Precision inside a residential treatment program

The challenge

A residential treatment program for children needed to understand not just whether it worked on average, but which children it worked for, and what specific features of treatment made the difference, especially for the risk of hospitalization after discharge.

The approach

BCT applied Precision Causal Modeling, predictive analytics, and machine learning to the program's existing data. Rather than reporting a single average, the analysis identified meaningful subgroups of children and isolated the treatment features associated with different post-discharge outcomes.

The results

• Identified eight matched subgroups of children with distinct needs and responses
• Surfaced the treatment features associated with post-discharge hospitalization
• Delivered more precise measures of program success than average-based reporting
• Gave leaders and practitioners evidence they could act on for specific children