Using multilevel modelling to explore decision making in child welfare

The goal to understand what factors predict child welfare service decisions may be addressed through the quantitative method of multilevel modelling (MLM). MLM provides an opportunity to examine whether child welfare decisions can be predicted by various characteristics and whether they vary by group factors, such as worker, team, department, organisation or geographical location. This quantitative method addresses data dependence through data collected at multiple levels, a specific data set structure and multiple statistical tests. Unless nested data are appropriately structured and analysed, such as in MLM, there is an increased risk of Type I errors, that is, false positives. MLM provides researchers of decision making with the opportunity to assess whether decisions are unique to certain group characteristics and the degree to which decisions vary. This article presents MLM as a method for exploring decision making through an example taken from the Canadian child welfare context, whereby clinical, child welfare worker and organisational characteristics are assessed for their relationship to decisions to transfer families to ongoing child welfare services. Results illustrate the importance of utilising MLM as a method for exploring decision making and potential future uses.