Page 88 - Reduction of coercive measures

Chapter 4
et al., 2017). In the second step, each predictor was entered separately into the model. Continuous predictors were centered prior to analysis. For the two variables number of coercive measures per subset, the same procedure was repeated. However, because the occurrence of 2 or more applied measures in the subset ‘applied at direct danger’ was rare (only 9 persons out of 209 persons), it was decided to dichotomize this count variable (0 = no coercive measures applied at direct danger; 1 = one or more coercive measures applied at direct danger) and analyze this variable with a binary logistic regression model as the type of generalized linear mixed model. In a third step, to determine the extent to which the disparate independent variables are mutually related to coercive measures, the factors have been jointly added to the analysis model. Within this third step the multivariate analysis was carried out in phases by first separate adding the resident related and then separate adding the staff related variables.
86