The CHAMP Decision Support Tool Development Program (C-DST-DP) is our vehicle for examining all predictive and causal models produced by the Predictive and Causal Modeling Laboratory to determine which have the most promise for use with children and families in settings where they typically receive care. We enable promising models to realize their potential by developing Decision Support Tools (DSTs) that guide providers’ decisions on interventions to reduce identified risks.

CHAMP Center DSTs are designed to be highly personalized by identifying child and family risks based on gathered information and recommending precise intervention strategies that address the specific causes behind their vulnerabilities. These DSTs take such information as input and calculate their risks for specific maltreatment-related outcomes (e.g., sexual abuse, self-destructive behaviors), as well as the causes of their vulnerabilities. Based on these calculations, the DSTs offer intervention recommendations based on known interventions effective for targeting those causes and also estimate the potential reduction in risk if the child and/or family receive those interventions. 

However, before any predictive or causal model is ready for inclusion in a DST, the models must be carefully evaluated by experts and other stakeholders familiar with the needs of at risk children and families and those providing their care. Accordingly, the C-DST-DP includes a Translation Team of such experts and stakeholders to carefully review models and offer their perspectives on their potential to inform care safely and effectively. Members of this team include clinicians, clinical investigators, agency administrators, services system administrators, and family representatives. The CHAMP Translation Team also includes data scientists from the CHAMP Center and professionals with experience implementing maltreatment prevention programs from the CHAMP Network to have the diversity of perspectives needed to translate models to tools for use in practice settings.

There are many reasons why models that seem promising based on their statistical results may not actually help reduce risks for children and families. For example, causes identified in models must be practically assessable in typical care settings to determine a child’s risk and the intervention needed to reduce it. Even if the factor can be feasibly assessed in such settings, an intervention targeting it must also be available and practically deliverable. All potential harms, risks, and biases should also be carefully considered before recommending a model for use in a decision-support tool (DST). 

The CHAMP Translation Team reviews all such considerations before recommending that a model move into DST production.