In a recently released article in Pediatrics(10.1542/peds.2017-2882), Dr. Rhema Vaithianathan and colleagues from New Zealand ask if administrative data available at birth can predict not just risk of substantiated Child Protective services (CPS) involvement, but also risk of injury, hospitalization and death. The question is meaningful because it can assist organizations in deciding how broadly or narrowly to target preventive services for high-risk families. The authors’ dataset is likely the envy of many researchers – it includes every live born infant in New Zealand and is “…linked to health, welfare benefits, CPS, and criminal justice registers (among others)…,” includes maternal data and all child hospitalizations, and is estimated to cover 94% of the birth cohort. The authors studied infants born in 2010 and 2011, which gave a cohort of more than 60,000 children per year. They first incorporated multiple variables in a logistic regression to predict risk of substantiated child maltreatment by age 2 years, and assigned the top 10% as very high risk and the top 20% as high risk. The model was powerful, and the results were at the same time both shocking and not surprising: children who were at very high risk (top 10%) were 4.8 times more likely to die in infancy, twice as likely to be hospitalized, and accounted for 57.1% of all accidental deaths in the cohort. The details for both the very high-risk and the high-risk groups are important and I hope you will delve into them. The authors carefully and clearly explain the limitations of their work, which is helpful since it is otherwise easy to feel that their results constitute a “slam dunk”.
I struggled with several issues while reading this powerful article. Is use of administrative data an ethical and just approach to identifying who receives intensive family support services? Regarding services that are either moderately intrusive or require family engagement, such as parental mental health care and home visiting, families might reasonably ask, “why me”? We need a good answer to this question. If the services are voluntary, might the highest risk families simply opt out? The authors report that 53.8% of very high-risk children (versus 10.1% of all others) had a mother who was reported to CPS during her own childhood. Could this use of a large administrative database to identify very high-risk children essentially institutionalize ascertainment bias? I believe there are approaches to some of these dilemmas. Adding less personal parameters to the determination for service receipt, such as zip code or census tract, could serve to de-stigmatize the “assignment” of services. Families could also be offered a menu of support services from which to choose, along with help in deciding which would be optimal for their own needs. Finally, as with any social service, a healthy dose of cultural humility will be essential for acceptance. While New Zealand offers a unique “laboratory” for this approach, the potential for saving lives suggests that we need to figure out how to adopt and adapt this important predictive model for use in the United States.