Predictive Analytics within a Prevention Science Framework

This first paper will define predictive analytics and illustrate the importance of clear evaluation criteria with recent examples of predictive analytical development efforts in prevention programs, child protection agencies and juvenile justice agencies. These examples will help illustrate the need for theoretically-informed development and evaluation of Predictive Analytics, then how a Prevention Science framework supports these needs as well as the system change efforts often needed to ensure Predictive Analytics can effectively and equitably help improve service delivery for children, youth and families.

Jerry Milner, Associate Commissioner at the Children’s Bureau, and Acting Commissioner for the Administration on Children, Youth and Families, intends to reform the current child protection system to promote family resiliency and change the focus to primary prevention of child maltreatment. Thus, most examples will focus on the development of Predictive Analytics to help prevent harm to children and youth. Possibilities include longitudinal studies referencing available social service and health information to estimate the likelihood child harm, child protective service agency’s use of actuarial risk assessment within a structured approach to decisions and service monitoring, exploratory research by adult protective agencies for similar reasons, and juvenile justice agency frameworks incorporating actuarial risk and offense data by to prevent system involvement among youth with moderate or lower likelihoods of future criminal behavior, and reduce youth placement out-of-home. Review of examples will begin with the type of practice question being addressed, the perceived role of Predictive Analytics in practitioner and/or system decision making, and methods and results used to test each Predictive Analytics approach. Finally, the Prevention Science framework will be used to critique each approach, the appropriateness of this question, and identify how this theoretical framework might affect future, similar efforts. Potential application of each model will be considered, to generate discussion about the trade-offs between accuracy and equity, and the balance of client and system benefits with the potential for unintended consequences.

The paper and presentation will conclude by reviewing how the Prevention Science framework can help ensure that Predictive Analytics and other screening criteria development efforts are effective and equitable. The theoretical framework helps guide these often empirical efforts by examining whether the question being explored is appropriate for a Predictive Analytics approach, how best to judge the accuracy and equity of results given the question and decision point being informed, and how to successfully apply an accurate and equitable tool using an organizational and systems-informed approach to implementation.