disorders are between two to five times higher for children with child welfare involvement than children without such involvement (Scott et al., 2010). Children who experience maltreatment are also more likely to have insecure attachments and negative representations of their birth parents when compared with peers who have not been maltreated (Cyr et al., 2010; Toth et al., 2000). Because children often generalize these early experiences with caregivers to new relationships, exposure to inconsistent and/or inadequate caregiving may result in difficulty forming secure attachments with foster parents or other caregivers (Milan & Pinderhughes, 2000).

Relationships with caring non-parental adults (e.g., within a mentoring relationship) may be key to mitigating some of the effects of maltreatment by providing a positive, potentially corrective, relationship experience (Ahrens et al., 2008; Munson & McMillen, 2009; Taussig & Weiler, 2017). Past relational distress and abandonment experiences for children in foster care, however, may pose a significant challenge for the development of a mentoring relationship built on trust and mutuality (Spencer et al., 2010). Indeed, some youth appear not to benefit from being assigned a mentor while others may experience early termination of the mentoring relationship that could lead to iatrogenic effects (Blakeslee et al., 2020; Grossman & Rhodes, 2002; Grossman et al., 2011; Kupersmidt et al., 2017; Stelter et al., 2018). The variability in program impact may partially be a function of youths’ interpersonal experiences prior to the intervention. The purpose of this study is to test the impact of youth mentoring among children with varied adult-youth relational histories who were maltreated and subsequently placed in foster care.

Literature Review

Effectiveness of Mentoring for Children in Foster Care

Formal mentoring programs for youth placed in foster care are growing due to the belief that a relational bond formed with a mentor can make a significant difference in the lives of youth who have experienced maltreatment. Theoretically, by providing consistent support within a nurturing environment, a quality relationship with a mentor may improve children’s perceptions of self and others and show youth that positive relationships with adults are achievable (Rhodes et al., 2006). Although there is an emphasis on expansion of mentoring programs for youth who were maltreated (Spencer et al., 2010), evaluations of such programs are sparse. Available research suggests that mentoring for youth in foster care can have positive impacts on mental health, self-esteem, peer relationships, educational outcomes and life satisfaction (e.g., Osterling & Hines, 2006; Rhodes et al., 1999; Taussig & Weiler, 2017; Taussig et al., 2019), but not all benefit equally (Blakeslee et al., 2020; Weiler & Taussig, 2019). Moreover, in long-term, open-ended mentoring programs, youth in foster care are likely to experience premature endings (Grossman & Rhodes, 2002; Stelter et al., 2018). While the evidence is growing, questions about whether, and for whom, mentoring is most effective remain.

Fostering Healthy Futures (FHF), a preventive intervention for preadolescents recently placed in foster care due to maltreatment, is one of a few interventions to examine the efficacy of a mentoring-based intervention specifically designed for children in out-of-home care due to maltreatment (see Taussig et al., 2007 for program description). Results of two randomized controlled trials of FHF found that children randomized to the intervention demonstrated fewer mental health problems, fewer trauma symptoms, and better quality of life (Taussig & Culhane, 2010; Taussig et al., 2019). Research examining moderating effects of FHF effectiveness, however, found that children exposed to relatively fewer adverse childhood experiences (ACEs; measured via a project-designed assessment) seem to benefit more from FHF with regard to trauma symptoms and quality of life than youth with greater ACEs exposure (Taussig et al., 2019; Weiler & Taussig, 2019). Although understanding how cumulative risk affects mentoring program efficacy is important, recent research has also emphasized the need to distinguish between different types of childhood adversity. For example, research indicates that ACEs related to childhood maltreatment and community violence are more likely to predict mental health challenges in adulthood (Lee et al., 2020).

Relational Histories as Moderator

Because mentoring-based interventions are primarily relational in nature, children’s disruptive or problematic relational histories may be especially likely to influence program impact. The ability to form and maintain healthy relationships, which could be referred to as relational capacity, is informed not only by genetic factors and temperamental traits (Donnellan et al., 2008; Picardi et al., 2011), but also through experiences with caregivers and other adults (i.e., attachment bonds; Bowlby 1969, 1988). The interplay of genetic, environmental and epigenetic processes on children’s relational capacity is complex and raises challenges to studying moderators of relationally-oriented interventions. Exposure to early stress in the environment (e.g., poverty) and within the family (e.g., maltreatment, witnessing violence) have direct and indirect effects on children’s attachment, relationship expectations and capacities, and development (e.g., Brown et al., 2020; Chisholm et al., 2005; Cicchetti et al., 2011; Hertzman & Boyce, 2010; Luijk et al., 2011; McCrory et al., 2010). While important, genetic variation appears less significant in the etiology of disorganized attachment, a form of insecure attachment that children who have been maltreated are especially at risk for (Cyr et al., 2010), than are the anomalous and pathological aspects of maltreatment while parenting (Cicchetti et al., 2011; p. 369; Luijk et al., 2011; Toth & Manly, 2019). For children who have been maltreated and subsequently placed in out-of-home care, their experiences with caregivers is critical; such is the focus of this study.

Consistent, warm, and nurturing caregiving relationships can influence children’s healthy expectancies from others (Ainsworth, 1989; Bowlby 1969, 1988). In addition to the foundation of trust, optimal quality in early caregiving may also help socialize children in forming new relationships and maintaining existing ones (Allen et al., 2007; Carlson et al., 2004). For instance, in one study examining adolescents’ attachment style and decision to enter mentoring relationships, adolescents with secure attachment were more likely to have a mentor (Georgiou et al., 2008). In another study of a national, community-based mentoring program in Israel, children with secure attachment histories benefited more from a positive mentoring relationship on global self-concept and reduced loneliness than those with insecure attachment histories (Goldner & Scharf, 2014).

In contrast, disruptions in early caregiving relationships due to death, maltreatment, and neglect may negatively influence the development of relational capacity in children (Cyr et al., 2010; Guild et al., 2017; Kim & Cicchetti, 2010; Milan & Pinderhughes, 2000; Moutsiana et al., 2014). In the context of maltreatment, children are more likely to hold negative representations of parents, self, and the parent-child relationship as compared to children who have not experienced maltreatment (Toth et al., 2000). Insecure and disorganized attachment patterns may initiate a maladaptive trajectory that places children who have been maltreated at greater risk for relational challenges and psychopathology (Barnett et al., 1999; Hesse & Main, 2006; Pickreign Stronach et al., 2011). For instance, maltreatment experiences can affect emotion dysregulation and social relations, such that higher dysregulation is associated with externalizing problems, peer rejection, and peer victimization (Cicchetti & Toth, 2016; Kim & Cicchetti, 2010; Shields & Cicchetti, 2001). Youth with unhealthy relationships with caregivers and other adults may develop a maladaptive, albeit self-protective (e.g., not easily trusting a new person, unconsciously pushing the adult away to avoid feeling hurt), approach to relationships, which may influence the degree to which they can engage with and benefit from relational interventions. At the same time, relational interventions may be the necessary anecdote for the adverse sequelae of child maltreatment (Toth & Manly, 2019).

Recent research has started to illuminate how mentees’ perceptions of relationships in their lives affect the quality of the mentoring relationship, although the implications are not totally clear. In one study of children involved in a school-based mentoring program, findings indicated that mentoring effects were dependent on youths’ pre-program relationships with parents, peers, and teachers. Schwartz and colleagues (2011) found that the intervention was most effective for children with adequate relationships as compared to those categorized as relationally strong or relationally weak. It is likely that most youth in the foster care system bring with them substantial relational risk, suggesting the potential benefits may be limited. In an exploratory study of a time-limited mentoring program for adolescent girls (not in foster care), girls who had poorer trust and communication with their mothers at the start of the program experienced less satisfaction with their mentoring relationship. On the other hand, girls who felt more alienated by their mothers before the program started reported higher quality mentoring relationships (Williamson et al., 2019). The seemingly contradictory findings may be explained, in part, by the conceptualization of these constructs and the developmental stage of the mentees. The authors suggest that girls’ trust and communication with their mothers (e.g., “She accepts me as I am”) may reflect larger attachment patterns, thereby influencing their working models and proclivity to experience a satisfying mentoring relationship. Alienation (e.g., I feel angry with her), though, may reflect a developmentally-expected shift away from parents during adolescence and not an attachment pattern that may pose a challenge to forming a mentoring relationship.

In addition to caregiver relationship quality, caregiver instability is a key component of children’s relational histories. Instability can include factors such as inconsistency in meeting the child’s needs, unavailability (low or sporadic frequency of physical proximity), and impermanence. Caregiver instability is associated with poor physical and mental health (Aarons et al., 2010; Bederian-Gardner et al., 2018; Newton et al., 2000; Villodas et al., 2016) and can be considered traumatic in its own right (Casanueva et al., 2014). Unfortunately, the more placement changes a child in foster care experiences, the more likely they are to have future changes (Osterman et al., 2007). Caregiver instability, measured by the number of caregiver transitions, was included as one indicator of risk in previous research examining the moderating effect of ACEs on FHF efficacy (Taussig et al., 2019; Weiler & Taussig, 2019), but it has not been examined as an individual risk factor within FHF research. Caregiver instability was assessed in another study examining the impact of a mentoring-based intervention for older youth preparing to transition from foster care. In that study, results indicated that youth with greater placement instability pre-program, as well as high placement restriction (i.e., restriction on communication, ability to move freely within home, and community participation) and traumatic stress, showed fewer overall benefits of self-determination and self-efficacy compared to youth with low-to-average risk (Blakeslee et al., 2020).

Finally, previous mentoring experiences may moderate the effect of a mentoring-based intervention on children’s mental health and related outcomes. For children in foster care with previous mentoring experiences, one hypothesis is that they may be in a better position to benefit. Certainly the larger relational context beyond caregiving matters for youth mentoring (Keller & Blakeslee, 2014; Schwartz & Rhodes, 2016). According to attachment theory, positive relational experiences with other caring adults (e.g., teachers, caseworkers) could contribute to the child’s future relational capacity. Yet, negative relational experiences with other adults that youth bring with them into mentoring relationships may understandably present barriers to building positive relationships with new adults (e.g., skepticism, difficulty trusting), potentially preventing them from deriving benefits from mentoring-based interventions. Conversely, these negative relational experiences may also present an opportunity for restoration or healing (Britner et al., 2013; Spencer et al., 2010). Children’s previous mentoring relationships, therefore, may perpetuate challenges or instill hope within youth. Because too few youth have mentors at all (e.g., Ahrens et al., 2008; Bruce & Bridgeland, 2014; Greeson et al., 2016; Munson & McMillen, 2008), an important initial question is whether any prior experience with a mentor moderates program efficacy.

The Current Study

Children’s interpersonal histories are clearly complex. For children in foster care, their relational histories may simultaneously offer opportunity and challenge as they enter into a new mentoring relationship. In an effort to further delineate the factors that impact mentoring effectiveness, the current study examined whether children’s relational histories (i.e., birth parent relationship quality, foster parent relationship quality, caregiver instability, and previous mentoring experience, all measured prior to the start of the intervention) moderated the impact of FHF on children’s mental health, trauma symptoms, and quality of life. Foster parent refers to any type of out-of-home caregiver, including relative and non-relative providers who cared for the child while in out-of-home care. We hypothesized that children with better baseline caregiving relationships, those with less baseline caregiver instability, and those who had a previous mentoring experience would exhibit fewer mental health problems post-intervention than those who had more negative or disruptive relational histories.

Method

Participants included children (ages 9-11) who participated in the Fostering Healthy Futures (FHF) randomized control trial (Taussig & Culhane, 2010; Taussig et al., 2019). The recruitment criteria were: (1) placement in any type of foster care by court order due to maltreatment within the preceding year; (2) current residence in foster care within a 35-min drive to intervention sites; (3) placement with current caregiver for at least 3 weeks; and (4) English proficiency. As the CONSORT diagram in Taussig et al. (2019) describes, 90.1% percent of eligible children (N = 511) and their caregivers agreed to participate; 16.6% were deemed ineligible after completing the baseline interview, resulting in a sample of 426 children.

Participants

Of the 426 participating, 233 were randomized to the intervention and 193 to the control condition. Forty-six (20 intervention and 26 control, a non-significant difference) were lost to follow-up at the 6-month post-intervention interview. The sample was 51.9% male, with a mean age of 10.28 years (SD = .90). The racial/ethnic distribution (nonexclusive) was 50.6% White, 51.5% Hispanic, 28.4% African American, 7.6% Native American, 1.4% Asian American, 1.4% Pacific Islander, and 0.7% Other. Children experienced the following maltreatment types (nonexclusive): 83.3% supervisory neglect, 63.1% emotional maltreatment, 48.4% physical neglect, 27.0% physical abuse, and 11.0% sexual abuse.

Procedures

A University Institutional Review Board (IRB) approved the study protocol, and informed consent and assent were obtained prior to children and their caregivers being interviewed. Following the baseline interview, eligible children were randomized to the intervention and control groups after stratifying on gender and county. Children randomized to both conditions received a comprehensive screening assessment, which was summarized and provide to children’s caseworkers, and participated in child welfare services as usual. Children randomized to the intervention were offered the opportunity to participate in FHF. The FHF intervention was intended to be above and beyond the services children typically received.

The current study utilized data collected at baseline (T1; 2-3 months prior to the start of the intervention) and at the 6-month post-intervention time point (T2; 17-20 months post-baseline). Children and their current caregivers were interviewed by separate interviewers and paid $40 for each interview. Children’s teachers were surveyed at T2 and they were paid $25 to complete the survey.

Intervention

The 9-month (30-week) FHF program is a preventive intervention that is theoretically grounded in positive youth development and comprised of two components: (1) weekly one-on-one mentoring; and (2) manualized skills groups for children in foster care (Taussig et al., 2007; Taussig et al., 2015). Mentors were graduate students in social work or psychology who received course credit for their work with children in the FHF program. With the support of regular training and supervision, mentors worked to (1) create empowering relationships, (2) ensure children received appropriate services, (3) help children practice social skills, (4) engage children in extracurricular activities, and (5) foster a positive future orientation (Taussig et al., 2007). The weekly 2-3 hour mentoring visits were tailored to children’s strengths and interests and sought to help children generalize the skills they were learning in their weekly groups. The FHF skills group follows a manualized curriculum that combines cognitive-behavioral skills group activities with process-oriented material on topics such as emotion recognition, change and loss, peer pressure, problem solving, and future orientation (Taussig et al., 2007). Children attended an average of 25.6 of 30 skills groups (median = 28.0, SD = 6.0, range of 1–30) and 25.9 of 30 mentoring visits (median = 27.0, SD = 6.4, range of 2–41).

Measurement

Demographic Variables.

Child age, sex (0 = female, 1 = male), and race/ethnicity (0 = White, 1 = Non-White) were collected via self- and caregiver-report.

Dependent Variables.

Dependent variables were identical to the primary outcomes measured in the FHF efficacy trials and are described as such (Taussig & Culhane, 2010; Taussig et al., 2019). A multi-informant index of children’s mental health functioning was created based on principal components factor analysis of the following variables: (1) self-reported mean scores on the Trauma Symptom Checklist for Children (TSCC; Briere, 1996) at T1 and T2; (2) the Internalizing scale of the Child Behavior Checklist (CBCL; Achenbach & Rescorla, 2001) completed by children’s caregivers at T1 and T2; and (3) the Internalizing Scale of the Teacher Report Form at T2 (Achenbach & Rescorla, 2001). A factor score was created and used in analyses. The factor score accounted for 62.8% of the variance at T1, with factor loadings at .79. At T2, the factor score accounted for 44.8% of the variance with factor loadings ranged from .62 - .71. Higher scores indicate greater mental health challenges. Two subscales of the children’s self-reported TSCC were used to assess Dissociation and Posttraumatic Stress. Dissociative symptoms include de-realization, emotional numbing, and memory problems, whereas post-traumatic stress symptoms include intrusive thoughts, nightmares, and avoidance of stimuli. Both subscales derive a standardized T-score. Higher scores indicate greater trauma symptoms. Finally, Quality of Life was measured by asking respondents to rate their satisfaction in several domains of their life (e.g., school, home, health, friendships; Andrews & Withey, 1976). The scale consists of 11 items that are rated from 1 (Mostly Unhappy) to 3 (Mostly Happy). The scale has good internal consistency (T1: α = .75; T2: α = .77); for analysis, the mean score was calculated. Higher scores indicated greater life satisfaction.

Moderators.

At T1, children’s relationship quality with birth parents and relationship quality with foster parents were measured with the Inventory of Parent and Peer Attachment — Short Form (IPPA-Short Form; Gifford-Smith, 2000). On a 3-point scale from 1 (Not at all True) to 3 (Often True), children were asked to rate the extent to which they thought each item was true about their birth parent and foster parent (e.g., “My parent cares about me.”). The scale includes 15 items comprising three subscales: (1) acceptance, (2) positive communication, and (3) negative representation, which was reverse coded to show positive representation. We created separate latent variables for each moderator; baseline relationship quality with birth parents (α = .92) and baseline relationship quality with foster parents (α = .83).

Baseline caregiver instability was determined through a protocol developed during the randomized controlled trial of FHF. Children were asked to complete a timeline of caregiver changes starting with who they lived with at birth. They were then asked who they lived with next, the relationship of that person, type of living situation (e.g. in foster care, reunified with parents), how old they were when they changed caregivers, and the reason for the change. Each child could list up to 12 caregiver changes. For this study, baseline caregiver instability refers to the number of caregiver transitions from birth to T1 (i.e., subtracting one from the total number of reported caregivers).

Prior mentoring experience was measured at T1 by asking children whether they recently had a non-parental adult that they could turn to for advice or support at the baseline assessment (“Over the past year, were there any adults who you could turn to for advice and support, such as teachers, coaches, big brothers/sisters, mentors, counselors, caseworkers, people at church?”). Prior mentoring experience was coded as 0 = no and 1 = yes. In the sample, a total of 248 youth (58.2%) reported they had mentoring in the year preceding their baseline assessment.

Data Analysis Plan

Prior to hypothesis testing, we completed preliminary analyses. We first examined descriptive statistics, correlation matrices of study variables, and missing data. Second, we used confirmatory factor analysis (CFA) to test the measurement model of the two latent constructs (i.e., relationship quality with birth parent; relationship quality with foster parent). Three indicators were used for each construct: acceptance, positive communication, and reversed negative representation.

To test hypothesized moderation of the intent-to-treat (ITT) effect of FHF intervention on outcomes (i.e., overall mental health, dissociation, posttraumatic stress, and quality of life), we used different approaches to accommodate each of the three unique variable types (i.e., relationship quality as a latent construct, instability as a continuous variable, and prior mentoring experience as a dichotomous variable). First, to test the moderating effect of relationship quality, latent moderated structural equations (LMS; Klein & Moosbrugger, 2000) were used because it assumes the nonnormality of the latent product terms when creating the moderator variable. Second, to examine the moderating effect of caregiver instability, we tested the effect of an interaction term (mean-centered caregiver instability X intervention status) on each outcome. Third, to test the moderating effect of prior mentoring experience, we conducted multigroup analysis to compare the intervention effect between the two groups (prior mentoring v. no prior mentoring) by constraining residual variance. Before completing the multigroup analysis, we tested independent model fit for those with and without prior mentoring experience in all four outcome models.

This study investigated all hypothesized moderation models using Mplus 7.2 (Muthén & Muthén, 1998–2012). In all models, treatment status (1= intervention, 0= control) was the independent variable, and we controlled for the corresponding variables measured at baseline, as well as child’s age, gender, and race/ethnicity. Model fit was evaluated using the fit indices recommended by Hu and Bentler (1999): comparative fit index (CFI) above .95, standardized root mean squared residual (SRMR) below .08, and root mean squared error of approximation (RMSEA) below .06. In the LMS analysis using XWITH statement, these fit indices are not available, so model fit was assessed using a log-likelihood (LL) ratio test, Akaike information criterion (AIC) and Bayesian information criterion (BIC) to determine the relative fit of alternative model (including interaction term) with null model (parsimony model with constrained interaction term) (Klein & Moosbrugger, 2000; Muthén, 2012). Significant loss of LL ratio and lower AIC and BIC indicates better fit (Santorra & Bentler, 2010). Finally, we determined the region of significance to identify the points at which the interaction effect is significant.

Results

Preliminary Analyses

Table 1 shows the descriptive statistics and bivariate correlations among study variables at baseline. As expected, the three indicators of relationship quality (acceptance, communication, and representation) were positively correlated with one another for birth parents and foster parents, respectively. The outcome variables were also correlated as expected (i.e., posttraumatic stress, dissociation, and mental health functioning were positively correlated, and quality of life was negatively correlated with the others). In regards to the relationships between the moderator variables and the outcome variables, each relationship quality indicator (with both birth and foster parents) was significantly correlated with children’s quality of life, posttraumatic stress symptoms, dissociation, and mental health functioning in the expected direction. Caregiver instability was negatively related to foster parent communication and positively correlated with mental health challenges. Finally, prior mentoring experience was not correlated with any other variables.

Table 1.

Correlation, Means, and Standard Deviations of Study Variables (N = 426)

	1	2	3	4	5	6	7	8	9	10	11	12
1. Previous mentor	1
2. CG instability	0.09	1
3. BP acceptance	0.01	−0.03	1
4. BP communication	0.05	0.02	0.74**	1
5. BP representation	0.02	−0.04	0.73**	0.54**	1
6. FP acceptance	−0.01	−0.06	0.10*	0.06	0.08	1
7. FP communication	0.03	−0.11*	0.02	0.11*	0.02	0.60**	1
8. FP representation	−0.04	−0.04	0.07	−0.02	0.19**	0.56**	0.30**	1
9. T1 MH	0.05	0.11*	−0.22**	−0.11*	−0.28**	−0.28**	−0.13**	−0.40**	1
10. T1 PTS	0.05	0.06	−0.22**	−0.12*	−0.29**	−0.25**	−0.14**	−0.34**	0.70**	1
11. T1 DIS	0.05	0.04	−0.20**	−0.10*	−0.24**	−0.17**	−0.10*	−0.34**	0.69**	0.73**	1
12. T1 quality of life	0.05	−0.03	0.22**	0.20**	0.16**	0.32**	0.29**	0.26**	−0.28**	−0.27**	−0.22**	1
Mean	0.58	2.78	2.65	2.44	2.48	2.73	2.34	2.51	0.00	49.44	50.23	2.68
SD	0.49	2.08	0.47	0.60	0.54	0.34	0.49	0.46	1.00	10.40	10.78	0.29

Note.

^*p<.05,

^**p<.001;

BP = Birth parent, FP = Foster parent, MH = Mental health problems, PTS = Posttraumatic stress, DIS = Dissociation.

The percentage of missing data on all study variables ranged from 0% to 18.8%. The Little’s Missing Completely at Random (MCAR) test indicated no systematic relationship between the missing data and the study variables (χ²(153) = 146.522, p = .63). Missing data were treated with full information maximum likelihood (FIML), a well-known estimator to minimize loss of information (Schafer & Graham, 2002) and efficient in model estimation (Enders & Bandalos, 2001).

Moderation Effect of Relationship Quality

We first conducted a CFA to examine the model fit of the two latent constructs of relationship quality. All factor loadings were significantly related to the relationship quality construct (p < 0.001) and the model exhibited adequate fit: χ²(5) = 4.369, p>.05, CFI = 1.00, SRMR = 0.015, and RMSEA = 0.00. The standardized factor loadings ranged from .80 to .93 (acceptance = .93, communication = .80, representation = .93) for birth parent relationship quality and from .47 to .92 (acceptance = .92, communication = .65, representation = .47) for foster parent relationship quality.

Next we tested the moderating effects. In the LMS with foster parent relationship quality as the moderator, the LL test showed no significant fit difference between the null model and alternative model across the four outcome models: (1) Mental health functioning, Δχ²(1) = 0.55, p >.05, (2) Dissociation, Δχ²(1) = 0.05, p >.05, (3) Posttraumatic stress, Δχ²(1) = 0.60, p >.05, and (4) Quality of life, Δχ²(1) = 0.30, p >.05. The AIC and BIC values overall were slightly higher in the interaction model than the nested model. Therefore, the model without the interaction term provided better fit compared to the model with the latent interaction.

Using the same method, in the birth parent relationship quality LMS, the LL test showed no significant fit differences between the null model and alternative model across the outcome models with the exception of the model with quality of life as the outcome: (1) Mental health functioning, Δχ²(1) = 0.01, p >.05, (2) Dissociation, Δχ²(1) = 0.05, p >.05, (3) Posttraumatic stress, Δχ²(1) = 0.12, p >.05, and (4) Quality of life, Δχ²(1) = 147.3, p <.05. In general, the AIC and BIC values were slightly lower in the interaction models in the first three outcome models, but those values significantly decreased in the quality of life outcome model. The model that included the latent interaction provided better fit than the model with no latent interaction. The moderation effect and its region of significance indicated the impact of FHF on quality of life was greater for youth reporting lower relationship quality with birth parents (see Figure 1).

Figure 1.

Region of significance of moderation effect of relationship quality with birth parents on youth life quality (QL).

Moderation Effect of Caregiver Instability

Next, we tested the moderating effect of caregiver instability on each of the four outcomes. Caregiver instability did not moderate the intervention effect on mental health functioning (B = 0.065, p = .168), but it did moderate the intervention effect on dissociation (B = 0.831, p < .05), posttraumatic stress (B = 0.842, p < .05), and quality of life (B = −0.031, p < .05). The moderation effect and its region of significance (see Figure 2) indicated that there was a greater effect of the intervention on dissociation and posttraumatic stress symptoms for children with three or fewer caregiver transitions at baseline. Similarly, the intervention effect on quality of life appears stronger for children with two or fewer caregiver transitions at baseline.

Figure 2.

Region of significance of moderation effect of number of caregiver transitions on youth mental health outcomes. Note. DIS = dissociation, PTS = posttraumatic stress, QL = quality of life.

Moderation Effect of Prior Mentoring Experience

Before testing the moderating effect of prior mentoring experience, we examined model fit for each group (prior mentoring v. no prior mentoring) on the four outcomes, separately. All individual models showed model fit that met the Hu and Bentler (1999) criteria. Within the multigroup analyses, the mental health functioning multigroup model showed a good fit (CFI =1.00, SRMR=.021, RMSEA=.00), while the posttraumatic stress and quality of life multigroup models showed moderate fit (CFI=.983, SRMR =.077, RMSEA=.089; CFI=.971, SRMR=.157, RMSEA=.072, respectively). The dissociation multigroup model showed inadequate fit (CFI=.934, SRMR=.137, RMSEA=.198). There were no significant differences in intervention effect between the two groups, except for a marginally significant difference in intervention effect of previous mentoring experience on posttraumatic stress (b = −3.06, SE = 1.775, [90% CI: −5.98, −0.14]), such that the intervention effect was higher for kids with previous mentoring.

Discussion

The results of this study indicated that, overall, FHF appears to have positive effects on youth in foster care who bring with them diverse relational histories. We hypothesized that youth would benefit more from the intervention if they had better relationships with caregivers, fewer caregiver disruptions pre-program, and prior mentoring experience, but, for the most part, this was not supported. Neither prior mentoring experience nor foster parent relationship quality significantly moderated the effects of the intervention on mental health, trauma symptoms or quality of life. Relationship quality with birth parents and caregiver instability moderated the impact of FHF on some outcomes. Specifically, the impact of FHF on quality of life was greater for youth reporting lower relationship quality with birth parents, and similarly, the impacts of FHF on quality of life, posttraumatic stress and dissociative symptoms were strongest for youth with relatively few caregiver transitions (about 3 or less).

The role that a youth’s history of caregiving transitions plays in a formal mentoring intervention effectiveness is logical and consistent with previous literature. Mentoring consists of bringing a new, temporary relationship with a caring adult into a young person’s life. Youth who have already had numerous transitions in caregiving relationships may be less willing to invest in a new relationship. For example, if they bring with them a model of relationships that includes abandonment by adults, this could influence how open they are to a mentoring relationship (Bowlby, 1988). Youth in foster care also are more likely to have had numerous professional helpers coming in and out of their lives, so a mentor may feel like one more temporary helper who will eventually leave them, although empirical research is needed to further investigate these potential processes.

Additionally, because prior transitions in caregiving among youth in foster care predict future transitions (Oosterman et al., 2007), it is likely that youth who experienced multiple transitions in caregiving prior to the intervention also experienced additional caregiver transitions during the intervention, which would likely influence the mentoring relationship both emotionally and logistically (e.g., moving may make it more challenging to meet regularly with a mentor). This is also consistent with some previous research that found in a community sample of youth who experienced more stressful life changes (e.g., moving, parental separation, loss of someone close, etc.) in the 6 months prior to the start of a formal mentoring intervention were more likely to be in relationships that terminated prematurely (Grossman et al., 2012). Within FHF, mentors persisted with youth during all caregiver transitions, including reunification and adoption, and prioritized this consistency for children, thereby limiting avoidable disruptions in the mentoring relationship and program. Future research should examine the extent to which pre-program instability versus during-program instability affects propensity to derive benefits from mentoring. In the same way, the context surrounding the transition (e.g., moving to a more restrictive placement, reunifying with birth parent(s)) as well as the extent and type of support provided during the transition could affect the youth’s experience of, and response to, the stressor. Digging deeper into theses nuances is another important future direction for research.

Results also showed that the intervention’s effect on quality of life was stronger for youth who had more negative relationships with their birth parents. This finding is inconsistent with attachment theory, which would suggest that those with more negative representations of their parents would have greater challenges establishing a relationship with a mentor, thereby limiting positive outcomes. Future research should examine the quality of the mentoring relationship to test whether children with thwarted relational capacities actually experienced lower quality mentoring relationship. If so, this finding highlights limits to attachment theory’s utility in explaining and predicting children’s capacity to form and benefit from non-familial mentoring relationships. If children’s negative relationships with caregivers do not prohibit the formation of high-quality mentoring relationships and children are able to derive benefits via the relationship, this is highly encouraging to the field of youth mentoring and would suggest greater investment in mentoring programs for children with maltreatment histories and/or poor parent-child relationships. It would also be important to consider other factors (e.g., child characteristics and personality, trauma symptoms) that contribute to mentoring relationship quality and the child’s capacity to form a relational bond with their mentor.

Additionally, one assumption is that a quality mentoring relationship precedes positive outcomes, yet benefits of mentoring programs may be derived through other processes (Cavell et al., 2021; Christensen et al., 2020; Rhodes, 2020). The fact that relationship quality with birth parents moderated the effect of FHF on quality of life and not the other outcomes (mental health, posttraumatic stress, and dissociation) raises questions about which aspects of the program affect quality of life. For example, engaging in fun activities with their mentor may be the respite children with challenging relationships with their birth parents need. In the same way, support derived within the skills groups could lead to feeling less stigmatized and greater satisfaction with peers (a component of quality of life). Perhaps the impact of the intervention on quality of life is not channeled through the mentoring relationship. Examining the relative impact of different aspects of the program on quality of life is an empirical question that should be explored in future research. Using optimization trial designs (e.g., micro-randomized trial, multiphase optimization strategy), for instance, would allow for comparison of different intervention components that contribute to quality of life, thereby informing developmental science and attachment literature.

It is also notable that, contrary to our hypothesis, prior mentoring experience did not predict whether youth benefited from FHF. This finding may speak to the FHF mentor’s capacity to form relationships with youth regardless of their previous experience with mentoring. At the same time, even among youth who have been mentored, there is significant variability across mentoring experiences; thus, it should not be assumed that prior mentoring experience means that a youth is more trusting of, or receptive to, subsequent mentoring. For example, a study examining potential interactions between the presence of a natural mentor and a history of childhood abuse did not detect any interaction or buffering effect of mentoring for youth with histories of abuse (Weber Ku et al., 2020). Other research on formal mentoring programs that examined whether having a “special” nonparental adult at baseline predicted better youth outcomes also failed to detect a significant effect (Herrera et al., 2013), suggesting that having had a mentor per se, may not significantly alter youth’s approach to relationships.

Implications

It is encouraging that FHF showed positive impacts on youth in foster care with diverse and challenging relational histories, contrary to hypotheses that FHF may primarily benefit those who already had more positive experiences with relationships. This finding may be a function of the unique aspects of FHF, including the use of graduate students in social work and related fields and combination with skills groups. In fact, in contrast to programs that do not target foster youth, mentoring interventions for youth in foster care tend to be multi-component (i.e., not only one-to-one mentoring) and/or use agency staff members or university students as mentors (Taussig & Weiler, 2017). Graduate student mentors in FHF receive ongoing supervision within an educational context which implies they have more opportunity to develop advanced skills than a typical mentor without specific training in the helping professions. For example, social work graduate students are not only supported by program staff who are licensed mental health professionals, but they are also engaged in coursework within their program of study that could supplement and enhance their understanding of child development, trauma, and clinical skills. As a result, FHF mentors may be skilled to work effectively with children with varied relational histories resulting in relatively consistent impacts. For stand-alone mentoring interventions serving youth in foster care, it may be helpful to examine the background, training, and support of mentors and consider the use of groups (e.g., skills-related, support). Although this study could not test the unique impact of groups, the use of groups may offer a necessary supplement to the one-to-one relationship. Additionally, the time-limited nature of the FHF program likely aids in promoting healthy good-byes and ensuring that premature terminations are very rare.

The fact that effects on quality of life were strongest for youth with more negative relationships with their birth parents also suggests that mentoring relationships may fulfill a need for positive relationships in their lives. If and when reunification with birth parents occurs, a well-trained and supported mentor may be able to serve as a significant source of support in addressing relationship difficulties with birth parents and in the transition to reunification. As discussed above, other individual and environmental factors and program components could also contribute to youth quality of life suggesting need for further study. In addition to the support mentors may provide during the reunification, regular engagement with peers in the skills group may also provide supportive stability, structure, and familiarity before, during and after the transition.

Findings also suggest that youth with more caregiver transitions may benefit less from the intervention on some outcomes, which has implications for mentoring programs. It goes without saying that trauma-informed training is helpful for most mentors, but is particularly important for mentors of youth in foster care. The traumatic stress that is instability has been conceptualized and described as a series of significant losses (e.g., loss of friends, loss of normalcy, loss of self-esteem), as well as emotional scars that influence the ability to trust people and build relationships (Unrau et al., 2008). Youth with many caregiver transitions prior to starting a mentoring program may be hesitant or resistant to mentoring sessions. At the same time, they may be open to the relationship but unsure how much to trust the mentor. Match support should be provided for mentors of children who present as wary about forming a new relationship. Programs may choose to include additional mentor training to address relationship challenges (e.g., frequently cancelled meetings, feeling overwhelmed with youths’ circumstances) that may inadvertently parallel youth’s previous relational histories and result in adverse outcomes for youth (Zilberstein & Spencer, 2014).

Limitations and Directions for Future Research

The study included several strengths including a rigorous design and low attrition. Despite the high recruitment rate of children in foster care, however, the study had limited power to detect small moderation effects. Further, in the current study, child-level factors, mentoring relationship quality and service utilization were unknown. Child-level factors, such as personality, temperament, and behavior, were not modeled in this study although they could contribute to children’s relational histories. It is possible that genetic, physiological, and/or behavioral differences that make children more or less sensitive to the foster care environment may affect the extent to which disruptions in caregiver relationships and transitions impact them (Belsky & van IJzendoorn, 2017; Ellis et al., 2011). The interaction between child characteristics and environmental fit may be an additional factor to consider (Siefer, 2000). Statistical models in which caregiver transitions and relationship serve as partial endogenous and exogenous influences in relation to these child-level factors could provide more insight. Additionally, it may be that children with the greatest mental health needs were receiving additional mental health services, thus disguising any differential impact of FHF on mental health functioning. It is also unclear to what extent observed benefits of FHF stem from the mentoring intervention versus the skills group, making it difficult to generalize results to mentoring-only interventions. Similarly, the mentoring program, drawing from a mentor pool of graduate students who received substantial training and rarely stopped mentoring before program completion, is not representative of community-based mentoring programs. Finally, because participants could not be randomly assigned to various levels of the moderator variables, causality cannot be assumed and results should be interpreted with this in mind.

In addition to what was described above, future empirical research should examine the relative impact of the unique aspects of the FHF program that are hypothesized to produce better mental health and related outcomes. For instance, using experimental designs, studies could compare intervention effects by mentor type (graduate students vs. community volunteers) or program structure (mentoring vs. mentoring plus skills groups; time-limited vs. open-ended). In addition, not all mentoring programs for youth in foster care target trauma symptoms and mental health. It may be unnecessary, for example, to use graduate students if targeting academic success or career exploration. Similarly, time-limited, instrumental relationships may be warranted for older foster youth about to transition out of foster care. A priority of future research must be identifying the key mechanisms that can meet the targeted outcomes within the target audience.

It also would be helpful for future research to explore how relationships and stressors outside of the mentor-mentee dyad influenced mentoring relationships within the intervention group, using both quantitative and qualitative methods. For example, with respect to the finding that youth with more caregiver transitions benefited less from the program, it would be helpful to understand the extent to which caregiver transitions were occurring during the intervention and whether this may have influenced mentoring relationship quality and program engagement. It would similarly be beneficial to explore strategies or approaches used by mentors that may be more or less effective for youth with different relationship histories. For instance, does mentor involvement with other caring adults in the child’s life make a difference for those with thwarted caregiving relationships? Finally, for youth with greater caregiver instability, identifying what is needed to derive benefit related to trauma symptoms remains an important area of study.

Conclusion

The use of mentoring-based interventions for children in foster care is a popular intervention strategy, but it has only limited empirical support. The FHF program is one of few evidence-based youth mentoring interventions for youth in foster care. The results of this study are encouraging; the FHF program seems to be positively impacting mental health-related outcomes for children with varied relational histories. Mentors and mentoring programs can play an important role in the lives of children in foster care.

Highlights:

1. Children’s early relational histories may affect the impact of mentoring programs.

2. Fostering Healthy Futures (FHF) improves mental health of children with varied relational histories.

3. FHF’s impact on trauma symptoms was stronger for children with fewer caregiver changes pre-program.

4. FHF’s impact on quality of life was also stronger for children with fewer caregiver changes pre-program.

5. FHF’s impact on quality of life was stronger for children with weaker birth parent relationships.