How Might the Effects of Evidence-Based Child Welfare Policies Implemented Differ Across States? A Combined Analysis Using Large-Scale Administrative Data Analyses and Computer Simulation Models

Introduction: The U.S. Child Welfare system serves over 1 million children at a cost of $20 billion annually. Evidence-based interventions have the potential to improve outcomes for this large, vulnerable population. Given the substantial investments necessary, Child Welfare agencies require evidence to support such decisions and must satisfy themselves that studies conducted in other settings provide evidence relevant to their own settings.

Methods: We constructed a microsimulation of all children in all counties from two U.S. states, focusing on time spent in the respective states’ Child Welfare systems. Children were characterized by age, race/ethnicity, and detailed Child Welfare placement status and history. The model tracked measures of stated Child Welfare system goals: placement stability, permanence, total sizes of the county, and state Child Welfare populations over time and their budget implications. We used complete Child Welfare data from the Multistate Foster Care Data Archive to estimate rates of entry into the system, movements between foster placements, rates of exit and reentry. To establish at-risk population sizes, birth and death rates, we used data from the CDC’s WONDER database, U.S. Census Intercensal County Population Data and the Kaiser Family Foundation’s State Health Facts database. To illustrate how outcomes differ with the implementation of an evidence-based Child Welfare intervention, we repeated our simulations using information on an evidence-based foster parent training intervention (KEEP) from a randomized controlled trial.

Findings: The microsimulation accurately replicated past Child Welfare trends in terms of system contacts as well as placement stability. It also captured substantial state and county-level heterogeneity in outcomes, accounted only partially by the demographic and race/ethnicity mixes in the counties. The implementation of KEEP increased permanence and placement stability in all systems for children targeted by the program. Because the size of these improvements varied substantially between state and county systems, the relative priority of adopting the intervention may likewise differ, an important insight for Child Welfare decision makers but one requiring further modeling studies of the relative benefits of alternative interventions within each system.

Conclusion: We demonstrate methods to employ existing, large-scale administrative data, controlled studies of evidence-based practices, and microsimulation models to project policy-relevant Child Welfare outcomes. Such methods when fully developed could be valuable and feasible in supporting Child Welfare-decision makers considering multiple interventions at outcomes in the contexts of their individual systems.