Evaluation of an evidence-based science approach for the establishment of national nutrition guidelines

Abstract

Inter-rater agreement was assessed using kappa statistics (K) and percent agreement. Filtering decisions were initially performed independently and then revised after a 'consensus meeting' to discuss disagreements as a comparison of the responses with an expert reference standard (consensus evaluations by the investigators). One-hundred and eighty-five and heart disease risk in children. The inter-rater agreement was similar among like groups and between corresponding filtering (P> 0.05). Higher inter-rater agreement was observed filtering, there was no significant difference between groups for the proportion of agreement with the expert reference standard (P> 0.05) after the consensus meeting. For the groups, false-negative rates ranged from 11.5--26.2% for title filtering, 11.1--23.3% for If descriptions of inclusion/exclusion criteria, withdrawals and drop-outs, justification of sample size, and statistical analysis were present in an article, it was more likely to be appraised as 'Important'. Experienced scientists and nutrition professionals (experts) have traditionally developed national nutrition guidelines after a non-systematic review of relevant literature. The overall objective of this thesis was to develop literature filtering and article appraisal tools for use by non-experts for the creation of an evidence-based national nutrition guideline. A specific objective and article filtering and appraisal using these tools by 3--4 pairs of dietitians (RDs) and 3--4 pairs of nutrition students (Grads) (non-experts). Filtering is the process of selecting only those articles that are relevant to a review. This study describes the development and evaluation of literature filtering and appraisal tools for use by non-experts. After consensus meetings, both although important differences remained compared to the expert reference standard. Further work is needed to determine if non-experts can adequately complete literature filtering and article appraisals.