We will build out the comprehensive infrastructure (data integration, research products, training, and policy relevant recommendations) which is required to support informed and open-source use of the new US Report And Placement Integrated Data System (RAPIDS). RAPIDS is a program (data construction algorithm) which can be used to transform existing NCANDS child maltreatment (CM) report and AFCARS data into a single, user-friendly dataset utilizable by researchers with NDACAN data use agreements. We will also build and test linkages to policy and census data files to enhance policy relevant analyses of CM incidence. The proposed “Open-Source Approach” project responds to specific goals of the Secondary Analyses of Data on the National Incidence of Child Maltreatment (HHS-2022-ACF-OPRE-FA-0149) call to aid collaboration among CM researchers and policy makers. This project is a collaboration among several prominent CM researchers. Project outcomes will significantly expand networking and collaboration among CM researchers by enabling them to analyze and use RAPIDS data enhanced with policy and community data. Analyses and testing of the data will also address several specific and timely CM incidence research questions.
The RAPIDS algorithms and the resulting dataset have been under development by a consortium of researchers at Washington University in St. Louis, University of Illinois Urbana-Champaign and University of Colorado School of Medicine for five years. RAPIDS data construction algorithms combine the National Child Abuse and Neglect Data System (NCANDS) and the Adoption and Foster Care Reporting System (AFCARS) data from 2005 forward into a single, linked longitudinal dataset that combines records at the child/event level for both Child File CM reports (one record per referral per child) and AFCARS foster care (FC) spells (one record per foster care episode [entry-to-exit] per child) while preserving temporal information between events. Having a historical database that accurately captures longitudinal sequences of nationwide records of all CM reports and FC cases over a decade and a half is a critical advance which will accellerate science in the area of CM surveillance in the US. Our tested algorithms also place children in sibling groups based on network analysis. RAPIDS data, in addition to the new linkages to policy and census information, are a significant improvement over the raw NCANDS and AFCARS data. A child’s entire lifetime trajectory (during the 2005-2019 timeframe) of CM reports and FC spells is available using RAPIDS. These linked data have the potential to support a wide range of relevant research questions.
Our “Open-Source Approach” objectives are: (1) build comprehensive user infrastructure for RAPIDS (e.g. fully documented algorithms, a user’s guide); (2) conduct and disseminate two RAPIDS analyses for research questions of relevance; these analyses include census data linkage and will form the basis for developing training materials; (3) determine and summarize the feasibility of integrating RAPIDS, CDC, the National Youth in Transition Database (NYTD), and other data sources including policies; (4) create training materials and conduct in-person user training workshop in 2024; and (5) develop broad recommendations to support future research.
RAPIDS capabilities are shown by our team’s initial study published in Child Maltreatment in which we found that surprising numbers of children in out-of-home care in child protective services (CPS) systems lack a preceding CM allegation, with large levels of state variability.1 We have also presented these early results at several conferences in 2021 and 2022.
The RAPIDS programs facilitate the creation of two primary types of datasets:
1. National individual-level longitudinal data: Compiling report records and foster care records spanning from 2005 to 2021.
2. National county-level longitudinal data: Linking report records and census data covering the period from 2007 to 2019.
In addition, the RAPIDS programs offer several add-ons to enhance the basic national individual-level longitudinal data. These include:
1. Analysis-ready dataset construction: Incorporating prior history variables and future event variables to create a dataset ready for statistical analysis.
2. Census data integration: Adding county-level variables based on residential counties at the time of the event (report, foster care entry, or foster care exit).
3. Family data augmentation: Enriching the dataset with information pertaining to a child's siblings, such as demographic characteristics, current event variables, and prior history variables specific to each individual child's siblings.
On this site, we will provide the RAPIDS programs, accompanied by pertinent documentation such as user guides and codebooks (anticipated release in March 2024).