Page 112 - Reduction of coercive measures

Chapter 5
the hierarchical data structure without any predictor variables. In this model, the intraclass correlation was computed for the level of coercive measures and for the level of residents by computing the proportion of ‘variance’ in the outcome variable attributed to each level (Davis & Scott, 1995). Second, the predictor Group (1 = program; 0 = control) was entered into the model to test the main hypothesis. Third, three dummy variables, together representing the four types of coercive measures, were entered as additional predictors in the model, in order to test whether, independent of the program effect, there was a difference between the types of coercive measures in the reduction of restraint use. Fourth, interaction terms between each dummy variable for type of coercive measure and the experimental group variable were added to the model to test whether the program effect was stronger for some types of coercive measures (protection from harm and danger as a result of challenging behavior) than others (reasons of surveillance techniques or physical support).
In a final step, analyses were added in which only the coercive measures registered prior to the intervention period were included. This was done to address the possibility that the intervention led to more awareness of coercive measures, and thus more registration and as a necessary consequence also more reduction of coercive measures. Additional analyses were carried out to address the alternative explanation for increases in reductions of coercive measures by heightened attention towards registration in the experimental condition.
110