Predictive risk modeling would improve Dominique Moody Safety Act
Some child maltreatment deaths are so horrific, and so clearly preventable, that the only honest response is to ask: what are we going to do differently?
Dominique Moody’s death is one of those for North Carolina. The 6-year-old died weighing just 27 lbs in a home infested with rats and both human and animal feces. She had open wounds and fractured bones in various stages of healing. There is evidence of a ligature mark on her right ankle. Her guardians would lock her in a dog crate and restrain her with duct tape.
Despite the limited information that has been released — and which the public should have every right to review — it is clear that our systems of protection failed her. First, child welfare placed her with a maternal aunt charged with her death, reflecting a national shift in child welfare to a “kin-first culture” that can treat child safety as a secondary consideration. Next, the courts formalized the custody transfer, with records suggesting limited vetting and no ongoing oversight. And an unknown number of adults failed to act, even after more than 50 calls to law enforcement and multiple child maltreatment investigations.
Deaths such as Dominique’s are understandably met with public outrage. Governors know child welfare is a public system nobody asks about during an election — until one tragedy means it’s all anyone talks about. A typical playbook follows. Sometimes we restructure. When I worked in New York City, families still referred to the agency as the “Bureau of Child Welfare” — despite the agency having been re-organized (and renamed) twice over. The name had changed. Little else had. Other times we convene commissions and produce recommendations that are quickly forgotten and never implemented. Child welfare leaders know to expect a short tenure on the job as the next death could be career-ending. The same cycles, across our nation and still, we are failing our children.
Yet, in this tragedy, North Carolina has already taken steps that can meaningfully improve both child safety and public accountability. To go beyond a reactive cycle that produces little result for these kids, North Carolina should look to implement predictive risk modeling (PRM).
The sponsors of the bipartisan Dominique Moody Safety Act (HB 1144) are already focused on the right question: When repeated community concerns signal a strong possibility of ongoing abuse and neglect, when the same family appears in our records again and again, at what point do we treat that pattern as the emergency it is?
Legislators deserve credit for getting this far, and getting there so quickly. Notably, the bill does not expand the reach or scope of the current system. But it does recognize that child safety is a core government function and one in which there must be far greater accountability. As Naomi Schaefer Riley of the American Enterprise Institute recently argued: “To be clear, government does not need to interfere in the lives of more children, it needs to interfere more in the lives of children who are already coming to its attention.”
But while HB 1144 — which would create regional teams to oversee particularly complex child abuse and neglect cases — makes good sense, the mechanism by which escalations would occur could be better. As currently drafted, the bill proposes a discrete checklist of specific risk factors such as “Involvement of a medically complex juvenile who requires subspecialty care by two or more specialties” and ”Families that have three or more prior child protective services reports with concerns for domestic violence or active or current drug activity.”
Versions of this checklist approach have been in use by child welfare agencies for decades, including in North Carolina. The research is clear: Even if the selection of escalated cases based on isolated factors can be automated, we always lose important information when we rely on simple list-based protocols.
As the federal Administration for Children and Families recently noted in a research brief on PRM: “Because these models pull information directly from existing records, they can incorporate hundreds of data elements and update in real time as safety and other information changes.” PRM is precisely the mechanism HB 1144 needs — and could also address longstanding intake screening challenges documented in a 2019 report to the North Carolina General Assembly.
The evidence from other states is compelling. A randomized controlled study in Colorado documents reductions in hospitalizations for child injuries. Data from Los Angeles County show that when supervisors and managers are immediately alerted to high-risk investigations through a PRM, maltreatment near-fatalities and fatalities decline. For those concerned about cost or feasibility, Idaho’s statewide implementation demonstrates these models need not take years to build or millions of dollars to deploy. And concerns about racial bias have not borne out. Allegheny County’s experience in Pennsylvania shows PRMs actually narrow racial disparities, guarding against the tendency to mistake race for risk.
Without modern decision-support tools, these patterns can remain invisible until a situation escalates into serious harm, a near fatality, or worse. We will never be able to prevent all child deaths. But we know more than we ever have about how to identify the children most at risk. Dominique Moody’s death has given North Carolina both the reason and the opportunity to put that knowledge to work. The only question is whether we will.
“Predictive risk modeling would improve Dominique Moody Safety Act” was originally published on www.carolinajournal.com.
