Profiling of county-level foster care placements using random-effects Poisson regression models

Concern regarding the removal of African American children from their homes in McLean county and Peoria County Illinois, and placing them into Foster care have been the center of considerable debate and concern in the state of Illinois. As a result, a county-level (McLean and Peoria counties) investigation and intervention was performed. To evaluate the success of this intervention, we developed a mixed-effects Poisson regression model for the analysis of these data, and used it to obtain case-mix adjusted empirical Bayes estimates of county-specific changes (2000–2002) in Foster care placement rates in all counties in the state of Illinois. Results of the analysis revealed that four out of the 85 counties with African American residents in the state of Illinois exhibited significant decreases in Foster care placement rates relative to the change in the overall state-wide rate between 2000 and 2002. The two counties that received the intervention (McLean and Peoria) were among the four counties out of 85 counties that exhibited significant decreases in Foster care placement rates (OR = 14.7, P < 0.0001). Conceptual and statistical aspects of this type of statistical profiling are presented and discussed.