Identifying treatment moderators of a trauma-informed parenting intervention with children in foster care: Using model-based recursive partitioning

Background : Trauma-informed parenting interventions have been used in child welfare to help caregivers respond to children in trauma-informed ways that can mitigate the effects of maltreatment and build strong caregiver-child relationships. Existing studies support its effectiveness with the foster care population. However, to further advance its development, one key step is to identify subgroups of participants that respond differently from the intervention. Objective: To identify pre-treatment moderators that can distinguish subgroups of caregivers and children that benefit differently from the intervention. Participants and setting: 414 foster care children (age 3 or younger) and their caregivers were assigned either to the trauma-informed parenting intervention in the Illinois Birth through Three Title IV-E waiver demonstration or to a comparison group that received services as usual. Methods: Model-based Recursive Partitioning (MOB) was used to identify treatment moderators and moderator interactions. MOB fits a parametric model and uses a data-driven method to find subgroups for which the specified parametric model has different parameters. Two parametric models (logistic and linear regression) were used in accordance with two outcomes: reunification (binary), and caregiver-child attachment (continuous). We examined 21 potential pre-treatment moderators in both models. Results: For reunification outcome, three treatment moderators interact to produce different subgroups of participants who responded differently to the intervention: (a) caregivers’ relationship status (kin vs. non-kin/permanent caregivers), (b) caregiver-child attachment, and (c) case history of physical abuse. For attachment outcome, caregivers’ age was found to be a treatment moderator. Future developments of trauma-informed interventions should consider these moderators.