The Society for Social Work and Research

2014 Annual Conference

January 15-19, 2014 I Grand Hyatt San Antonio I San Antonio, TX

Optimizing Differential Response Thresholds in Child Protection: How Much Is Enough?

Thursday, January 16, 2014: 1:30 PM
HBG Convention Center, Room 001B River Level (San Antonio, TX)
* noted as presenting author
Adam Darnell, PhD, Senior Research Associate, Casey Family Programs, Seattle, WA
John D. Fluke, PhD, Associate Director for Systems Research and Evaluation, University of Colorado School of Medicine, Aurora, CO
Katherine Casillas, PhD, Assistant Professor, University of Colorado, Denver -- Anschutz Medical Campus, Denver, CO
Background and Purpose

A critical component of differential response (DR) systems is the risk threshold that determines whether families are assigned to investigation response (IR) or an alternative response (AR) at Child Protective Services (CPS) intake.  States and counties vary widely in both policy and implementation of criteria for referral to the non-investigation pathway, including the possibility of ‘drift’ from the intended AR population.  Research is needed to describe differences in criteria for and utilization of AR and examine how these differences relate to child safety.  This study uses data from the National Child Abuse and Neglect Data System (NCANDS) to examine the relationship between county rates of AR utilization and child safety.  We expected that as the risk threshold for accepting cases into AR increases, a larger proportion of higher risk cases would be included in the AR caseload.  Thus we hypothesized that counties with greater utilization of AR, as a proxy for a lower risk threshold for AR, would have reduced child safety outcomes.


The NCANDS child data files were obtained for the years 2005 through 2011.  These states were selected because they had active DR systems statewide over these years and also represented both county-administered and state-administered systems. 

AR utilization was computed as the number of responses receiving an AR disposition, as a percentage of all annual responses. Child safety was operationalized as six-month rates of repeat reports and substantiated repeat reports. Both of these variables were computed separately for cases receiving AR on their initial report and cases receiving an IR on their initial report.  The arithmetic difference between re-report rates for cases receiving AR versus IR in their initial response was also computed. 

Besides AR utilization, covariates representing county population size, demographics, and socioeconomic factors were also included. The relationship between AR utilization and child safety variables, controlling for community contextual factors, was examined using Generalized Linear Models to specify Poisson regression.


Results of descriptive analysis indicated marked variability between states and counties in AR utilization rates as well as substantial increases in AR utilization over time for most jurisdictions.  Results of primary analyses supported the hypothesized positive relationship between AR utilization and re-reporting, and strong positive relationships between AR utilization and the AR-IR safety difference indicator. Further ancillary analysis suggested that county-level variation in AR re-reporting may also be a function of county racial/ethnic composition where greater levels of persons of color were associated with higher re-reporting, controlling for county poverty levels.


While limited to the experience of four states, the results are quite robust suggesting that the risk pool of CPS cases for re-reporting between AR and IR may not be different and is not being systematically distinguished at CPS intake. Optimal levels of AR utilization based on re-reporting risk are therefore likely to remain discretionary to local and state policy depending on how much re-reporting among AR cases policymakers are able to accept.