The development of the questionnaire was done in a multiphasic process. In the first phase, a thorough review of the literature was conducted to identify the questionnaires and modalities that have been used in the Middle East [9-16]. The items in these questionnaires were then checked for their suitability and adaptation by an experienced dietician and a public health researcher (NS & AW) inclusion in the ATNQ. In addition, several more items were developed, based upon the Ajzen’s theory of Planned Behaviour, which states that an action requires three pre-meditated components; attitudes, knowledge and practice . Overall, these items assessed the participants on attitudes, knowledge and practices related to nutrition and diet in their clinical practice. Responses on these items were assessed using a five-point Likert Scale ranging from strongly agree to strongly disagree.
In the next phase, we recruited 18 dieticians, nutritionists (n=6), medical students (n=3) and medical doctors (n=9). The participants were requested to respond to the ATNQ and then comments on its suitability, strengths and weaknesses. Using open ended questions, they were also requested to comment on the items to be excluded or rephrase sentences for an improved comprehension. They were also requested to suggest more items that could be added in the questionnaire. Typical comments raised were to mention measurement units as mmol/l instead of mg/dl; less suitability of the questionnaire for medical students and a high number of questionnaire items. It was also suggested that questions pertaining to physical activity, renal-nutrition, bariatric surgery, physical activity and knowledge acquisition behaviors be added in the questionnaire. After the pilot study, we made necessary changes in the questionnaire, yielding a total of 52 items in the finalized questionnaire (Table 1). It is important to note that the data collected from the pilot survey was not included in the final dataset.
Based on our preliminary analyses, we expected a two to four factor solution for this questionnaire, where the items presented wide communalities. We judged a sample size of 200 to be adequate for our study based on Comrey & Lee’s (1992) recommendation for sample size calculation for factor analysis. They recommended that a sample size of 50 to 100 is poor, 200 is fair, and above 300 is good [19,20]. In addition, Mundform et al., recommended a sample size of 55 to 75 participants for scale items presenting with low communalities, two to four factor solution, variable to factor ratio of 12 and a good-level criterion (0.92) [19,20].
Thereafter, we initiated the cross-sectional survey where a total of 200 dieticians, nutritionists and medical doctors were invited to participant in the survey, using convenient sampling method. Participants were recruited using an electronic survey developed using Survey Monkey platform. Professionals from several institutes and hospitals were contacted to participate in the survey during face to face meetings conducted at the Ministry of Health in Kuwait. Before participating in the survey, all the participants signed informed consent forms. Participation in the survey was voluntary, anonymous and the participants could leave the study at any time. Average time for completion of the questionnaire was around 20 minutes. Ethical approval for this study was provided by Ethical Review Board of Ministry of Health of Kuwait, Kuwait.
All data were analyzed using the SPSS v.25. Firstly, the data was subjected to dimension reduction using the Principal Component Analysis (PCA) and orthogonal rotation . This process ascertained the dimensionality of the questionnaire by guiding the number of factors to retain and redundant items to be excluded. Before running the PCA, its suitability was assessed using the Kaiser-Meyer-Olkin (KMO) sampling adequacy statistic (> 0.60) and Bartlett’s test of sphericity. Number of factors to retain was based on three criteria; variance explained by each factor, Eigen value > 1 and the Cattell’s Scree Plot. Naming of each factor retained was done subjectively by analyzing the theme of most items included in the questionnaire. Suitability of each item was assessed using several criteria. For each item to be suitable for inclusion in the final scale, it was ensured that the KMO sampling adequacy value was > 0.6 for each item in the anti-image of the covariance matrix; communality value was > 0.2 and the factor loading was > 0.32.
Reliability analysis was done to evaluate the internal consistency of the overall scale, where a value > 0.60 as considered to be acceptable . Convergent validity was assessed using the Pearson’s correlation indices obtained using the inter-item correlations. Moreover, contribution to the overall Cronbach’s alpha value yielded by the scale was also assessed.