What is predictive risk modeling and how can it be applied to child welfare?

A single call to the child protection hotline can set in motion events that have the potential to change a child’s and family’s future, including saving that child’s life. Erin Dalton, deputy director of the Allegheny County (PA) Department of Human Services’ Office of Data Analysis, Research and Evaluation, introduces the hotline as a critical decision point that has significant implications for both vulnerable children as well as for the system. Ms. Dalton presented with Dr. Rhema Vaithiananthan at a two-day convening hosted by Casey Family Programs in May 2018, where child welfare leaders and experts came together to share innovative strategies designed to improve child safety and reduce child abuse and neglect fatalities. 

Why is the hotline such a critical decision point in child welfare?
RUNTIME: 04:49
RUNTIME: 04:49

Dr. Rhema Vaithiananthan, professor of Economics at the Auckland University of Technology and co-director of the Centre for Social Data Analytics then provided an overview of predictive risk modeling and its applicability in child welfare. She explained how the application of a predictive risk model at the hotline can quickly harvest historical case records that screeners do not have the time to review, but are necessary to support effective decision-making at the hotline. The risk score generated by the model tells the hotline screener whether a particular referral has similar patterns to other cases that have the highest likelihood of placement within two years; the screener is still responsible for making a well-informed decision about the referral. The model was tested in Allegheny County, Pennsylvania, with past referrals, and evaluators found that 50 percent of children with the highest score were removed from their home within two years, while  only 1 percent of the children with the lowest score were eventually placed into out-of-home care.  

What is predictive risk modeling and how can it be applied to child welfare?
RUNTIME: 08:25
RUNTIME: 08:25

Erin Dalton finished the presentation by discussing the impact that predictive risk modeling has had in Allegheny County. Since implementation of the model in 2016, higher-risk children are being screened in more often and lower-risk children are being screened out more often. In addition, it has reduced the variation in decision-making so that children with the same level of risk are treated in a similar way. Ms. Dalton stressed the need to always be transparent about the processes, learn from other fields, listen to critics, and continue to evaluate.

Erin Dalton
RUNTIME: 05:45
RUNTIME: 05:45

Download PDF presentation

Send this to a friend