(1) Does the launch of DR implementation correspond with a reduction of CPS investigations over time?
(2) Does the launch DR also correspond with an increase in the proportion of investigations that are substantiated over time?
Methods: The study used data from the National Child Abuse and Neglect Data System (NCANDS) child files for 2000-2010 for 66 counties that launched DR between 2000-2010 (representing 15 states). The author conducted independent data collection in 2013 to verify county-level differential response implementation history in the United States since 1998. The dependent variables of interest were the rate of neglect investigations within a county’s child population and the rate substantiation among a county’s investigated neglect cases. The rates were constructed by aggregating child-level maltreatment records by county. The analysis included two phases. First, the unadjusted means of investigation and substantiation rates were plotted by measuring time in two ways, calendar years and length of time before and after DR implementation. These plots informed the construction of inferential piecewise mixed-effect models to test the effects of DR on investigation and substantiation rates. The piecewise models employed contrast tests to detect whether differences in slopes correspond with the launch of DR. Calendar years were included in the models to control for potential confounding history effects.
Results: Results from both the descriptive trend plot and piecewise model suggested a reduction in investigation rates levels, with a steep decline within the first three years after the launch of DR. Specifically, the piecewise model indicated significant linear and quadratic effects of DR implementation on CPS investigation rates (p = .03 and p=.006 respectively) and the omnibus contrast tests suggested significantly lower investigation rates post-DR implementation (p=.002). Conversely, the trend analysis suggested substantiation rates remained unchanged after DR implementation. The inferential piecewise model showed only marginally significant effects (p=.08 for the omnibus contrast test). Post-hoc exploratory analysis revealed that compared to counties that never implemented DR, DR-counties had significantly higher rates of substantiation even prior to the launch of DR.
Conclusion and Implications: This study comports with other literature that has suggested that DR reduces CPS investigations. The unexpected null results pertaining to the effects of DR on substantiation rates conflict with prior evidence and suggest the need for more exploration. The findings also emphasize the strength of longitudinal studies to assess the impact of large-scale policy changes.