California Child Welfare Predictive Risk Model Proof of Concept

Modelled closely on the Allegheny Family Screening Tool, in 2016 Vaithianathan and Putnam-Hornstein also began PRM work in California. Given that the state does not have an integrated data system, the goal of this project was to assess whether a PRM built exclusively from child protection records could approach the accuracy of the AFST. This proof-of-concept also included comparisons between scores generated through a PRM and risk levels assigned through the use of the SDM® Family Risk Assessment tool (i.e., “SDM® risk tool”) which has been in use in California since 1998. Findings indicated that the PRM was more accurate than the SDM® risk tool in identifying children who would have chronic or intensive involvement with the child protection system. The state is currently developing implementation plans with county stakeholders.


Vaithianathan, R., Maloney, T., Putnam-Hornstein, E., & Jiang, N. (2013). Children in the public benefit system at risk of maltreatment: Identification via predictive modeling. American Journal of Preventive Medicine, 45(3), 354-359.

This paper co-authored by Rhema Vaithianathan and Tim Maloney (CSDA co-director 2016-2019) outlines the method of using an automated predictive risk model to identify children at high risk of maltreatment, as well as discussing the strategic targeting of prevention activities toward these individuals.

Cuccaro-Alamin, S., Foust, R., Vaithianathan, R., & Putnam-Hornstein, E. (2017). “Risk assessment and decision making in child protective services: Predictive risk modelling in context”. Children and Youth Services Review 79

The authors, including Rhema Vaithianathan, review the literature and provide a context for predictive risk modelling in the current risk assessment paradigm in child protective services. They describe how predictive analytics or predictive risk modelling using linked administrative data may provide a useful complement to current approaches.

Chouldechova, A., Putnam-Hornstein, E., Benavides-Prado, D., Fialko, O., & Vaithianathan, R. (2018). A case study of algorithm-assisted decision making in child maltreatment hotline screening decisions. Proceedings of Machine Learning Research, 1-15.

Diana Bendavides-Prado, Rhema Vaithianathan, Oleksandr Fialko (CSDA research fellow 2017-2018), Alexandra Chouldechova and Emily Putnam-Hornstein describe their work on developing, validating, fairness auditing, and deploying a risk prediction model in Allegheny County, PA, USA. They discuss the results of their analysis to-date and highlight critical problems and data bias issues that present challenges for model evaluation and deployment.

Predictive Risk Modeling: Practical Considerations, Children’s Data Network

Related news

Quick facts

Location: California, United States

Partner/s: Led by Children’s Data Network together with California Child Welfare Indicators Project, California Department of Social Services, California Department of Public Health, County Child Welfare departments in Los Angeles, Monterey and San Francisco, CSDA

Timeframe: 2016-June 2018