On Thursday, May 17 the American Enterprise Institute (AEI) focused on, Preventing Harm to Children through Predictive Analytics.
A somewhat formal definition of predictive analytics would define it as a practice of taking data and the information gathered from that data to determine patterns and predict outcomes and trends. It predicts or forecasts what might happen in the future. Used in some settings, such as law enforcement and policing practices, some see it as a useful tool within child protection and child welfare.
The AEI discussion (re-scheduled from January) was a two-part discussion focused first on the use of predictive analytics in Hillsboro County Florida with presentations and discussions led by Andrew Barclay, co-founder, Foster Court Improvement, Bryan Lindert, Eckert Kids and Greg Povolny, Mindshare Technology. It was moderated by James Payne, Public Consulting Group. The second panel focused on a leader’s perspective with participants including Joette Katz, Connecticut Department of Children Family Services, she was joined by Susan Smith, also from the Connecticut Department, David Sanders, Casey Family Programs and chair of the Commission to Eliminate Child Abuse and Neglect Fatalities, and John Specia, Texas Department of Family and Protective Services. This panel was moderated by Maura Corrigan a fellow at AEI and former head of the Michigan Department.
Much of the focus by Eckard and other experts from the first panel discussed how the project began in Hillsborough County Florida (a number of child fatalities within a short period of time) and how they started to conduct an analysis and review of past data. That effort has continued and expanded into other areas including an on-going analysis of re-entries into foster care.
A critical component for the predictive analytics is to gather as much data across as many
systems as possible. As a result of their work they believe they have built a model that can assist caseworkers in their decision-making. Their presentation focused less on global policy making and action and more in terms of supervision that can be provided to caseworkers in specific cases, practices and decisions. In addition to their on-going work in child protection they are also using analytics to examine reunification practices and re-entries into foster care.
The first panel presentation was on the importance of how the data can build a profile that can assist caseworkers and importantly, supervision. It provides a “second set of eyes.” This first panel did not look at overall policy but practice and decision-making.
In the second panel presentation there was a discussion by the representatives from Connecticut and Texas about how such tools might and should be used. David Sanders reflected on some of the research and findings presented during the Commission to Eliminate Child Abuse Neglect Fatalities. There seemed to be agreement on this panel that it can be an important tool but such programs can’t replace judgment rather it is a tool to help workers make judgments. And to determine high risk and high priorities.
Speakers emphasized key provisions in implementing any kind of predictive analytics. These include having clear information, values and commitments (for example addressing disproportionality), guiding principles that have to be respected, the importance and significance of good data and research, the technical infrastructure to implement the practice, and support staff including IT and research.
Here is the event summary and video.