Disappointing results from an evaluation of an expert-assisted case management system
Goldbeck, L., Laib-Koehnemund, A., & Fegert, J.M. (2007). A randomized controlled trial of consensus-based child abuse case management. Child Abuse & Neglect, 31, 919-933.
This study examined the effects of expert-assisted case management of child abuse and neglect, as perceived by case workers. Child protection cases in the German child welfare and healthcare system were randomly assigned to either an intervention group (n = 42) or a control group, which received routine case management (n = 38). Problems included alleged or confirmed physical, emotional or sexual abuse and/or neglect. The case workers, from 12 different institutions, had varied professional backgrounds, social work or health-related. The experts, also with different professions and affiliations, all had outstanding practical experience in child protection, and were trained in the method of assisted case management by the research team. Cases were assessed at time of referral (T1), and again after 6 months (T2), based on two instruments developed by the authors and completed by workers. Variables measured were satisfaction with the degree of child protection, level of certainty in the process of investigation, risk assessment and intervention planning, quality of inter-institutional communication, and involvement of children and families. The intervention group received 2-6 case review sessions (M = 2.6) which included case-specific goal attainment scaling.
There was very little evidence that the intervention was effective. Compared to the control group, the workers in the intervention group reported: (1) for the sub-sample of cases with more severe abuse or neglect, a significantly greater reduction between T1 and T2 in certainty in estimating suspected child abuse, (2) fewer legal prosecutions of the perpetrators, and (3) lower involvement of families in planning the interventions. There were also “tendencies” (p<.10) toward greater increases in both satisfaction with the degree of child protection, and certainty about intervention planning.
Much of the evaluation methodology was sound. Intra-family abuse cases and extra-family cases were separately randomized, and no significant differences were found between the resulting groups at T1. Five-point Likert scales were used to record most case worker judgments. Separate analyses were performed for cases with moderate or severe levels of abuse. Statistical procedures, though not elaborate, were quite straightforward; between-group differences at T2 were tested by χ2 (categorical variables) or repeated measures ANOVAs—effect by group and time (scaled variables).
However, there were some problems. Except for abuse severity, there was no control for case variables. There was no correction to significance levels for multiple tests; the program impact may have been even smaller than the reported results suggest. Several factors might explain these flat findings. Self-selection bias could lead to a sample of highly-motivated and experienced case workers, with potentially less to learn from outside experts. Workers were not blind to group membership; some had cases from both groups, raising the possibility of a transfer effect.
There were also some surprising gaps in the overall program design. No setting or worker variables were discussed at all. Nothing was said about any possible matching of worker and expert disciplines, only that experts were from outside the workers’ own agencies. The dependent variables were case worker reports of perceptions and actions taken; there were no independent measures of any actual outcomes for the children.
In summary, though the idea of using expert assistance is intuitively attractive, there is absolutely no evidence here to suggest that this translates into better decisions for the children concerned.