![[ Visit Client Website ]](images/banner.gif)
Establishing a comprehensive historical database that accurately captures longitudinal sequences of nationwide child maltreatment reports (CMRs) and foster care (FC) cases represents a crucial advancement in child maltreatment research in the United States. Currently, national CMR and FC records are maintained separately by two systems: the National Child Abuse and Neglect Data System and the Adoption and Foster Care Analysis and Reporting System. The RAPIDS project has amalgamated these systems into a unified longitudinal dataset, overcoming existing data limitations that hindered linkage and enhancing usability. RAPIDS data contain child/event-level records of all CMRs and FC entry-to-exit episodes from 2006-2021. RAPIDS also includes identification of sibling groups based on network analysis, county-level census data, and state policy data. The present study illustrates the benefits of RAPIDS data in predicting current and future CMR outcomes.
Methods:
We selected all US children aged ≤10 with CMRs (index CMRs) in 2018 (N=2,371,119). The age limit was set to ensure comprehensive coverage of their records from birth onwards with no left-censoring. We assessed five CMR outcomes: two current outcomes (substantiation and FC entry immediately following the index CMR) and three future outcomes (repeat-CMR, substantiated repeat-CMR, and future FC entry within two years from the index CMR). Using logistic regression, we examined improvements in predicting each of these outcomes by sequentially integrating six groups of predictors derived from: (1) index CMR records, (2) prior CMR records, (3) FC records, (4) siblings’ records, (5) county-level census data, and (6) state policy data.
Results:
For all five outcomes, significant improvements in model fit were observed with each stepwise addition of the six groups of predictors (both AIC and BIC values reduced by >10, and likelihood ratio tests showed p-vales<.0001). We further examined the relative contributions of predictors by assessing the increase in the area under the curve (∆AUC). Current index CMR records exhibited the greatest contributions to current outcomes (substantiation [∆AUC=.2466] and FC entry [∆AUC=.2267]), while prior CMR records played a larger role in future outcomes (repeat-CMR [∆AUC=.0722], substantiated repeat-CMR [∆AUC=.0613], and future FC entry [∆AUC=.0658]). Predictors from FC records were notably impactful for FC outcomes (FC entry [∆AUC=.0121] and future FC entry [∆AUC=.0052]). Using siblings’ records to predict outcomes yielded the greatest benefits for younger ages in terms of increases in AUC. Finally, the contributions of county-level data (county-level poverty and urbanicity with ∆AUC=.0022-.0061) and state policy data (state CPS administration, mandated reporting laws, alternative responses, and substantiation evidence levels with ∆AUC=.0008-.0075) were significant.
Conclusions/Implications:
Our findings highlight the advantages of RAPIDS data in predicting current and future CMR outcomes, underscoring the importance of children’s longitudinal history for outcome prediction. Additionally, despite the youngest group facing the highest CMR/FC risk, their limited life histories pose challenges in predicting their future outcomes. Our findings suggest that leveraging the records of their older siblings can effectively address this challenge. The significant contributions of county-level census data and state policy data underscore the importance of connecting CMR/FC records with external sources.