Risk-Targeted Services and Substantiated Re-Reports: A Propensity Score Matching Study of Poverty and Behavioral Health Services In Child Welfare

Background and Purpose

Child neglect affects approximately 79% of substantiated or indicated maltreatment victims in the U.S., making it the most prevalent form of child maltreatment, and is associated with impaired child development, and family instability. CPS agencies provide targeted services based on identified family risk factors, aiming to address these risks and prevent maltreatment recurrence, however, the extent to which such services reduce substantiated re-reports remains critically understudied, particularly using rigorous methods to address selection bias with national data.

Methods

Using the National Child Abuse and Neglect Data System (NCANDS, 2022–2024), we identified each child's first substantiated neglect case in 2022 (N=431,614). Two analytic subsamples were constructed among families with identified risk factors who received any CPS services: a poverty-risk group (n=33,598, 26 states) and a behavioral health (BH)-risk group (n=41,363, 31 states). States were included only if service reporting quality met a 70% threshold. Within each group, we compared families who received risk-matched targeted services to those who received other CPS services but not the targeted type, isolating the effect of service-need match rather than service receipt per se. Propensity score matching (PSM, caliper=0.05) was applied within each subsample, balancing on child age, gender, race/ethnicity, prior victimization, domestic violence, maltreatment type, and risk-relevant caregiver factors. Two outcomes were examined: neglect-specific re-reports and all-type re-reports, tracked over a 365-day follow-up period. Multiple sensitivity analyses were conducted for both the poverty and BH risk groups to assess robustness.

Results

In the poverty-risk group, receiving poverty-targeted services (employment, education, housing) was associated with a significantly higher likelihood of neglect-specific re-reports (AME=+0.037, p=0.029), and marginally higher for all-type re-reports (AME=+0.033, p=0.082), with consistent findings across all sensitivity analyses. In the BH-risk group, receiving BH-targeted services (mental health, substance abuse, counseling) was not associated with the likelihood of neglect-specific re-reports (AME=−0.016, p=0.435) and all-type re-reports (AME=−0.015, p=0.433). Though neither was statistically significant; sensitivity analyses displayed directional consistency.

Conclusions and Implications

These divergent findings suggest that service-need matching may operate through different mechanisms depending on service delivery setting and population risk profile. Possible explanations include: first, increased system visibility among families receiving poverty-targeted services, which may elevate re-report detection rather than reflect worse outcomes; second, insufficient service intensity (e.g., dosage, duration) to sufficiently reduce underlying risk. The two explanations carry distinct policy implications and warrant further investigation. These findings also raise a measurement concern: re-report rate may capture increased detection rather than true maltreatment recurrence, limiting its interpretability as a service effectiveness outcome. Future research should incorporate service dosage, contact frequency, and report source data to disentangle these mechanisms.