AUTHORS:
Cody Inman, Delegated Commissioner, Administration on Children, Youth, and Families, Administration for Children and Families, HHS
Alex J. Adams, Assistant Secretary, Administration for Children and Families, HHS
Key Points
- The Administration for Children and Families (ACF) awarded $6 million to 10 child welfare jurisdictions to test predictive risk modeling (PRM) and related predictive analytics approaches across the child welfare continuum—from prevention and hotline screening to placement stability and permanency.
- Grants were awarded to the following states and jurisdictions: Indiana, Missouri, Nebraska, New Jersey, North Carolina, Ohio, Oklahoma, Texas, Muscogee Creek Nation, and the District of Columbia.
- The awardees’ projects will demonstrate technology options including machine learning, natural language processing, automated case summaries, dashboards, cross-agency data integration, text analytics, and predictive models embedded within Comprehensive Child Welfare Information Systems (CCWIS).
- The projects can advance A Home for Every Child on both sides of the home-to-child ratio: increasing or preserving safe, available family homes—the numerator—while safely reducing the number of children who enter or remain in foster care—the denominator.
- Other states and Tribes can advance PRM even outside of the grant opportunity. ACF has clarified pathways for eligible PRM costs to receive Title IV-E support through applicable CCWIS or administrative cost-allocation rules.
From Opportunity to Demonstration
ACF released a March 2026 issue brief, Modernizing Child Welfare Technologies and Tools: Opportunities for Predictive Risk Modeling (PRM) to Improve Child Safety and Outcomes. The report described how PRM can use information already captured in child welfare systems to provide workers and supervisors with timely, data-informed decision support. It highlighted several potential use cases for PRM and highlighted case studies from several pioneering county child welfare agencies as well as the first statewide application in Idaho.
To advance this work, ACF issued a Notice of Funding Opportunity (NOFO) to provide seed funding for child welfare agencies to test use cases of PRM. ACF received 21 applications from 20 jurisdictions and ultimately awarded funding to 10 jurisdictions: eight states, the District of Columbia, and a Tribal government.
These 10 new demonstration projects move this PRM work from opportunity to implementation. They are designed to build an evidence base across different jurisdictions, workflows, and technology environments while maintaining safeguards, transparency, evaluation, and human oversight. Predictive tools are intended to support—not replace—professional judgment; final child welfare decisions remain with trained professionals. This issue brief will highlight how these 10 jurisdictions are envisioning PRM as a potential roadmap for other states considering PRM as an opportunity to advance their child welfare practices.
The Child Welfare Continuum and Potential Uses
The demonstration projects are organized across a child welfare continuum that begins before foster care entry and extends through permanency and post-permanency. The continuum is useful because PRM is not a single tool or use case. Jurisdictions can apply different analytic methods at different decision points, depending on the problem they are trying to solve and the data and technology already available.
Across the continuum, potential uses include:
- Prevention / Family Support — Identify needs earlier and help prevent escalation or unnecessary foster care entry.
- Hotline & Intake — Prioritize referrals and help distinguish higher-risk from lower-risk allegations.
- Investigation / Safety — Assess risk and safety, surface relevant case history, and support timely supervisory review.
- In-Home / Case Planning — Target services, monitor changing risk and case complexity, and support case planning.
- Placement & Stability — Improve child-caregiver matching, identify risk of disruption, and target supports that stabilize placements.
- Reunification / Permanency — Identify barriers to permanency, reduce re-entry, and support timely, safe exits from foster care.
- Post-Permanency — Monitor stability and identify families or youth who may benefit from additional support.
Several capabilities, including workforce productivity, resource allocation, quality assurance, data governance, administrative automation, and evaluation, cut across every stage of the continuum. By funding projects that span these areas, ACF can better understand where predictive analytics delivers the greatest value and how different approaches perform across child welfare settings.
What the 10 Jurisdictions Will Test
The demonstrations intentionally differ in use case, technical approach, and point of intervention. Table 1 summarizes the portfolio and shows how each project may contribute to the home-to-child ratio.
Table 1. Summary of Jurisdictions Receiving PRM Grants
Jurisdiction | Use Case | Technology / Data Approach | Home-to-Child Connection |
|---|---|---|---|
| Indiana | Placement matching and stability; reduce disruptions and improve timely, appropriate matches. | Machine learning; historical and newly structured case data; workforce-facing decision support and transparency features. | Preserve caregiver capacity and support faster permanency. |
| Missouri | Identify elevated risk of placement disruption; provide actionable case summaries to workers and supervisors. | Enterprise analytics; natural language processing; predictive risk scoring; automated case summarization and recommendations. | Improve stability and reduce avoidable loss of foster-home capacity. |
| Nebraska | TRACK: earlier intervention, case planning, resource allocation, safety, and permanency. | Predictive modeling; interactive dashboards; data visualization; integrated data; internally managed analytics capacity. | Prevent entries where possible and accelerate safe permanency. |
| New Jersey | Support decisions about safe CPS case closure and identify families needing additional support. | Child welfare administrative data; existing safety/risk information; model validation, governance, monitoring, and workforce support. | Reduce repeat involvement and avoidable out-of-home placement. |
| North Carolina | Strengthen hotline intake, assessment, and supervisory decision-making statewide. | PRM embedded in PATH-NC CCWIS; data-quality assessment; governance; county implementation support; continuous quality improvement. | Improve targeting of intervention while reducing unnecessary system involvement. |
| Ohio | Identify youth at risk of foster care re-entry and earlier service, education, and supervision needs. | Predictive analytics; model development and recalibration; performance monitoring; longitudinal evaluation. | Prevent re-entry and support durable permanency. |
| Oklahoma | Identify family needs earlier and improve cross-system service connections so families can remain safely together. | Predictive analytics; cross-agency data integration; improved frontline data access. | Reduce foster care entries through prevention and earlier support. |
| Texas | Validate and modernize the existing investigation risk assessment and improve prediction of future maltreatment. | Investigation records; historical outcomes; unstructured case narratives; modern text analysis; regional validation. | Improve service targeting and prevent avoidable entry or re-entry. |
| Muscogee Creek Nation | Strengthen prevention-focused decisions through a tribally led approach integrated with Peacemaking/Circle Processing. | Tribal child welfare data; locally controlled governance; predictive analytics; model validation and continuous monitoring. | Use culturally grounded early supports to keep children safely with family. |
| District of Columbia | Predict prevention-service needs; improve placement matching and stability; support hotline screening and maltreatment risk assessment. | Predictive AI embedded in the Microsoft-based STAAND CCWIS; integrated administrative data; ongoing validation and evaluation. | Addresses both sides of the ratio through prevention, permanency, and placement stability. |
One Shared Outcome: More Homes Relative to Children Who Need Them
The demonstrations differ intentionally. Some focus upstream on preventing foster care entry. Others focus on placement stability, re-entry, case closure, or faster permanency. Still others test how analytics can improve the speed and consistency of frontline and supervisory decisions. Together, they support the organizing goal of A Home for Every Child: homes waiting on children, rather than children waiting on homes.
The home-to-child ratio makes this strategy concrete. Jurisdictions can improve the ratio by increasing and retaining safe foster and kinship homes—the numerator—while safely reducing entries, preventing re-entry, and accelerating permanency to reduce the number of children in foster care—the denominator. This grant cluster is especially well positioned to test the denominator side through prevention, better risk assessment, reduced re-entry, and faster permanency. Placement matching and stability projects can also strengthen the numerator by reducing disruptions and helping preserve caregiver capacity.
Beginning in October 2026, the home-to-child ratio will provide a common system-level measure for reporting whether jurisdictions are moving toward a sufficient supply of family homes relative to the number of children who need them. With all 50 states, the District of Columbia, and Puerto Rico joining A Home for Every Child, ACF aims to evaluate which strategies, including PRM, contribute to increased ratios.
Project-specific evaluations can complement that measure with information on child safety, placement stability, workforce efficiency, model performance, and other outcomes.
Opportunities for States to Advance PRM Beyond the Demonstration Grants
The 10 demonstration projects will help ACF and the field learn what works, in what settings, and with which safeguards. But they are not the only pathway for a jurisdiction interested in PRM. ACF has clarified that predictive analytics and PRM functionality may be financed within or alongside a CCWIS. Costs must be appropriately allocated to benefiting programs, but costs allocable to Title IV-E may qualify for federal reimbursement under applicable CCWIS or administrative cost-allocation rules. Eligible Title IV-E administrative costs generally receive 50 percent federal financial participation.
Other states and Tribes can therefore move forward now by identifying a high-value use case, assessing available administrative data, establishing governance and validation requirements, and determining the appropriate Title IV-E funding and technology pathway. PRM need not require wholesale technology replacement; analytic tools can be developed and tested within existing IT environments and during longer-term CCWIS modernization.
Conclusion
These ten PRM demonstrations create a practical national learning agenda for PRM in child welfare. They will test different cases, technologies, governance approaches, and points along the child welfare continuum while keeping professional judgment and human oversight at the center of decision-making.
Will these tools help jurisdictions safely improve the home-to-child ratio? To do so their implementation would lead to fewer children entering or remaining in foster care, more stable and available family homes, and/or faster safe permanency. As ACF begins reporting the ratio this fall, the demonstration sites will be positioned to show how modern analytics contribute to measurable progress under A Home for Every Child and provide other states and Tribes with practical models they can adapt using available Title IV-E administrative funding pathways. The demonstrations will also contribute to a growing body of knowledge about implementation and outcome improvement that ACF will make available to support peer-to-peer learning, replication, and continuous improvement across jurisdictions.