Fallon, B., Chabot, M., Fluke, J., Blackstock, C., Sinha, V., Allan, K., & MacLaurin, B. (2015). Exploring alternate specifications to explain agency-level effects in placement decisions regarding Aboriginal children: Further analysis of the Canadian Incidence Study of Reported Child Abuse and Neglect Part C. Child Abuse & Neglect. http://doi.org/10.1016/j.chiabu.2015.04.012
Data from the Canadian Incidence Study of Reported Child Abuse and Neglect (CIS) have shown that Aboriginal children are overrepresented in out-of-home care; this study examines the effects of agency-level variables on the overrepresentation. A consistent predictor found in earlier studies has been the proportion of investigations at an agency involving Aboriginal children; this study is a further exploration of the relationship, using previously unavailable organizational variables from CIS-2008.
The present analysis was based on a subsample of 1710 substantiated maltreatment investigations opened to ongoing child welfare services. The outcome variable was placement in out-of-home care at the conclusion of the initial investigation. The following possible predictors were included; four, shown in italics, were found to be highly statistically significant:
Four case level variables: Physical harm; Mental/emotional harm; Ethnicity; Caregiver cooperation.
Seven organizational variables: >45% of investigations involve Aboriginal children; Government vs. community-run agency; Range of services vs. child protection only; Differential response; Alternative dispute resolution; Agency on reserve; Geographic remoteness.
Consistent with earlier findings, the risk of an Aboriginal child being placed was found to be significantly greater in agencies serving large proportions of Aboriginal children. Here, the authors speculate that the proportion of Aboriginal investigations may be a proxy for a lack of services and resources in Aboriginal child welfare agencies. The second organizational predictor was new; the risk of placement was greater in government-run than in community-run agencies. The authors conclude that further analysis is needed to fully understand organizational-level variables.
All variables were binary. The statistical procedures used in the data analysis—multi-level logistic regression (MPlus software), with multiple imputation to correct bias due to missing data—seemed reasonable. The problems were in the choice of variables.
Case level: In earlier studies of CIS-2008 data, socioeconomic risk factors were significant predictors of placement; the effects of poverty and neglect are mentioned in the discussion here. However in this study no risk factors at all were included with the case variables; this might affect the significance found for the organizational variables.
Organizational level: The most useful finding would be a significant predictor of overrepresentation that could be modified to reduce the problem; percentage of Aboriginal investigations does not itself meet this need. However, if this variable actually can be shown to be a proxy for lack of services and resources, it would be very useful, as these variables depend on agency funding, and could be a target for corrective action. The authors do discuss the inadequate funding of Aboriginal agencies, but without specific reference to the agency categories in their own data. How are the community agencies in this study funded? Federal underfunding of on-reserve child welfare services has been criticized elsewhere; could that be a factor in these present findings? This is clearly an important topic, and deserves further study.