Children with behavior problems are known to experience greater instability in out-of-home placements. Using a nationally representative sample (n=5501), this study identifies characteristics of children upon entry into out of home care and predicts the number of placements experienced over the next 36-months. .

The data used in this study are drawn from the nationally representative National Survey of Child and Adolescent Well-Being (NSCAW). The subsample is composed of children with clinical level internal or external Child Behavior Checklist (CBCL) scores. Children were classified as having a clinical CBCL and by having fewer than four placement moves. About half of all children in foster family homes had a clinical CBCL, with about 60% of them having fewer than 4 moves. About 33% of the children in kinship care had a clinical CBCL but 90% had fewer than four moves. About 80% of the children in residential care had a clinical CBCL with only 55% having fewer than four moves. The mean number of moves for children with a clinical CBCL was 3.45 compared to 2.72 for those without a clinical CBCL and the medians were 3 and 2 moves, respectively. Bivariate analyses indicated that a child with special needs and high rates of risky behavior (i.e., sexual activity, substance abuse, or delinquency) were strongly associated with placement moves.

A Poisson regression model was used to estimate the number of moves; demographic characteristics and other variables suggested by the literature were used as predictors. Poisson regression was selected to account for the use of non-continuous, count data as the dependent variable. Poisson regression also offers an ‘offset' option which allows the model to account for the correlation between time and number of moves. The logarithm of the number of days in out-of-home care was the offset variable. Sensitivity analyses were conducted by constraining the amount of time in care—limiting it to at least 3-months and at least 6-months--to determine the effects on the results.

The unconstrained Poisson regression results indicate that a child's lack of hopefulness of being reunified with his or her family of origin predicts movement in placement. Additionally, having special developmental needs predicts an increased likelihood of movement in placement. Other variables such as gender and age identified as significant in predicting placement movements in the general population of out-of-home children were found to be nonsignificant among behavior disordered, out-of-home children. The two time constrained regression models generated slightly different results.

The measures of central tendency indicate that children with clinical behavior problems do not have dramatically more moves than children without. This reaffirms that many placement moves have other causes. The variations in regression results suggest that standard procedures for studying placement moves that required a minimum duration in care should be checked against models without constraining assumptions. At the service level, this study provides evidence for training of substitute caregivers of special needs children and children facing diminishing likelihood of reunification.