54 Chapter 3 Phase 2: definite selection of items A. Missings and distribution of the scores. Overall, there were few missings (N = 45, 0.3%), obtained from 26 (7%) respondents. All scores on the VAS items showed non-normal distributions. Scores on all items (VAS and Likert scale) were assessed as being sufficiently distributed by our expert panel (ie every response option was selected by a substantial number of respondents). B. inter-item correlation. All items showed an inter-item correlation of >0.2 with at least one other item. None of the items showed an inter-item correlation >0.9 with any other item. C. Explorative factor analysis. EFA of OQUA-v4 generated 11 relevant factors with an eigenvalue >1. The scree plot showed no obvious ‘elbow’. Therefore, we decided to retain 11 factors. Two of these factors had no items with a factor loading >0.5. All items were grouped into the nine remaining factors which were then labelled: eight factors related to severity and frequency of complaints and one factor related to impact. Most factors related to severity and frequency of complaints consisted of only two or three items per factor and had a low (<0.70) value of Cronbach’s Alpha. Both parts of the questionnaire (complaints and impact) can be seen as separate constructs. The complaints-part represents a formative model: items do not necessarily correlate with each other.6 The impact-part represents a reflective model: a change in the observed variables is caused by a change in the construct. Each item in a reflective model is interchangeable. Factor analysis should only be applied to a reflective model.6 Hence, item reduction in the complaints-scale was conducted based on expert opinion. Complaints-scale. Items were critically reviewed by the expert panel. Seven items were removed because they did not seem to measure what they were supposed to measure or were found irrelevant (item no. 6, 22, 30, 32, 33, 35 and 38). Nine items about hearing loss were replaced by five items from an existing, validated and broadly used questionnaire; the ‘Amsterdam Inventory for Auditory Disability and Handicap’13. These items have proved to cover the different aspects in hearing (e.g. speech intelligibility in noise and in quiet, localisation, detection and discrimination of sounds). One item for every aspect of hearing loss was used.14 Three items about popping of the ear were combined into one item (‘my ear pops’). Impact-scale. EFA was conducted with the 12 items of the impact-scale. EFA generated two factors with an eigenvalue >1. The scree plot showed an obvious
RkJQdWJsaXNoZXIy MjY0ODMw