Do Upstream Family Supports Reduce CPS Involvement? Evidence from a Quasi Experimental Study

Background and Purpose

Community‑based, voluntary program models that strengthen family protective factors represent a promising approach to child maltreatment prevention. However, rigorous evidence of effectiveness remains limited, partly due to a range of ethical and practical constraints related to experimental research designs. In 2021, the Ohio Children’s Trust Fund launched the Family Success Network (FSN) through the federal Child Welfare Community Collaborations (CWCC) grant as a strategy to reduce child maltreatment risk and child protective services (CPS) involvement in three pilot counties. FSN provides families with a single access point to support including family coaching, parenting education, financial literacy, and concrete resources. This study evaluated whether high levels of FSN utilization at the community level were associated with changes in CPS outcomes using a quasi-experimental approach.

Methods

We conducted a difference‑in‑differences (DiD) analysis using zip‑code–level monthly administrative data provided by the Ohio Department of Children and Youth from January 2017 through June 2025. The study population included families with children under age 18 residing in 1,240 Ohio zip codes. The intervention group consisted of 19 zip codes in the three pilot counties with high FSN utilization, defined as instances in which FSN interactions exceeded 0.05% of the 2021 population of families with children. Zip codes with minimal exposure and those affected by later FSN expansion were excluded. The comparison group included 1,002 Ohio zip codes with no FSN implementation. To mitigate potential baseline differences between intervention and comparison areas, we applied inverse probability of treatment weighting (IPTW) based on propensity scores derived from demographic, socioeconomic, and pre‑intervention child welfare indicators. DiD models incorporating IPTW were estimated using generalized estimating equations with standard errors clustered at the zip‑code level.

Primary outcomes included monthly counts of screened‑in CPS referrals, substantiated maltreatment reports, and removals to foster care. Models were adjusted for zip‑code–level sociodemographic characteristics using 5-year estimates from the 2021 ZIP Code Tabulation Areas (ZCTAs) obtained from the Census Bureau’s American Community Service Data (U.S. Census Bureau, 2023), county CPS indicators, and county‑level economic conditions.

Results

A visual inspection of plots for the pretreatment period (2017-2021), suggested parallel trends in average mean intakes and substantiated cases for high FSN and comparison non-FSN zip, conditional on the inverse probability weights. On average, controlling for zip code- and county-level factors, high FSN utilization decreased the number of intakes by 13% (IRR: .87 (95% CI: .83, .93) and were associated with a slight but non-statistically significant decrease in the number of substantiated reports (IRR: .95 (95% CI: .88, 1.03).

Conclusions and Implications

The study demonstrates the feasibility of using administrative data, quasi‑experimental designs, and causal inference methods to evaluate community‑wide interventions under real‑world conditions. Findings provide evidence that a voluntary, family‑centered prevention model can reduce CPS involvement at scale, highlighting the potential of FSN and similar initiatives to strengthen upstream prevention. Study limitations and implications for additional research, such as the need to disaggregate effects to determine for whom and under what conditions the program is most effective, are discussed.