A Comparative Study of the Use and Effectiveness of Different Risk Assessment Models in CPS Decision Making Processes

Dataset Number: 84
Investigator(s): Baird, C.
Abstract: A Comparative Study of the Use and Effectiveness of Different Risk Assessment Models in CPS Decision Making Processes was a three-year project conducted by the Children's Research Center and funded by the National Center on Child Abuse and Neglect. The central objective of this study was to determine the relative validity and reliability of three different risk assessment systems used in Child Protective Services (CPS). In addition to comparing system validity and reliability, this research was intended to: 1) identify correlates of child maltreatment in each jurisdiction studied; 2) determine the degree to which risk assessment models are transferable among jurisdictions; 3) determine the extent to which site-specific risk assessment systems might offer improvements over existing systems; 4) determine the degree to which worker characteristics are related to case outcomes; 5) determine if services have a significant effect on outcomes; and 6) determine which risk factors are valid across all sites, and which have little or no relationship to outcomes.
   The three risk assessment instruments used in this study included the Washington Risk Assessment Matrix, the California Family Assessment Factor Analysis, and the Michigan Structured Decision Making System's Family Risk Assessment of Abuse and Neglect. Both the Washington and the California models are consensus-based assessment systems, while the Michigan model is an actuarial instrument developed from empirical studies of child maltreatment.
   Data were collected from four project jurisdictions: Alameda county, California; Dade county, Florida; Jackson county, Missouri; and Macomb, Muskegon, Ottawa, and Wayne counties in Michigan. Due to complexities in gathering the data, the vast majority of cases in this study came from urban settings. In total, 1,400 families constituted the validity study sample. All were investigated in the fall of 1995 or early 1996. The number of cases for which each risk assessment was actually completed, however, ranged from 1,335 to 1,396 (Baird and Wagner, 2000). Outcome information was collected over an 18-month follow-up period, including rates of subsequent investigations, substantiations, and placements in out-of-home care. The sample for the reliability component of the research consists of 20 cases per site - ten substantiated and ten unsubstantiated cases (except in the case of Michigan where 20 substantiated cases were chosen for reliability testing) - for a grand total of 80 reliability study sample cases.
Download User's Guide (Data Codebook): PDF

  Back to List

PDF (Portable Document Format) files can be viewed and printed with Adobe Reader, a free utility available at the Adobe Web Site.