Children and Family Research Center/University of Illinois at Urbana-Champaign
The way states submit data to the federal Adoption and Foster Care Analysis and Reporting System (AFCARS) seriously handicaps efforts to track children from foster care entry to exit and distorts assessments of the states’ performance, according to this report.
The study began when researchers wondered why Illinois – which has been touted for reforms that led its foster care population to drop by two-thirds from 1997 to 2005, while its permanent placements rose significantly – showed up on a list of states that failed to meet all seven standards of the federal Child and Family Services Reviews (CFSRs). Many CFSR standard measures are based on AFCARS data.
This yearlong investigation, funded by the Pew Charitable Trusts, found that when researchers applied alternative analytical methods to AFCARS data for Arizona, Illinois, Missouri, Ohio and Wisconsin, those states did better on some measures than either AFCARS data or the state’s CFSR indicated.
AFCARS, which began in 1993, was designed to generate point-in-time and retroactive case counts for foster care entries and exits. The design did not allow for tracking the progress of cases over time. When CFSRs began in 1997, the U.S. Department of Health and Human Services had to rely on the “statistical snapshots” taken at the end of AFCARS reporting periods, according to the study.
For example, CFSR end-of-quarter caseload measurements are based on a cross-section of a state’s foster care population at the start of a quarter. The number of children who entered care that quarter is added, and the number of children who exited care that quarter is subtracted. However, any data drawn from this end-of-quarter caseload (such as median length of stay) are more heavily “biased toward the experiences of children with the least satisfactory permanency outcomes,” the researchers write. That’s because end-of-quarter caseloads include a preponderance of backlogged cases, but exclude cases that have entered and exited the system in the most timely way – within the span of the quarter.
Looking only at children exiting the system would not be an adequate solution to this problem, they write, because that could falsely suggest increasing lengths of stay, as states focus more on adopting or otherwise permanently placing children who have been in foster care the longest.
The researchers suggest that all calculations should be based on samples of children entering care.
For example, states meet the CFSR standard for “time until reunification” if at least 76.2 percent of the children are reunified with their families within 12 months of their latest removal from the home. This measure is now based on the analysis of reunifications during the previous year. The researchers propose tracking the proportion of children who get reunified within one year of the date that they first entered the foster care system.
The two methods yield very different results, because some youth enter the system, are returned home, then removed again.
In one of the study’s examples, using the current CFSR method, one state fell further below the national standard for reunification, from 48.7 percent in 2000 to 36.9 percent in 2002, then rebounded to 45.8 percent at the end of its last AFCARS reporting session. Using the alternative measurement, the same state showed a gradual improvement in the proportion of children reunified within 12 months, from 21.8 percent to 23.6 percent over the same period.
While the second gain is not dramatic, it does not selectively exclude groups of children in care. The current measurement includes only reunified children, while the proposed measure tracks the experiences of all children.
The study documents similar problems with national standards in measuring time to adoption, permanency types or outcomes, and the number of children discharged from and subsequently returned to care.
According to the researchers, filtering AFCARS data through the “prospective lens of a longitudinal data system” using available statistical software would efficiently “provide an important check on the reliability of the existing CFSR retrospective measures.”