Casey Family Programs’ (CFP) vision is to reduce the need for foster care while supporting children and families. It provides foster care and related services to children, families, and child welfare systems. It is also a data hub for information about those children. While front-line practitioners use a mix of services to support each case, better use of data could help them refine their services instead of using a “one size fits all” approach.
Community Science provided data science expertise and was tasked with facilitating collaborative inquiry to help CFP precisely target the correct quantity and quality of services. We were responsible for co-creating predictive and prescriptive models, applying machine learning algorithms, and training CFP on the step-by-step modeling process. This project was unique in that CFP desired to build the skills, tools, and capacity to continue the modeling process in-house—making the entire project a hands-on collaboration.
We worked with the Director of Systems, Data, and Reporting to build models to assess cases at intake and during service to pinpoint what types of services work for certain subsets of children to best support a path to legal permanence. Part 1 of the project involved creating the model from CFP’s original data. In Part 2, we developed a plan to integrate data-driven learning into all CFP programs. Such predictive prescriptive analytics can help front-line practitioners’ decisions about the care of foster children be more accurate on a case-by-case basis. Part 3 of the project engaged practitioners and applied the model to real-time decision making and learning. The training resulted in building CFP’s capacity to use the model going forward.
This project was important in the field of child welfare with the longstanding challenge of best determining what will work for each case. To date there are evidence-based programs and practices, but the “standards of care” that drive service delivery are not precise enough to address the unique needs of each case. As a result of this project, CFP identified and determined the unique set of circumstances for nine different sub-populations of cases. Additionally, the client was able to leverage the findings to further understand their program effectiveness for those sub-populations.