Applying Predictive Analytics
Identifying family needs and improving the allocation of resources
What are the risk factors for families with frequent involvement with the child welfare system? A large child welfare jurisdiction asked Chapin Hall to help them answer this very big question.
In response, our Implementation Collaborative’s data analytics team developed predictive analytic models to address risk of repeat reports for child abuse or neglect. We then worked with the agency to incorporate these findings into system operations. With this information, the child welfare agency is able to direct additional and timely resources to families who have the highest indicators of frequent system involvement.
Chapin Hall is using predictive analytic models to provide a data-informed understanding of individual, family, and community needs to improve child and family outcomes. By applying methodological expertise and implementation experience to agency operations, we are integrating responsible, rigorous predictive analytic approaches to enhance resource allocation and service effectiveness. This allows service providers to:
- Maximize the potential of evidence-based interventions,
- Make resources available to families based on their level of need, and
- Gauge provider performance equitably.
New applications of this analytic work are underway, including helping agencies build internal capacity to develop, refine, and routinize these models. Algorithms are applied to performance assessment of private providers for comparison with other providers serving similar cases.
Chapin Hall is committed to the responsible use of predictive analytics. We are defining an approach to reduce the likelihood of reinforcing systemic bias in the models. This includes:
- Models are transparent and available for review.
- Stakeholders weigh in to ensure the validity of the data inputs and outputs, and the relevance of models for practice enhancements.
- A dedicated advisory committee with diverse representation ensures equity concerns are addressed.
- Procedures for responsible oversight of the applicationof the models are developed.
- Models guide service planning and resource allocation decisions rather than legal or permanency determinations.
This work at Chapin Hall is led by Policy Fellow Dana Weiner and Senior Researcher Brian Chor. Dana’s work focuses on evaluating programs and improving the ability of state child welfare agencies and juvenile justice systems to use data to inform decision-making and support implementation. Dana is also a faculty member at the Northwestern University Feinberg School of Medicine and the former Senior Policy Advisor to the Director of the Illinois Department of Children and Family Services. Brian provides expert consultation to adopt predictive risk modeling using administrative data to improve case management and decrease repeat occurrences of maltreatment, placement, and involvement with the child welfare system. He is also a licensed clinical psychologist and a child welfare and child mental health services researcher, and a former researcher for American Institutes for Research.
Contact Weiner or Chor for more information about the use of predictive analytics.