Abstract
Limited research is available examining distal child welfare outcomes after participation in evidence-based parenting interventions. To address this gap, this study employed a multi-tiered analytic approach to examine child welfare outcomes after participation in Attachment and Biobehavioral Catch-up (ABC). Using propensity score analytic techniques to establish a matched comparison group, logistic regressions examined subsequent maltreatment reports and substantiation, and survival analyses observed time to and likelihood of reunification for children who received one of three ABC curriculums compared to comparison group children (child welfare services as usual). In total, 205 children were included in the impact analysis (n = 66 treatment; n = 139 comparison); the majority of the children were White (53.7%), non-Hispanic (84.4%), males (59.5%) with an average age of 6 months (M [SD] = .50 [1.0]). Over half (56.1%) of the study sample was in out-of-home placement; 23.5% of the removed children experienced reunification. No statistically significant group differences were observed on the likelihood of subsequent or substantiated maltreatment reports. All three ABC curriculums were associated with a statistically significant increased likelihood of reunification, when compared to their matched counterpart. Additional research is warranted, though results indicate ABC may be a promising intervention to help enhance the likelihood of reunification.
Introduction
According to the World Health Organization (2022), child maltreatment refers to the abuse and neglect of children under the age of 18, including all types of physical and/or emotional abuse, sexual abuse, neglect, and other exploitation that results in actual or potential harm to a child (World Health Organization, 2022). Prior literature has documented chronic child maltreatment can have profound and lasting impacts on a child’s development and overall quality of life (Anda et al., 2006; Shonkoff & Garner, 2012). Alarmingly, in 2019, close to four million reports of child maltreatment involving over seven million children were received by child protection reporting centers across the United States (U.S. Department of Health & Human Services, Administration for Children and Families, Administration on Children, Youth and Families, Children’s Bureau (2021, [USDHHS, 2021]). While the majority of reports result in unsubstantiated findings, close to 16% of these reports result in substantiated findings (USDHHS, 2021).
The outcomes for children who have experienced maltreatment and other related adversities can be far-reaching and devastating, affecting every aspect of a child’s life and well-being. These outcomes can manifest in various ways, such as: physical health consequences; emotional and psychological impact; cognitive and academic; behavioral issues; social and interpersonal challenges – among others. Because of the severe and dysregulating consequences of child maltreatment, it is imperative to continue to examine efficacious treatment and interventions for both maltreating caregivers and children experiencing maltreatment. This imperative is particularly salient for families already involved in the child welfare system to avoid chronic or repeated maltreatment, ensure safe reunification, and – ultimately – enhance the child’s safety and well-being.
Child maltreatment interventions provided to families already involved in the child welfare system primarily include caregiver or parenting based interventions (i.e., parent skills training or parent coaching). While the provision of parenting interventions seems to be the industry standard for child welfare agencies, much remains unknown regarding how these interventions truly impact child maltreatment outcomes. Rather, most evaluations of these types of interventions measure increases in pro-social parenting behaviors as indicated by psychometric measures, rather than true child maltreatment instances (Whitcombe-Dobbs & Tarren-Sweeney, 2019). This finding is supported by a recent systematic review conducted by Whitecombe-Dobbs & Tarren-Sweeney (2019), which identified nine studies that examined whether parenting interventions implemented with child welfare involved caregivers reduced further maltreatment. The review noted only four of the nine studies, which examined two different parenting interventions (e.g., Parent-Child Interaction Therapy and SafeCare) demonstrated statistically significant results between intervention groups in post-intervention child protective services reports, and none of the interventions have been found to reduce all types of child maltreatment.
Understanding the effectiveness of parenting interventions or training programs on child maltreatment outcomes is critical, as each year numerous caregivers will voluntarily (or via mandate) participate in such programs in hopes to successfully preserve or reunify their family unit (Barth et al., 2005). Unfortunately, much of the literature regarding the effectiveness of parenting interventions is mixed and a large gap remains regarding parenting interventions and reductions in subsequent maltreatment for child welfare involved children and caregivers (Whitecombe-Dobbs & Tarren-Sweeny, 2019). In an effort to synthesize literature on intervention effectiveness and establish an evidential threshold for interventions and services intended to enhance supports for vulnerable children and families and to prevent foster care placements, the Administration for Children and Families - as a component of the Family First Prevention Services Act of 2018 – established the Title IV-E Prevention Services Clearinghouse (Title IV-E PSC) which put forth rigorous evidential standards to assist in organizing and cataloging child and family interventions. Guided by the Title IV-E PSC standards (Wilson et al., 2019), the current study utilized a multi-tiered analytic approach to examine child welfare outcomes, specifically repeated instances of both maltreatment reports and substantiated maltreatment reports, and the likelihood of reunification after participation an evidence-based parenting intervention: Attachment and Biobehavioral Catch-up.
Attachment Biobehavioral Catch-up
The Attachment Biobehavioral Catch-up (ABC) intervention, currently a level-1 evidence-based practice according to the California Evidence-Based Clearinghouse for Child Welfare (California Evidence-Based Clearinghouse for Child Welfare, nd), is an intensive 10-session, manualized parenting intervention program (Dozier et al., 2008). ABC was initially designed for families with infants aged 6–24 months (referred to as ABC-Infant [ABC-I]). However, since origination, ABC has been modified and adapted for use during the prenatal period through early infancy (prenatal – 6 months, referred to as modified-ABC [m-ABC]), as well as for toddler/early childhood aged children 24–48 months (referred to as Toddler Childhood [ABC-T]). For the present study, all three versions of the ABC curriculum were administered (i.e., m-ABC, ABC-I, and ABC-T). The curriculum content and structure for ABC-I and ABC-T are nearly identical; curriculum variations and adaptations are present in the m-ABC curriculum.
The overarching goal of the ABC intervention is to enhance parental sensitivity and child behavioral and physiological self-regulation. ABC employs the use of trained ABC parent coaches (i.e., home visitors) who provide “in-the-moment” coaching comments to parents; these comments are meant to promote ABC’s target behaviors of providing nurturance and following the child’s lead. Trained ABC parent coaches also provide guidance to caregivers to increase parental sensitivity and reduce intrusive or frightening caregiver behaviors (Lind et al., 2020). All ABC sessions, regardless of age variant curriculum, are videotaped or recorded for scoring and playback to the families; the three core intervention targets are: (1) providing nurturance (nurturance); (2) following the child’s lead (sensitivity); and (3) avoiding frightening behavior (intrusiveness) (Dozier & Bernard, 2017). An in-depth summary of the ABC intervention is outside the scope of this article, however, for additional information on the ABC intervention and session topics please refer to Dozier and Bernard et al. (2019). For purposes of clarity, it is important to mention that since the implementation and evaluation of the study described in this manuscript, the ABC developers have modified and renamed two of the three curriculums examined: mABC is now referred to as ABC-Newborn; and ABC-T is now referred to as ABC-Early Childhood. Theoretical or conceptual differences between the old and new curriculums are outside the author group’s purview, and, therefore, this manuscript has maintained the original intervention names that were associated with the implementation and evaluation. Any readers interested in implementing the ABC intervention should analyze potential differences in nomenclature and reporting criteria accordingly.
Research into the ABC-I model is robust. Results from numerous randomized control trials demonstrate the ABC-infant model is effective at increasing secure and organized attachment (Bernard et al., 2012), regulating cortisol production patterns, increasing child perceptive language skills, improving infant cognitive flexibility, and improving internalizing and externalizing problem behaviors (Grube & Liming, 2018). While ABC-I is a heavily researched intervention—and an intervention that has been used with at-risk and child welfare-involved families—there remains a gap in the ABC literature as it relates to distal child welfare outcomes and the impact ABC-I – along with the other ABC curriculums –has on subsequent maltreatment outcomes and reunification.
Important for all research examining parenting intervention effects on child-level outcomes is the articulation of the conceptual mechanism of effects – especially in the context of child welfare. Prior literature has suggested three classifications of factors impacting reunification—child characteristics, family characteristics, and child welfare service characteristics (Teixeira et al., 2022). Of particular relevance to this study are child characteristics. Research indicates that problematic behaviors exhibited by the child can adversely affect reunification (Davidson et al., 2019; Parolini et al., 2018). A central objective of the ABC intervention is to enhance the target child’s self-regulation abilities. Though there is very limited evidence examining a young, substance-affected child welfare involved population such as the present study, this study is underpinned by the concept that improved child self-regulation through the ABC intervention may have positive implications on reunification via improvements in the child’s emotional regulation capabilities –regardless of participating caregiver (Labella et al., 2020; Dozier et al., 2011).
Advancing the field and adding to the knowledge base on the ABC intervention, this study rigorously examined, by age-variant of the ABC intervention, distal child welfare outcomes, including: repeat maltreatment reports, repeat substantiated maltreatment reports, likelihood of reunification, and time to reunification. The researchers utilized a multi-tiered analytic process and hypothesized that treatment group children would be: less likely to have repeat maltreatment reports; less likely to have repeat substantiated maltreatment reports; more likely to experience reunification, if removed; and experience shorter days in out-of-home care, if removed, when compared to their propensity-score matched, child welfare counterpart (treatment as usual [TAU]).
Methods
Study Design & Eligibility
To examine these hypotheses, the researchers conducted a quasi-experimental, longitudinal design using propensity score matching (PSM) techniques to establish an equal comparison group. The present study is a sub-study of a larger, federal Regional Partnership Grant (RPG), funded by the Administration for Children and Families, Children’s Bureau, targeting young (0–48 months), substance-affected children and their families involved in the child welfare system. Enrollment for the parent study began in December 2019 and concluded in April 2022.
Participants
The parent study was implemented at one urban site in a Midwestern state. To be eligible for the parent study, the focal child must have been: (1) at-risk of removal or already removed; (2) have a case plan goal of reunification at the time of referral receipt; (3) have an open child welfare case; (4) substance-affected; and (5) parent must not have been incarcerated or have a “no contact” order in place. The present study included children from the parent study who had matched state administrative data and met study criteria.
Procedures, Data Matching & Analyses
The parent study was approved by the BLINDED & BLINDED institutional review boards. None of the authors or study researchers are affiliated with the ABC intervention or program developer and have no financial or other interest to report on the use and evaluation of ABC. The present study utilized state administrative data to examine the impact of the ABC intervention in multiple ways.
All variables utilized in this study were obtained or computed from the state’s child welfare administrative database. At the conclusion of the parent study (final treatment group case exited services in July 2022), the study authors requested child welfare state administrative data for all children with an “accepted” child welfare referral who met study eligibility criteria from October 1, 2017 (RPG start date) through September 1, 2022 (study end date) in order to accurately establish a matched comparison group. Specifically, records within the state administrative database were eligible for matching if they met the following criteria: (1) child had a birth date between the birthdate range of the treatment group children; (2) child lived in an eligible county; (3) referral must have been “substantiated”; and (4) substance abuse must have been a contributing factor to the respective substantiated referral. Screened out child welfare referrals were not received from the child welfare agency. For children with duplicate records, the authors selected the earliest eligible referral within the study timeframe; this date was used as the “referral date” for the comparison group. Using a unique family identifier provided in the state administrative database, the authors were able to easily identify sibling groups. In alignment with the parent-study selection criteria for treatment group children in multi-child households, if a referral contained multiple children/siblings, the study authors hand-cleaned the file to maintain the oldest eligible child in the sibling group. In the case of twins or triplets, the study authors randomly selected a child via a random number generator. Therefore, state administrative records for all children who met the above criteria, between October 1, 2017, and September 1, 2022 were eligible for matching. Child identifiers for treatment group children and their siblings were requested to ensure that treatment group children and/or children who may have been exposed to the ABC intervention (i.e., siblings identified at program enrollment) were not allocated to the pool eligible for matched comparison. These children were removed after cleaning and prior to conducting the matching procedure. In total, 6184 unique children remained in the eligible matching pool.
STATA/SE version 17.0 (StataCorp, 2021) was used to establish the propensity scored matched (PSM) comparison group from the pool of children who met the aforementioned study criteria. The advantage of utilizing PSM techniques, particularly when evaluating child welfare outcomes, allows researchers to estimate causal treatment effects using observational data while reducing selection bias in quasi-experimental designs (Barth & McCare, 2008; Gou et al., 2006). Broadly, PSM generates a statistical model for the treatment group selection process and then employs this model to match treatment cases with the most similar comparison cases within the eligible sample pool (Akin et al., 2017). A one-dimensional propensity score value is generated that statistically accounts for characteristics on each of the selected covariates for the respective case (Barth & McCare, 2008). In practice, cases are matched on their propensity score; treatment group cases match to their comparison group counterpart that shares key case and demographic characteristics known to influence outcomes.
For this study, the following covariates were used to generate a child’s propensity score: (1) child’s date of birth (e.g., MM/DD/YYYY); (2) date of referral (e.g., MM/DD/YYYY); (3) child’s race (5 dummy-coded race variables); (4) child’s biological sex (dummy-coded: female); (5) child’s ethnicity (dummy-coded: Hispanic); (6) substance abuse exposure type (2 dummy-coded variables); (5) domestic violence (DV) exposure (dummy-coded: DV exposed); (7) financial hardship exposure (dummy-coded: exposed to household financial hardship). Since the STATA/SE PSMATCH2 package does not allow for categorical variables, the authors had to recode child’s race into separate racial categories: White, Black, Asian, Pacific Islander, American Indian, and multi-racial (1 = identified as respective racial category); White served as the reference group and therefore was not needed in the matching procedures (due to multicollinearity). Similarly, substance-abuse type included exposure to: drug, alcohol, and both (1 = yes, exposed to respective substance-abuse category); substance abuse exposure to drugs served as the reference group and was not needed in matching procedures. Using the generated propensity scores, the authors conducted nearest neighbor matching using a 2:1 ratio (2 comparison group children: 1 treatment group child) and the field-standard caliper size recommended by Gou et al. (2006).
A well-documented critique of PSM is that an influential covariate may not be available in the dataset utilized; therefore, the study authors were limited in matching procedures by the variables available within the received state administrative dataset. Further, the authors conducted a power analysis in STATA/SE after generating the PSM comparison group to confirm that a 2:1 matching ratio would ensure adequate power for the respective analytic methods. Using Cohen’s effect size guidelines, power analyses revealed that a sample size of 205 and 115 with an alpha level of .05 would be adequately powered (β > .80) to detect medium effect sizes (d > .50) for both the main regression analyses and the supplemental survival analyses, respectively. Attachment and Bio-behavioral Catch-up (ABC).
As mentioned previously, the ABC intervention is a manualized, evidence-based parenting program comprised of ten, 60-min curriculum sessions, routinely delivered at weekly intervals by a trained ABC parent coach (Dozier & Bernard, 2019). Historically, the ABC intervention was intended to be delivered in the home of the target caregiver and young child, however, due to COVID-19 emergency orders during implementation of the parent study, the ABC intervention and study protocol was adapted by the intervention developers to be delivered via telehealth modalities in addition to in-person. Fidelity to the intervention was monitored – through an annual contract - via the developer’s research staff at the University of Delaware’s Developmental Psychology Lab (i.e., where the ABC intervention is housed). The ABC intervention, per intervention guidelines, was delivered to the target caregiver and one focal child. The target caregiver was defined as the caregiver who the child spends 50% or more of their time with; for the majority of the children included in this study, the target caregiver was the foster caregiver. Focal children who participated in at least one ABC session with an associated child welfare case were eligible for inclusion in the present study.
The parent study was implemented at a single urban, child and family services organization in a Midwestern state. Referrals to the parent study were received primarily from the state’s Department of Human Services (DHS), local substance abuse treatment provider, local mental health organization, and local child and family services organizations. A total of four trained ABC clinicians served all treatment group cases; clinicians were female (100%, N = 4) with 50% (n = (2) identifying as White, 25% (n = (1) identifying as Black/African American, and 25% (n = (1) identifying as multi-racial. All implementing clinicians had a master’s degree in social work or a related field, completed the intervention-required year of supervision, and received additional training in the mABC, ABC-T, and telehealth-adapted (referred to as TeleABC) models.
Pertinent to the respective parent study, the COVID-19 pandemic left little choice for many in-person child welfare services – particularly in-home based interventions – therefore, in order to provide continued, evidence-supported interventions to children and families in need, the majority of the field had to make an immediate shift—essentially overnight as the pandemic emergency orders were issued—in order to meet the needs of child welfare-involved families; this respective parent study was no different. The initial parent study began as a randomized control trial (RCT), however, adaptations to the intended study design were necessary to meet the needs of child welfare and the pandemic safety restrictions. Therefore, the evaluation design transitioned from an RCT to a QED, shortly after implementation, to allow for more families to receive services during a global crisis and to TeleABC.
It is important to note that recent research has demonstrated statistically significant positive parental outcomes when ABC was delivered via telehealth and in-person/telehealth hybrid format (Schein et al., 2023) providing evidence that the ABC intervention maintains effectiveness on the curriculum’s core components across delivery formats. Though additional research is needed to corroborate these findings, the early studies show promising results. Further, the ABC manualized session content was not altered based on administration format (i.e., in-person or telehealth). Due to multicollinearity with ABC curriculum variation and administration format, the present study was unable to include a covariate to account for administration modality; the majority of treatment cases completed teleABC (n = 52, 78.8%), with seven cases completing ABC in-person (10.6%) and six cases completing ABC in hybrid format (9.1%).
After removing duplicate cases (i.e., treatment group children who participated twice with different caregivers, n = 7), original control cases (i.e., children who were randomized into the control condition prior to COVID-19 evaluation design modifications), siblings of focal children, and treatment group children who did not receive any dosage of the ABC intervention (n = 5) from the sample, the analytic treatment group sample was 66. In total, 66 treatment group children were included in the main analysis; 47 treatment group children (71.2%) experienced removal and were included in the reunification supplemental analysis.
Treatment as Usual (TAU)
Comparison group children received child welfare services as usual through one urban, Midwestern county. Given the study design, detailed data on the comparison group (i.e., types of services received) were not available to the study authors. In total, there were 139 comparison group children included in the main analysis; 68 comparison group children (48.9%) experienced removal and were included in the reunification supplemental analysis. A CONSORT diagram of the participant flow is illustrated below, in Figure 1. CONSORT Diagram of Study Participants. Notes. Red outlined boxes indicate cases removed from analyses; Green outlined boxes indicate treatment group cases; Blue outlined boxes indicate cases from child welfare state administrative databases; Yellow outlined box indicates final study sample. * = Initial parent study design was an RCT; due to COVID-19, evaluation design was modified to a QED. ** = Associated referral date was prior to October 1, 2017 (study begin date). *** = Cases completed enrollment but did not attend any ABC sessions (i.e., no intervention exposure). ^ = Total, unduplicated child welfare referrals during study timeframe (Oct. 1, 2017 through Sept. 1, 2022).
Variables
Outcome (Dependent)
Multi-Tiered Analysis Plan Examining Child Welfare Outcomes: Research Questions, Sample Size and Analytic Method.
Notes. ABC = Attachment and Biobehavioral Catch-up; * = Main Impact Analysis; treatment group children received at least one session of the ABC intervention; control group children were matched using Propensity Score (PS) techniques and received child welfare services as usual.
The supplemental analysis (survival analysis) examined reunification and time to reunification. Given that a subsample of the child participants had experienced removal (n = 115, 56.1%), analytic examination of reunification only included children who had experienced removal during the study timeframe (i.e., children at-risk of removal were ineligible for inclusion). The federal reunification variable available in the state administrative database was utilized. For the respective referral record analyzed, children who exited care to reunification were coded “1”; children who were discharged from care for other reasons along with children who were still in care at the study end date were coded “0”. The researchers examined time to reunification, in days, by calculating the number of days in out-of-home care. Given the objective of examining the impact of the ABC intervention, it is logical to assume that the ABC intervention, for children in the treatment group, could not have had an effect on the focal child until the child was enrolled in intervention services. Therefore, the researchers subtracted the child’s child welfare discharge date from the focal child’s date of enrollment into services. For comparison group children, time to reunification was calculated by subtracting the child’s discharge date from their removal date. If a child was still in out-of-home care at the study end date, then the study end date (September 1, 2022) was imputed (i.e., children still in care used a censored date).
Independent
To best examine the treatment condition, a four-category treatment variable was utilized. The authors were intentional about using a more granular independent variable – instead of the traditional treatment versus control dichotomous variable – to meet the recent guidelines put forth by the Title IV-E Prevention Services Clearinghouse handbook (Kerns et al., 2021). The four-category treatment variable allows for separate examination of each of the ABC curriculums (e.g., mABC, ABC-I, and ABC-T). Comparison group children were the reference category (coded “0”); children who participated in at least one session of the ABC intervention were coded 1 (mABC), 2 (ABC-I), and 3 (ABC-T). Five treatment group children were excluded because they did not receive any portion of the ABC intervention.
Supplemental analyses were conducted for each of the research questions to see if treatment effects maintained or were enhanced for children who successfully completed the intervention. Children who successfully completed the intervention (i.e., completed all 10 ABC curriculum sessions), regardless of age-variant curriculum, were dichotomously coded (1 = successfully completed intervention; 0 = All other children [did not successfully complete ABC intervention and comparison group]).
Covariates
The authors intentionally controlled for variables known to impact child maltreatment and child welfare outcomes. Dichotomous covariates included: biological sex (male = 0; female = 1), Ethnicity (non-Hispanic/Latino = 0; Hispanic/Latino = 1), and whether or not the child had any prior child welfare referrals (no = 0; yes = 1). A six-category race variable (White = 1; Black = 2; American Indian = 3; Asian = 4; Pacific Islander = 5; multi-racial = 6) was computed by the study authors which combined the child’s primary and secondary racial categories; children who identified as more than one race were categorized as multi-racial. Lastly, a child’s age (continuous, in years) at study referral date was included.
Given the study design and state administrative database, none of the variables utilized in the study analyses had missing data. Therefore, all variables had 0% missingness. Further, though no statistically significant baseline differences were present between the treatment and comparison group children, the evaluators still included multiple baseline variables in the analyses given the evidence base demonstrating the impact of certain demographic and case variables on child welfare outcomes. Specifically, race, biological sex, and age at referral.
Analytic Approach
Sample Descriptives and Baseline Equivalence Results by Full Sample and Treatment Condition.
Notes. mABC = focal child is birth through 6 years of age; ABC – Infant = focal child is 6–24 months of age; ABC - Toddler is 24–28 months of age. * = Categorical treatment variable was used for baseline analysis; TAU = Treatment as usual.
aIndicates variable was analyzed on full sample, N = 205.
bIndicates variable was only applicable to/analyzed for children who had experienced removal, n = 115.
Adding to the evidence on the ABC intervention, the intentional multi-tiered analytic approach permitted examination of the potential impact of the ABC intervention, by age variant, on multiple child welfare outcomes, specifically: repeat maltreatment reports, repeat substantiated maltreatment reports, and likelihood of and time to reunification. To the best of the authors’ knowledge, this is the first study of its kind. The main analytic models examined repeat and substantiated repeat maltreatment reports. Not unusual in child welfare data is an over-abundance of zeros on the outcome variable (i.e., participant did not experience examined outcome); to account for the large proportion of zero values, the authors employed multivariate logistic regression with bootstrapping, given that the bootstrap approach does not make specific distributional assumptions and allows for a more conservative interpretation of statistical output. Specifically, 24 of the 205 children (11.7%) had a repeat maltreatment report, with only 13 (6.3%) of the sample experiencing a substantiated repeat maltreatment report. Supplementary analyses were run to see if the treatment effects maintained and/or enhanced when examining successful ABC completion verse all other children (i.e., control group and ABC dropouts).
Lastly, the authors conducted a rigorous supplemental analysis on the subsample of children who had experienced removal (n = 115, 56.1%) to observe the impact of the ABC intervention, by age variant curriculum, on a child’s likelihood of reunification and time to reunification. The authors utilized Cox proportional hazards model (i.e., survival analysis) to answer the third research question. The use of event history techniques, such as Cox proportional hazard models, permits researchers to include cases where the event has yet to occur (i.e., child has not experienced reunification/is still in out-of-home care) – critical when examining child welfare outcomes. Further, Guo and Wells (2003) explain that Cox regression, compared to other techniques, produces more robust findings and unbiased estimates of the associated model factors estimating the timing of an event – in this case, reunification.
Results
The majority of children were White (n = 110, 53.7%), non-Hispanic/Latino (n = 173, 84.4%), males (n = 122, 59.5%) with an average age of .50 (standard deviation [SD] = 1.0) years old (e.g., 6 months old) at the time of their substantiated referral. Given the study eligibility criteria, the entire sample was substance-affected with the majority of children being exposed to parental drug use (n = 165, 80.5%). Twenty-one percent (n = 43) had domestic violence reported as a contributing factor on their child welfare referral. When examining the outcome variables of interest, 11.7% (n = 24) had repeat maltreatment reports after their study referral, with only 6.3% (n = 13) of the sample having substantiated repeat maltreatment reports after their study referral. For the study referral examined, 56.1% (n = 115) of the sample was in out-of-home placement; of the children in out-of-home care, 23.5% (n = 27) experienced reunification during the study timeframe. Bivariate analyses did not reveal statistically significant differences in prevalence or substantiation of maltreatment re-reports; however, treatment group children were significantly more likely to experience removal from their home (71.2% vs. 48.9%, χ2 (df) = 10.18 (3), p = .017) and to experience reunification (38.3% vs. 13.2%, χ2 (df) = 13.18 (3), p = .004). Sample descriptives by treatment condition are presented in Table 2 including the bivariate baseline equivalence results for treatment verse control groups.
Main Impact Analyses
Logistic Regression Output (1000 Bootstrap Iterations) – Repeat Maltreatment Reports.
Notes. N = 205; SE = standard error; CI = confidence interval. Reference category is listed in parentheses. 1000 bootstrap iterations. * = bootstrap standard error.
Logistic Regression Output (1000 Bootstrap Iterations) – Repeat Substantiated Maltreatment Reports.
Notes. N = 205; SE = standard error; CI = confidence interval. Reference category is listed in parentheses. 1000 bootstrap iterations. * = bootstrap standard error.
Supplemental analyses examining children who successfully completed the ABC intervention compared to all other children were also non-significant. Models 1b and 2b in Tables 3 and 4, respectively, present the logistic regression results for repeat and substantiated repeat maltreatment reports for children who completed the ABC intervention verse all other children in the study.
Supplemental Analysis
Fifty-six percent (n = 115) of the study sample experienced removal and were eligible for inclusion in the survival analysis. In total, 115 children experienced removal (47 treatment group children [mABC = 20, ABC-I = 15, ABC-T = 12]; 68 control group children). Of the children who were removed, 23.5% (n = 27) experienced reunification during the study period (18 treatment group children [mABC = 5; ABC-I = 7; ABC-T = 6]; 9 control group children).
The observation period ranged from 1 to 1186 days with the mean number of days being 427.1 days (SD = 250.4) in out-of-home care. Children in the treatment group (i.e., attended at least one session of the ABC intervention) averaged 330.0 days (SD = 170.1) in out-of-home care compared to 494.1 days (SD = 275.0) for comparison group children. Bivariate percentile analyses revealed that 30% of treatment group children who received: mABC were reunified by 431 days (standard error [SE] = 51.26); ABC-I were reunified by 271 days (SE = 104.85); and ABC-T were reunified by 378 days (SE = 43.60). For the control group, it took 766 days for 30% of the children to experience reunification.
Survival Analysis Output– Time to and Likelihood of Reunification by ABC Curriculum.
Notes. HR = hazard ratio; SE = standard error; CI = confidence interval; * = dichotomous race variable (White [1] vs. non-White was utilized for survival analysis due to skewed distribution of categorical race variable).
None of the covariates included in the supplemental analysis significantly influenced reunification at the p < .05 level. Additionally, no statistically significant differences were found between children who successfully completed the ABC intervention and all other children (HR = 1.35, p = .478, 95% CI [0.59 – 3.13]).
It is important to note the study authors conducted separate survival analyses to examine likelihood of reunification between each ABC age-variant curriculums. No statistically significant differences were observed; the statistical output is summarized for transparency and evidential clearinghouses and other database repository purposes. Non-significant results examining ABC – Infant and ABC – Toddler curriculums compared to mABC, while controlling for all other covariates, are as follows: HR = 2.37, p = .240, 95% CI (0.56 – 9.93) and HR = 2.16, p = .140, 95% CI (0.78 – 5.99), respectively). Non-significant results examining ABC – Infant children compared to ABC-Toddler children are as follows: HR = 1.10, p = .882, 95% CI (0.33 – 3.67). Detailed results are not presented due to space constraints and are available upon request.
Discussion
Overall, impact analyses revealed promising results demonstrating the effectiveness of the three age variant ABC curriculums on the examined child welfare outcomes. Given a gap in the current parenting skills intervention knowledge base, this study intentionally focused on examining the impact of the ABC intervention on subsequent child maltreatment reports, substantiations, and likelihood of and time to reunification. Often parenting programs or interventions evaluate proximal outcomes (e.g., increases or decreases in pro-social parenting behaviors and/or child behaviors) rather than evaluating distal outcomes of the interventions’ impact on subsequent child maltreatment and reunification outcomes. This study specifically sought to address these gaps. Interpretation of the study results should be done with caution given the small sample size; however, several results are worth notating.
First, results revealed no statistically significant intervention effects between groups on subsequent maltreatment reports or substantiation – regardless of ABC curriculum. This study adds to the existing literature demonstrating that among a highly vulnerable child population, no statistically significant differences were observed on repeat and substantiation status of maltreatment reports. In other words, when examining repeat maltreatment reports and substantiation, this study provides evidence that among the target population all three ABC curriculums had similar outcomes (i.e., no statistical differences) as children who received child welfare services as usual. The null findings may, in part, be associated with the small sample and being underpowered to detect small effects sizes. Though this finding warrants further investigation on the scope of ABC’s impact on subsequent maltreatment reports and substantiation status, it supports the notion that the short-term (e.g., 10-session) ABC intervention should potentially be offered as a standard child welfare service for young, substance affected children – particularly when paired with this study’s finding of statistically significant increased rates of reunification.
Related to this finding, it is worth noting that treatment group children who were referred to and participated in the ABC intervention may have been more high-risk from the onset, thus prompting the initial referral to the program. Evidence to support this suggestion is found in the baseline sample descriptives and bivariate results which illustrated significantly higher prevalence rates of removal among treatment group children (71.2% vs. 48.9%; χ2 [df] 10.2 [3], p = .017); additional variables to potentially indicate greater risk for the children in the treatment group were not available in the state administrative dataset and thus limited the analyses.
Another potential explanation for the increased baseline removal rates observed among ABC children could be due to unidentified surveillance bias and the associated increased monitoring/observation of the child. Recent research (Green et al., 2017; Holland et al., 2024) supports the possibility of surveillance bias among child welfare involved children participating in home visiting programs and the increased likelihood of subsequent child welfare reports; however, it is presumed that the child participated in the home visiting program with their biological caregiver. Therefore, future research examining the potential of surveillance bias and ABC involvement is warranted, particularly because the ABC intervention allows for the target caregiver to be either the foster or biological caregiver.
Next, this study found that treatment group children who experienced removal, regardless of ABC curriculum, were statistically more likely to reunify during the study period when compared to their control group counterpart who received child welfare services as usual. This adds to the established literature indicating parenting programs can positively impact reunification (Akin & McDonald, 2018). Foster care agencies seeking to implement an evidence-based parenting program to increase reunification efforts between children and their families, particularly among substance-affected families, may look to ABC implementation. This study’s findings, in conjunction with recent qualitative results indicating caregiver support of ABC (Bagwell-Gray et al., 2023), should lead child welfare policymakers or administrators to consider promotion of the use of ABC in foster care agencies.
Lastly, for children who experienced removal, ABC children averaged nearly 165 days less (approximately 6 months) in out-of-home care, compared to children who received child welfare services as usual, which is a considerable amount of time in the life of a young child. Moreover, promising results were observed for ABC-I children, revealing that 30% of ABC-I children had experienced reunification after 271 days in care. However, these results necessitate further research to corroborate these findings, though, it appears – as a whole – ABC does have a positive impact on both time to and likelihood of reunification for children in out-of-home placement.
Again, the authors would like to caution the interpretation of the study results and highlight that there are many, cross-system individual- and family-level factors that contribute to reunification and this respective study examined the impact of one parenting intervention. The findings of this study provide support that participation in the ABC intervention for young children in out-of-home placement increases their likelihood of experiencing reunification, compared to child welfare services as usual. One potential mechanism for this finding is the mediating role of the child’s ability to regulate their behaviors and emotions. Consequently, the results of this study reinforce the foundational concepts of the ABC intervention--by enhancing emotional regulation skills in target children, ABC potentially contributes to fostering positive outcomes in reunification efforts, a foundational outcome for child welfare services. However, these findings and conceptual mechanisms warrant further investigation particularly among young children who are substance-affected.
This study does have several limitations worth noting. First, though rigorous, a quasi-experimental design does not equate to an RCT, and the results should be interpreted accordingly and not be inferred as causal. Study findings should be understood as tentative and necessitating further investigation and replication within varying sample populations. PSM techniques were employed to minimize potential selection bias and increase the rigor of this impact analysis, however, the study authors cannot know for certain if the two study conditions were truly equalized by the matching process. Next, caution should be taken when interpreting the results due to the small sample. Important variables may have been missing or inaccurate. For instance, given the study start date and data sharing agreements, lifetime child welfare state administrative data were not received; therefore, it is possible that our covariates did not adequately account for prior child welfare referrals and that some children who met study criteria had referrals, maltreatment reports, and/or removals prior to October 1, 2017. However, given the target age range of child participants prior data points would have been limited for this young sample. Other variables known to influence likelihood of reunification, such as substance abuse treatment receipt, were not available and not included in the analyses, which is a limitation. Additionally, the researchers were limited in the variables collected and stored within the state administrative database; variables indicating the type of services or interventions received by the children in the comparison group were not available. Further, this study overlapped with the COVID-19 pandemic, and the global pandemic circumstances may have uniquely impacted the study participants in a way that was not measurable, potentially resulting in spurious outcomes. These results were obtained from one Midwestern county among substance-affected families involved in child welfare, and, therefore, are not generalizable. Lastly, it is important to note that by virtue of the analytic technique selected (survival analysis), time under observation (i.e., study timeframe) could have a significant impact on the outcomes-particularly when study enrollment is constant and continuous; therefore, the longer the observation window, in theory, the greater the likelihood of experiencing the outcome. Given this, future research is needed to corroborate these findings using a larger observation window for child welfare involved children and families.
Conclusion
Despite the limitations, this study makes a novel contribution to the field by examining multiple distal child welfare outcomes among a high-risk, substance-affected population for all three curriculum versions of a widely used parenting intervention: Attachment and Biobehavioral Catch-up. Although no differences were found between study groups for subsequent maltreatment reports and substantiation, promising results were found for time to and likelihood of reunification for ABC participants, regardless of age-variant curriculum. That is, when compared to children who received child welfare services as usual, children who received mABC, ABC-I, and ABC-T were all significantly more likely to experience reunification during the study period - and reunified at faster rates. Future research should investigate if ABC booster sessions would help mitigate the risk for repeat maltreatment reports, regardless of removal status. Moreover, efforts should be made to investigate the impact of policy and how state and federal-level policies can be leveraged to decrease subsequent maltreatment reports and increase permanency among this vulnerable, child welfare-involved population having positive and lasting implications on the child’s safety and well-being.
Footnotes
Declaration of Conflicting Interests
The author(s) declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.
Funding
The author(s) disclosed receipt of the following financial support for the research, authorship, and/or publication of this article: Original study data was collected as a component of a Regional Partnership Grant initiative funded by the Children’s Bureau, Administration on Children, Youth and Families, Administration for Children and Families, U.S. Department of Health and Human Services, under grant number CU17000305.
