Fostering Healthy Futures for Preteens (FHF-P)

Study 1

Summary

Taussig and Culhane (2010) and Taussig et al. (2012) conducted a randomized controlled trial with 156 children aged 9 to 11 years who were placed in foster care. The study was conducted from July 2002 to November 2010. Assessments of psychological outcomes (e.g., mental health symptoms, life satisfaction) were conducted at baseline, postintervention (immediately after the 9-month program) and at six-month follow-up (six months after completing the intervention).

Taussig and Culhane (2010) and Taussig et al. (2012) found that the intervention group relative to the control group had significantly

• Higher self-reported quality of life (a risk & protective factor) at immediate posttest
• Fewer mental health symptoms at 6-month follow-up
• Fewer self-reported dissociative symptoms at 6-month follow-up
• Less mental health treatment at 6-month follow-up.

Evaluation Methodology

Recruitment: The study was conducted from July 2002 to November 2010 during which participants were recruited in 5 cohorts over 5 consecutive summers from a list of all children aged 9 to 11 years who were placed in foster care in participating counties. Criteria included (1) placement in foster care by court order due to maltreatment within the preceding year; (2) currently residing in foster care within a 35-minute drive to skills group sites; (3) living with their current caregiver for at least 3 weeks; and (4) demonstrating adequate proficiency in English. Participation was voluntary and could not be court ordered. All children who had closed cases at the start of the study (i.e., had been placed successfully resulting in case closure) were excluded from the study. Of the 197 children found initially eligible, 180 completed baseline assessments (8.6% refusal rate) and 24 were found to not meet eligibility criteria due to no longer being in foster care, being developmentally delayed or not having accurate information on welfare records. When multiple members of a sibling group were eligible, one sibling was randomly selected to participate in the RCT.

Assignment: Children were stratified by sex and county and then randomized to the control group (n=77) and intervention group (n=79).

Assessments/Attrition: Assessments were conducted at baseline, postintervention (immediately after the 9-month program) and at 6-month follow-up (6-months after completing the intervention). At postintervention, 88% of the control group and 95% of the intervention group completed assessments; at 6-month follow-up, 88% of the control group and 97% of the intervention group completed assessments.

Sample: The sample was 50.5% male, 43% White, 30% Black, 50% Latino with a mean IQ score of 96. Among the mothers of these children, 38.5% had a mental illness and 55% had a criminal history according to child welfare records.

Measures: Primary outcome measures included mental health functioning, life satisfaction and use of mental health treatment, which were gathered through separate interviews with children and their caregivers. Interviewers were blinded to the condition, however at times the participants spontaneously disclosed their condition. Teachers of participating children were also surveyed at 10 months after baseline and one year later.

Mental health functioning was assessed using the following: (1) child self-report on the posttraumatic stress and dissociation scales of the Trauma Symptom Checklist for Children, a widely used symptom-oriented measure of mental health problems; and (2) a multi-informant index of mental health problems. The mental health index was created based on principal components factor analysis of the children's mean scores on the Trauma Symptom Checklist for Children and the internalizing scales of the Child Behavior Checklist and the Teacher Report Form completed by children's caregivers and teachers. The Child Behavior Checklist and Teacher Report Form are well-normed measures of child emotional and behavior problems.

Children also completed the Life Satisfaction Survey, a quality-of-life measure that asks respondents to rate satisfaction in several different domains (e.g., school, home, health, friendships). Children's use of mental health services and psychotropic medications was assessed based on the following: (1) caregiver report of services and medications used within the past month; and (2) child report of services and medications used within the past 9 months at post-intervention and the past 6 months at 6-month follow-up.

Secondary outcomes measures included child-reports of positive coping, negative coping (both from the Coping Inventory), self-worth, social acceptance (using scales from the Self-Perception Profile for Children) and social support (using People in My Life - Short Form).

The study indicated that all measures and scales had been used in prior published work and reliability and validity had been established.

Additionally, child welfare records provided information on the number of placement changes, placement in residential treatment, and case closure (permanency).

Analysis: Linear regression was used to estimate effect sizes for continuous outcome variables and Poisson regression was used to estimate relative risks for dichotomous variables using baseline scores and variables that differed between groups at baseline as covariates. An additional analysis was conducted on placement and permanency outcomes with a subset of children who had open child welfare cases at the start of the study timeframe (n=54 for control and n=56 for intervention). An analysis was conducted on this subsample of children with open cases (n=110). The 110 children who remained in this analysis with open cases at baseline had any one of three placements - foster, kinship and residential. Data for this analysis were gathered at baseline and one-year postintervention or at case closure. An additional analysis was also conducted on the subsample of children (n=61) in foster placement with open cases at baseline (i.e., excluding kinship and residential placements).

Intent-to-Treat: The analysis was conducted on the intent-to-treat sample and included participants regardless of intervention dose received.

Outcomes

Implementation Fidelity: Children attended an average of 25 of the 30 skills groups and 26.7 of the 30 mentoring visits. Mentors received weekly individual and group supervision and attended one didactic seminar. Of the 108 discrete activities in the skills group sessions, 103.8 (96%) were completed.

Baseline Equivalence: Out of 31 variables tested for the whole sample, 3 significant differences and 1 marginally significant difference were found between intervention and control group children on demographic variables. Youths in the intervention group were more likely to have higher IQ scores (p=.04), to have been physically abused (p=.05), and to have mothers with criminal histories (p=.01). Further there was some evidence to suggest that intervention youth were more frequently exposed to illegal activity (p=.08). All 4 of these variables were used as covariates in the models. No differences were found for primary outcome variables.

Differential Attrition: The study found that rates of attrition did not differ by condition at either postintervention or 6-month follow-up but did differ by some measures. When comparing attriters to non-attriters on baseline and outcome measures, those not interviewed at both time-points had lower IQ scores and those not interviewed at 6-month follow-up scored higher on the mental health factor score.

Posttest: At immediate posttest, of the eight primary psychological outcome measures, only one was statistically significant - intervention participants scored higher on self-reported quality of life than control participants (d = 0.42). None of the secondary outcome measures were significantly different between groups.

At 6-month follow-up, three of the eight primary psychological outcome measures were significantly different between intervention and control groups. Intervention participants scored lower on multi-informant mental health symptoms (d = -0.51), lower on self-reported dissociative symptoms (d = -0.39) and were less likely to receive mental health treatment (relative risk = 0.75). None of the secondary outcome measures were significantly different between groups.

No outcome was significantly influenced by the intervention at both posttest and follow-up.

Long-Term: Long-term outcomes were not assessed on the whole sample. However, for the subgroup analysis, of the three placement/permanency outcome measures analyzed over the 18-month period for a subset of children with open cases at baseline, one was significantly different between groups with the intervention group less likely to be placed in a residential treatment facility. These children with open cases at baseline were placed in any one of three placements at baseline - foster, kinship, residential. When a subsample of children in foster care at baseline were analyzed (i.e., not including children in kinship and residential placements), all three outcomes (number of placements, residential placement, permanency) were significantly different between groups in the expected direction. In addition, a measure of baseline behavior problems was found to moderate the effect of the intervention on placement changes.

Study 2

Summary

Taussig et al. (2019, 2021) conducted a randomized controlled trial with 426 children aged 9 to 11 years who were placed in foster care. The study began in August 2002 in Denver, Colorado, and expanded to four metro area counties in 2007. Participants were randomized to intervention or control conditions. Outcomes measured from baseline through 12 years post-intervention included mental health problems, posttraumatic stress symptoms, and self-reported and official delinquency.

Taussig et al. (2019, 2021) found that, compared to the control group, participants in the intervention group showed reduced

• Mental health symptomatology, especially trauma symptoms (i.e., posttraumatic stress and dissociation) at 6-10 months post-intervention
• Mental health service utilization at 6-10 months post-intervention
• Self-reported and court-reported delinquency at long-term follow-up.

Evaluation Methodology

Design:

Recruitment: Participants were recruited in 10 cohorts over 10 consecutive summers (the first five cohorts comprised Study 1 and the second five, Study 2) from a list of all children, aged 9-11, who were placed in foster care in four metropolitan counties. Children were recruited if they (a) had been placed in any type of out-of-home care (e.g., foster care, kinship care, residential treatment) by court order due to maltreatment within the preceding year, (b) resided, at the time of recruitment, in out-of-home care within a 35-minute drive to skills groups sites, (c) had lived with their current caregiver for at least 3 weeks, (d) were not developmentally delayed, and (e) demonstrated adequate proficiency in English (although their caregivers could be monolingual Spanish speaking). In the first five cohorts, when multiple members of a sibling group were eligible, one sibling was randomly selected to participate in the RCT. For those recruited from the last five cohorts, eligible siblings were paired for randomization and both were included in the trial (there were 22 sibling pairs included in the study). Participation was voluntary and could not be court-ordered. Initially, 567 were recruited for the study, and 56 refused to participate while 85 did not meet inclusion criteria after a baseline interview.

Assignment: After stratifying by gender and county, 426 children were randomized to conditions, 233 to treatment and 193 to control. All children were manually randomized, by cohort, in a single block.

Assessments/Attrition: Assessments occurred at baseline (Time 1), six-month follow-up (Time 2), 1.5-2.5 year follow-up (Time 3), and long-term follow-up (Time 4). For the long-term follow-up, surveys were done at ages 18-22, which varied across participants in the time from intervention but lasted up to 12 years. Also for the long-term follow-up, court records were obtained for seven years and three months after the program start. Of the 426 randomized to condition, retention rates were 89% at Time 2, 84% at Time 3, and 92% at Time 4. Taussig et al. (2021) reported an analysis sample of 391 (92%) for the survey data and 425 (99.8%) for the records data. The authors added one other note about attrition for one time point: "Not shown are the exclusion of data from 18 participants (9 control and 9 intervention) subsequent to their participation in a booster mentoring program when they were between the ages of 14 and 17."

Sample: Half of the youth self-identified as Hispanic/Latinx, half as Caucasian, and over a quarter as Black/African American (racial/ethnic categories were not mutually exclusive). Average IQ was 4.5 points lower than the standardization sample's mean. According to child welfare records, two-thirds of the children's biological mothers had a substance use history, 60% had a history of criminal activity, and 43% had a history of mental illness.

Eleven percent of the children had documented sexual abuse, over a quarter had a history of physical abuse, almost two-thirds had experienced documented emotional abuse, and almost all had experienced some type of neglect. Children's families had an average of 4.6 referrals to social services before being removed from their homes. Over three-quarters of the youth had been in therapy, and one-fifth had been on psychotropic medication. At the time of the baseline interview, 42% were living in nonrelative foster care, 54% were placed with kinship care providers, and the remaining 4% were in some type of congregate care (i.e., group homes or residential treatment).

Measures:

Taussig et al. (2019) interviewed children and their current caregivers/parents separately, typically at the child's residence. Interviewers were blind to condition at the follow-up (Time 2) interview. Current teachers of participating children were also surveyed 10 months post-intervention (Time 2).

To determine type(s) of maltreatment experienced and maternal characteristics (e.g., substance use, criminal history), trained research assistants coded each child's legal petition and social history (child welfare records' narrative of the history and events preceding the legal filing that led to the child's removal from the home) using a modified version of the Maltreatment Classification System. The developers of the rating system report an overall kappa of .60 and adequate estimates of inter-rater agreement (.67-1.0). All records were consensus coded by at least two trained staff, and discrepancies were resolved through consultation with one of the senior investigators. Four maternal characteristics (substance use, criminal history, mental illness, and history of maltreatment) and seven types of child maltreatment (physical abuse, sexual abuse, emotional abuse, failure to provide, lack of supervision, and moral-legal maltreatment) were dichotomously coded as present or absent.

Mental health functioning was assessed using the Posttraumatic Stress and Dissociation scales of the child self-report Trauma Symptom Checklist for Children (TSCC; Briere, 1996), a widely used symptom-oriented measure of mental health problems, as well as the internalizing scales of the Child Behavior Checklist (CBCL) and the Teacher Report Form (TRF), both well-normed measures of child emotional and behavior problems. A multi-informant Mental Health Index was created based on principal components factor analysis of the children's mean TSCC scores, and the internalizing scales of the CBCL and TRF.

Children completed the Life Satisfaction Scale, which asks respondents to rate satisfaction in several different domains (e.g., school, home, health, friendships, leisure activity). The authors report that the original items demonstrate good internal consistency (a = .81) and construct validity. In the present study, internal consistency was similar, a = .75).

Children's use of mental health services and psychotropic medications were assessed based on child and caregiver reports at Time 1 and Time 2. Because of concerns related to foster parents not knowing the child's history of therapy and medication use, we asked the caregivers to report on current use at both time points, whereas children were asked to report on lifetime use at baseline and past 6-month use at T2. Mental health service and psychotropic medication use were dichotomized at T1 and T2, such that if either the child or caregiver reported use in the timeframe, they were coded as 1; if neither informant reported use, they were coded a 0.

Taussig et al. (2021) examined two types of measures, self-reported delinquency and court-reported delinquency charges. Self-reported delinquency was measured with 15 items from the Adolescent Risk Behavior Survey and consisted of separate scales for total, non-violent, and violent delinquency. The authors cited other studies that have shown adequate reliability and validity of the measures. Court-reported delinquency charges came from state-wide court records over the seven-year period beginning at three months after the intervention start. The offenses were classified as non-violent, violent, and total. A limitation of both these measures noted by the authors (p. 19) was the lack of information on diversion, probation, or imprisonment that might prevent subsequent offending.

Analysis: Taussig et al. (2019) used linear regression models to predict continuous outcome variables and logistic regression to examine odds ratios for dichotomous outcomes. All regression models adjusted for baseline scores on the corresponding outcome measure. Interaction analyses were conducted to examine whether a priori-selected demographic, adverse childhood experiences (ACEs), or baseline functioning variables moderated the impact of the intervention on outcomes. In order to examine the impact of non-independent data from siblings, analyses were run with the full sample and then replicated with a sample that contained only one randomly selected sibling. Because there were no differences in effect sizes in any of the analyses after dropping one sibling, the results reported were for the full sample.

Taussig et al. (2021) used multilevel models to examine the non-linear trajectories in individual delinquency across ages of 9-22 and the four waves of data. The models controlled for baseline delinquency as well as sex, race, county, adverse childhood experiences, and maternal criminal history (which differed between conditions at baseline). Given that the outcomes measured the percent of violent, non-violent, and total offenses committed by each participant, a Poisson distribution was used with the multilevel models. The inclusion of siblings could create a clustering problem, but tests with one randomly selected sibling did not indicate any significant differences.

Intent-to-Treat: Taussig et al. (2019) reported that analyses used the intent-to-treat sample, though no data were imputed. In response to a Blueprints request, the lead authors wrote in an email (December 11, 2020) that when using all cases with Full Information Maximum Likelihood Estimation, "all prior effects were sustained." Taussig et al. (2021) used maximum likelihood estimation to include 391 participants, even those "with some missing data."

Outcomes

Implementation Fidelity: Ninety-five percent of those assigned to treatment started the program, 92% completed it, and there was over 85% attendance.

Baseline Equivalence: Taussig et al. (2019) found that, out of 25 variables, including baseline outcomes (mental health index, PTSD symptoms, dissociation, quality of life, mental health therapy, and psychotropic medication) and child, maternal and child welfare characteristics, there was a significant baseline difference between groups on two. Intervention youth were more likely than control youth to have mothers with criminal histories, and intervention youth were more likely to have a history of receiving mental health therapy.

Taussig et al. (2021) found that, out of 28 variables, three differed significantly across conditions. Intervention youth were more likely to have mothers with a criminal history but also to have lower self-reported total and non-violent delinquency.

Differential Attrition: In Taussig et al. (2019), chi-square analyses suggested that the rate of attrition did not differ by treatment condition. Across treatment conditions, those interviewed at follow-up were compared with non-interviewed children on all baseline characteristics, and no significant differences were detected. In addition, interaction analyses were conducted to examine whether mental health or maternal characteristics moderated the impact of intervention status on attrition. Two of the five interaction analyses (mental health index and posttraumatic stress symptoms) were significant suggesting that the trial retained more intervention children with mental health problems and adverse maternal characteristics, whereas children with these characteristics in the control group were more likely to have attrited. Additionally, in response to a Blueprints request, the lead author sent results for complete tests of differential attrition (December 28, 2020). In 28 tests using the interaction of condition by each baseline measure to predict attrition, three interaction coefficients were significant.

Taussig et al. (2021) tested the interaction of each baseline measure by condition in predicting attrition. There was one significant interaction term in 28 tests that involved maternal criminal history such that intervention youth with a maternal criminal history were less likely to drop out.

6-10 Months Post-Intervention: Findings showed significant effects in favor of treatment for 4 of 6 outcomes. Results demonstrated that at 6-10 months post-intervention, compared to the control group, participants in the intervention group showed reduced mental health symptomatology (d = -.25), especially trauma symptoms i.e., posttraumatic stress (d = -.20) and dissociation (d = -.29), and mental health service utilization (OR = .62). Intervention and control groups did not differ on quality of life or psychotropic medication use.

In addition, 42 moderator analyses showed three significant findings. Lower levels of ACEs were associated with stronger treatment effects on youth-reported posttraumatic stress symptoms and quality of life. Among children exposed to relatively few or an average number of ACEs, intervention participants reported fewer symptoms of posttraumatic stress than the control group. In addition, among children exposed to few ACEs, intervention participants reported higher quality of life at 6-10 months post-intervention, as compared to children in the control group.

Long-Term: Taussig et al. (2021) examined the measures of self-reported delinquency from ages 10-22. Table 2 shows that the linear effect of age did not differ across conditions for any of the three measures. However, for total delinquency and non-violent delinquency, the squared effect of age differed significantly across conditions. Based on these results, Figure 2 depicts lower levels and quicker declines in the two delinquency measures for the intervention group than the control group. By age 18, total delinquency was 82% lower in the intervention than in the control group.

For the three measures of court-reported delinquency charges from ages 9-18, the models showed significant differences across conditions in the linear effect of age for total charges and violent charges and a significant difference across conditions in the squared effect of age for violent charges. Figure 3 depicts more charges for the intervention group at ages 9-13, fewer charges at ages 14-17, and more charges again at age 18. The outcomes in mid-adolescence were 15-30% lower for the intervention group but not consistent across all the ages.