Emergency Department Physician Reporting of Suspected Child Maltreatment: Effects on Child Safety and Wellbeing

• In Event: Mandatory Reporting, Institutional Variation, and Child Wellbeing: Novel Evidence from the Child Welfare System

Abstract

Medical providers represent the primary source of reports of suspected child maltreatment to child protective services (CPS) among young children prior to school entry. Prior literature documents wide variation in reporting policy and behavior, concluding that more targeted surveillance of children on behalf of medical providers could prevent later crises that necessitate removal. Yet, remarkably little is known about the causal effect of physician reporting to CPS. This study leverages the fact that children are quasi-randomly assigned to Emergency Department (ED) to physicians who vary widely in terms of their tendency to report suspected maltreatment to CPS.  To answer this question, we link administrative records on the universe of ED visits for Medicaid-enrolled children ages 0-17 that occurred between 2010 and 2023 to child welfare data from the Wisconsin Administrative Data Core. As reported children systematically differ from non-reported children in ways that are unobservable in our data, we harness two-stage least squares, instrumenting for reporting with the leave-one-out reporting tendency of the ED physician observed in Medicaid claims. Our first stage estimates indicate that our instrument is strong and relevant (70-90 first stage F-stat) and robust to a range of exclusion restrictions, minimum reporting thresholds, outcome periods, and across multiple subsamples of children. Our results indicate that children who are on the margin of being reported exhibit a substantial increase in the risk of having a substantiated investigation and of being placed in foster care within 60 days of the focal ED visit (20 and 14 pp, respectively). We additionally examine subsequent patterns of injuries, healthcare utilization, and mortality. To investigate racial disproportionality in reporting, we compare report rates of non-white to White children who share (1) high-sensitivity/high-specificity injuries that are commonly flagged for Child Abuse Pediatric team investigations, and (2) fall in the same quintile of predicted maltreatment propensity based on historical records of the child and of their caregivers. We conduct the former in consultation with an ED physician and for the latter, leverage gradient boosting machine learning methods optimized to predict relatively rare outcomes. Our results physician reporting in CPS involvement and provide novel insight into the costs and benefits of mandatory reporting policies.

Author

• Jessica Erin Pac, University of Wisconsin-Madison
  Presenting Author