106 CHAPTER 6 monitors the quality of the provided care. The number of HCPs’ sampled from each specialty was disproportionally stratified to also include sufficient respondents in the small HCP groups, such as optometrists, pulmonologists, and somnologists. HCPs were arranged on the registration date. Therefore, systematic sampling was applied within these HCP groups to include an equal number of novel and experienced participants. All these included HCPs received an invitation email with a personal URL link for the online SAF-TSUQ. Participants with unfinished questionnaires received reminders after 1 and 9 weeks. HCPs could unsubscribe from the questionnaire by email. Anonymous data collection for internal validation of the SAF-TSUQ questionnaire ran from July to September 2019 (ET). Four gift vouchers of €50 were raffled among respondents. Completion time of the finished questionnaires was calculated and outliers above 1 h were excluded. Statistical analysis The questionnaire was analyzed (ET and JvB) using R (version 3.6.1) [22]. Questionnaires of HCPs reporting not using the telemedicine platform were excluded. Psychometric analyses were conducted to assess construct validity and the internal consistency reliability of the questionnaire. Items with answering categories other than the standard format or items not answered by ≥95% due to the prior screener question were excluded for psychometric analysis. To keep “I do not know or not applicable” responses in the psychometric analyses they were recoded as “3,” reflecting the neutral answer category. The data set was assumed to be suitable for factor analysis (FA) if the Kaiser– Meyer–Olkin (KMO) measure of sampling adequacy was at least 0.60 and Bartlett’s test of sphericity was significant (p < 0.05) [23,24]. Principal component analysis (PCA) with direct oblimin rotation, common FA and confirmative FA (CFA) were conducted. Items in the common FA needed a minimum factor load of 0.40, items below this threshold did not fit to a scale. The CFA was performed on the original theorized model (six factors); loadings needed to exceed 0.45. Spearman’s rank correlation coefficient was calculated to find inter-item overlap between items (rs > 0.70 indicating one of the items is redundant). Cronbach’s alpha (α > 0.70) and item-total correlations (ITC > 0.40) were calculated to assess the internal consistency of the different questionnaire sections. RESULTS Phase 1: questionnaire development Figure 6.1 displays the number of items in each of the developing steps.
RkJQdWJsaXNoZXIy MTk4NDMw