study definition of a cold case. To begin the development of a predictive model, 65 potential variables in the AFCARS data were examined for their relationship to the outcome of interest: turning 18 while in care, exiting without permanency with all parental rights terminated, or emancipating after a foster care stay longer than three years. Using multivariate logistic regression techniques, a final list of seven variables was identified that predict case outcomes (“cold” vs. not). Those seven predictors of a cold case, in order from the most influential to the least, are presented in Table 1.
The final model uses all seven variables as well as their two-way interactions (see Appendix A for complete statistical documentation of the multivariate logistic regression cold case predictive model). In other words, there is an added benefit in accurately predicting case outcomes knowing whether a case had both a TPR and was eligible for federal funding. Applying this model to the 2007 AFCARS cases accurately predicted the outcome in 90% of cases. The model was also validated on 2006 AFCARS data, achieving a prediction accuracy of 89%.
Table 1. Seven Predictors of a Cold Case
Lack of federal funding reimbursement
Number of months in care since the current removal
Lack of termination of parental rights
Caretaker (in the removal home) year of birth
Current placement in an institution
Age of the child (on 3/31/09)
Number of placement settings in the current removal
Finally, the model was used to predict CFSR composite #3 outcomes for current foster care cases that had been ongoing for at least 24 months on March 31, 2009. That resulted in a ranking of current foster care cases in terms of their likelihood of becoming a cold case. Next, the 500 “coldest” foster care cases in Georgia on March 31, 2009 (cases with the highest likelihood) were selected for file review. Of the list of 500, half would be randomly assigned (within county) for review and half would serve as comparison cases (not reviewed).
Supreme Court Fellows
During the second month of the project, Supreme Court Fellows (lawyers with expertise in child welfare) were recruited and hired to review the sample of cold cases described above. Experts were recruited by posting notices on various child welfare list serves across the state. Interested candidates were directed to a website for fellowship details and application instructions. Resume reviews and interviews were conducted by the Project Director and staff from the Administrative Office of the Courts of Georgia (AOC).
AOC staff developed and administered an eight question test to all potential candidates for the purpose of assessing their knowledge of state and federal child welfare issues. Interviews were scheduled with the highest scoring candidates. Phone interviews were conducted with candidates during the first week of March. On March 12, 2009 eleven candidates were hired as Supreme Court Fellows. Two Fellows would split a fellowship with each working part-time, and one was selected to serve as the Project Lead. See Table 2.
The Fellows were a mix of Special Assistant Attorneys General (SAAG), who serve as the Social Service Agency’s attorney, and private attorneys. Each Fellow agreed to dedicate ten hours per week (five hours for part-time) to project tasks, including: file reviews, interviews with DFCS representatives, completing data collection forms, writing
Table 2. Cold Case Project Supreme Court Fellows
Patricia Ketch Buonodono Melinda Cowan Rachel Davidson (part-time) Darice Good (part-time) Karlise Grier Diana Rugh-Johnson Trân Lankford Dorothy Murphy Brooke Silverthorn Leslie Stewart Ashley Willcott, Project Lead