Geospatial Variation of CPS Reports of School Aged Children

Background & Purpose:

Child protective services (CPS) agencies are reliant of calls regarding child maltreatment concerns to be made to them. The majority of CPS reports come from mandated reporters who are professionals who regularly interact with children. This is particularly true for school-aged children who have more interactions with mandated reporters than younger children. However, there is limited information if there is variation in these reports rates by community and reporter type. The purpose of this study was to examine the relationship of CPS report rates for school-aged children by reporter type with community-level characteristics and to evaluate if there was geospatial heterogeneity among these relationships.

Methods:

We analyzed CPS administrative data from California (2018-2022) for children aged 5-17. We calculated rates of CPS reports overall and by reporter type (education, law enforcement, medical, non-mandated) at the census tract-level per 1,000 children in the population. These CPS report rates were linked to Healthy Places Index (HPI) census tract-level scores regarding education, economic, transportation, housing, insurance access, and clean environment. First, we ran negative binomial regression models to assess the relationships between report rates and a census tract’s HPI scores, race/ethnicity composition, and urbanicity producing incident rate ratios (IRR). We assessed the data for spatial autocorrelation by calculating Moran’s I statistic. To account for the identified spatial autocorrelation and to test for community-level heterogeneity by producing census tract-level coefficient estimates, we ran geographically weighted regression (GWR) models for each of the report rates. We compared model fits using Akaike Information Criterion (AIC).

Results:

CPS report rates varied greatly by census tract and reporter type. Overall report rates had a negative relationship with better economic scores (IRR: 0.67, 95% CI: 0.65-0.68) and a positive relationship with higher percentages of non-Hispanic White residents (IRR: 2.11, 95% CI: 1.97-2.25). The direction of these relationships held across each of the reporter types. In contrast, relationships with housing and urbanicity varied among reporter types. For example, medical reporter report rates had a statistically significant positive relationship with tract urbanicity (IRR: 1.19, 95% CI: 1.10-1.28) and education report rates had a negative relationship (IRR: 0.86, 95% CI: 0.82-0.92). Spatial autocorrelation was identified for each CPS report rate (all reports, Moran’s I = 0.452, p < .001). AIC indicated that GWR models were better fits across reporter type models. GRW identified spatial heterogeneity across the model coefficients.

Conclusion & Implications:

CPS report rates vary across communities, and this is consistent when disaggregated by reporter type. These variations have relationships with community characteristics, such as economics, housing, and urbanicity, but there is heterogeneity among these relationships by reporter type and by geography. These findings suggest that prevention efforts and mandated reporting trainings may need to be customized for different communities and types of mandated reporters. Further research is necessary to understand the geospatial variation of child maltreatment and child protection responses.