Predictive Risk Modeling for Child Protection

Every year, more than 3.6 million referrals involving about 6 million children are made to child protective services (CPS) in the United States. It is estimated that around one in three American children will have been investigated by CPS for maltreatment by age 18. Flooded with calls, and short on resources, CPS systems are struggling to identify and protect children at risk. This document describes how decision-support tools, such as predictive risk models, can help CPS workers navigate an environment where it is hard to separate “noise” from “signal.”