Using Predictive Analytics to Improve Outcomes in Child Welfare

This brief provides an overview of the responsible use of predictive analytics in child welfare. “Predictive analytics” refers to the practice of extracting information from existing data sets and identifying patterns that may help to predict future outcomes. In the context of child welfare, predictive analytics is associated with identifying levels of risk for maltreatment. This type of risk modeling has tremendous potential to provide empirical support for the allocation of resources, helping child welfare leaders manage complex and competing demands.

What We Did

Researchers reviewed the strategies that data analysts use to improve the reliability of decisions by child welfare leaders, presenting specific examples of the application of the approach. They then used this review to present an overview of the responsible use of predictive analytics in child welfare. Researchers also examined ethical and methodological considerations for predictive analytics.

What We Found

While predictive analytics has tremendous potential to assist with addressing problems faced by children and families, policymakers and system leaders must understand the method and its appropriate applications. Effective and responsible use of predictive analytics in child welfare can provide critical information needed to assess risk and allocate resources. Using predictive risk modeling—which applies predictive analytics to routinely collected data and looks at large numbers of previous cases–researchers can predict the likelihood that children and families will experience specific outcomes. By applying predictive analytics with transparency, integrity, and responsibility, policymakers and system leaders can make decisions that improve well-being.

What It Means

  • The effective use of predictive analytics requires a collaborative, transparent, and iterative process to plan and support implementation.
  • Researchers should partner with those using predictive analytics to develop and refine approaches and models; understand factors associated with clearly defined outcomes; direct attention to cases requiring more intensive service or supervision (not to impose additional requirements on families); and minimize the effects of race and ethnicity on future decision making.
  • Predictive analytics should utilize datasets with appropriate breadth, depth, and quality.
  • Child welfare systems should employ assessment strategies that capture elements of well-being, including strengths, protective factors, and functioning.
  • Practice protocols and policy should prioritize human judgment when integrating predictive analytics into service delivery and agency operations.
    • To ensure ethical use of predictive analytics models, researchers should be transparent in model development, model application, and model refinement.
      Recommended Citation
      Chadwick Center & Chapin Hall. (2018). Making the most of predictive analytics: Responsive and innovative uses in child welfare policy and practice. San Diego, CA & Chicago, IL: Collaborating at the Intersection of Research and Policy.
Making the Most of Predictive Analytics