Validating a Predictive Risk Model for Child Abuse and Neglect With Adolescent Outcomes.

Area of Science:

• Child Welfare Research
• Adolescent Health
• Risk Prediction Modeling

Background:

• A predictive risk model (PRM) was developed to identify children at risk of future child protection involvement following maltreatment investigations.
• This study evaluated the PRM's effectiveness in predicting adverse outcomes like arrest and death in adolescents.

Purpose of the Study:

• To assess the external validity of a PRM in predicting non-foster care related adverse events (arrest, death) in adolescents.
• To determine the utility of the PRM for guiding prevention-oriented services for maltreated adolescents.

Main Methods:

• Utilized child welfare and vital statistics records for adolescents (n=72,340) investigated for maltreatment between ages 11-17.
• Calculated risk scores using a random forest algorithm based on initial maltreatment report data.
• Linked adolescent records to subsequent arrest and death records up to age 21.

Main Results:

• 5.8% of investigated adolescents experienced a juvenile arrest or death before age 21.
• 43.9% of adolescents who experienced an arrest or death were in the highest risk decile of the PRM.
• The PRM demonstrated strong predictive accuracy for these adverse events.

Conclusions:

• The PRM, initially trained for foster care placement, showed significant external validity in predicting adolescent arrests and deaths.
• A notable time interval exists between maltreatment investigation and adverse events, presenting an opportunity for timely interventions.
• Findings support the use of PRMs to target support services for at-risk adolescents and their families.