As The Chronicle of Social Change has been reporting over the past two years, various jurisdictions have been exploring new tools to focus the attention of child welfare systems on the children most at risk of subsequent abuse or neglect. The mainstream media has begun to notice, as demonstrated by CNBC’s recent report on Los Angeles’ contract with software company SAS to develop such a tool for its child welfare system.
These new approaches generally rely on predictive analytics, which means using patterns in data to predict future outcomes. Despite the recent media coverage, there is still some confusion about what is meant by this term, how it differs from current approaches like Structured Decision Making (SDM), and the distinctions between the various new approaches.
To understand these new approaches, it is important to understand how child protective services (CPS) works now. I can describe the process in the District of Columbia, where I once served as a caseworker.
CPS workers in the District use checklists to interview children, parents, teachers, and others about an allegation of abuse and neglect. They generally rely on the verbal answers to these questions, although they do receive access to data from schools and public assistance agencies.
The District of Columbia, like jurisdictions in over 20 states, uses an SDM tool to help social workers decide how to proceed at the conclusion of an investigation. Social workers fill out checklists on the computer. SDM assigns points to each risk factor, such as “primary caregiver has historic or current drug or alcohol problem.”
Based on the worker’s checkoffs, the software spits out a recommendation to remove the child or keep her at home.
SDM has at least two major flaws. First, it can be manipulated to recommend the action that the social worker wants to take. Second, it does not obtain any new data but simply uses what the social worker plugs in. That’s why in my experience in D.C., SDM was treated as a meaningless form to be filled out, not as a tool for making decisions.
A new set of predictive tools, known as predictive analytics, is being developed for governments in Los Angeles, Allegheny County, Pennsylvania, and New Zealand. These tools are designed to produce a numeric risk score for each child assessed. A high risk score would target the child and family for special attention, which might include intensive services and monitoring.