52 Chapter 3 serves to identify a set of latent variables (i.e. factors) underlying the items. A condition to perform EFA is that responses are normally distributed. Generalised least squares were used to correct for non-normally distributed data.11 The number of relevant factors was determined based on factors with an eigenvalue >1 and the ‘elbow’ in the scree plot. Orthogonal rotation (i.e. Varimax) was used to rotate the component matrix.6 After orthogonal rotation, items were grouped into factors based on their factor loading. A minimum loading of 0.5 was taken as a threshold.12 Cronbach’s Alpha was used to examine the coherency within a factor (i.e. items within a factor should represent a common latent variable). A value between 0.70 and 0.90 was considered acceptable.6 Chart 1. The number of participants in the various phases of the study. Results Patient characteristics Table 1 presents the patient characteristics of the study samples in the various phases, chart 1 shows the number of participants per phase. Table 1. Demographic characteristics in the various phasesa of the study. Characteristics Phase 1A: Interviews Phase 1B: Pilot study 1 Phase 1D: Pilot study 2 Phase 2: Field-testing Number of participants 16 32 39 352 Items investigated 0 34 47 50 Age, years Mean Range (min.–max.) Missing 47 24- 69 0 (0%) 55 20 – 88 0 (0%) 48 16-76 0 (0%) 49 16 – 93 10 (2,8%) Gender Male Female Missing 7 (44%) 9 (56%) 0 (0%) 10 (32%) 21 (66%) 1 (3%) 20 (51%) 19 (49%) 0 (0%) 174 (49,4%) 177 (50,3%) 1 (0,3%) Clinic Secondary Tertiary 0 (0%) 16 (100%) 29 (91%) 3 (9%) 0 (0%) 39 (100%) 49 (13,9%) 303 (86,1%) a Phase 1C was omitted in table 1 because in this phase only expert opinion was used. No patients were included in this phase. b Missing values are caused by respondents who forgot to fill in their age. c Missing values are caused by respondents who forgot to fill in their gender.
RkJQdWJsaXNoZXIy MjY0ODMw