Abstract: Session 1: The Report and Placement Integrated Data System (RAPIDS) (Society for Social Work and Research 29th Annual Conference)

Schedule:

Friday, January 17, 2025

Aspen, Level 2 (Sheraton Grand Seattle)

* noted as presenting author

Brett Drake, MSW, PhD, Professor of Data Science for the Social Good in Practice, Washington University in Saint Louis, St. Louis, MO

Hyunil Kim, PhD, Assistant Professor, University of Illinois at Urbana-Champaign, IL

Rebecca Orsi-Hunt, PhD, Associate Research Professor, University of Colorado, Anschutz Medical Campus, Aurora, CO

John Fluke, PhD, Professor, University of Colorado, Aurora, CO

Dylan Jones, MSW, PhD Student, Washington University in St. Louis, St Louis, MO

Background/Significance:

The National Data Archive on Child Abuse and Neglect (NDACAN) includes national, near-universal data on hotline calls in the National Child Abuse and Neglect Data System (NCANDS) Child File. The Adoption and Foster Care Analysis and Reporting System (AFCARS) includes near-universal information on each child who contacted the Foster Care system. These data are available for most states in a linkable fashion back through 2006.

These data exist in yearly files and can be combined. The Child File data are stored in a user-friendly “one record per child per report” format. Unfortunately, the AFCARS data are stored in a “one record per child who contacted the foster care system per year” format. This means that individual AFCARS files may contain one entire foster-care experience, a fragment of an experience, or multiple experiences, whole or fragmentary. In addition, when the data files are merged, paradoxes and inconsistencies occur at an infrequent but troublesome level. While single-year linkages of AFCARS and the Child File have commonly been done, there have been no national long term (e.g. decade or more) uses of these data due, we believe, to the difficulties in integrating these datasets longitudinally.

Methods:

RAPIDS is a suite of programs written in SAS between 2017 and 2024 which resolves these problems. It is currently free for others to use and is available on LDBase. It was written by a consortium from the George Warren Brown School of Social Work at Washington University in St. Louis, the University of Illinois School of Social Work and the Kempe Center. The most complex part of RAPIDS involves combining AFCARS records and shifting them from a “one record per child per year” format to a “one record per full foster care episode” format. In addition to addressing these issues, RAPIDS also includes variables drawn from other data sources, such as Census data or state policy data. Finally, useful “value adds” are incorporated, for example, we use social network analysis to provide a new “Family ID”, and we include temporal variables for each record, such as prior and next child welfare and foster care placement dates (if any).

Results:

The final RAPIDS dataset spans the 2006-2021 timeframe and includes over 60 million records. RAPIDS is now publicly available on LDbase, accompanied by a codebook. This formalization has been made possible by the NICHD-supported Child Innovation in Child Maltreatment Center (IP50HD096719) and ACF NAICS funding (90FA3001). This presentation will introduce participants to RAPIDS, including how to obtain it from LDBase and use it.

Conclusion/Implications:

RAPIDS provides an off-the-shelf and very fast tool for building a longitudinal child welfare database which alpha testing has shown reduces coder burden for research projects by over 95%. Speed of data access is also reduced, for most projects, by a factor of 95-99 times. Using RAPIDS, previously prohibitive longitudinal analyses are relatively straightforward. With periodic updates, we hope that RAPIDS will serve the research community as a resource for child welfare system analyses for decades to come.