Repeatability and validity of a food frequency questionnaire in free-living older people in relation to cognitive function

Abstract

Objectives: To determine the repeatability and validity of a self-administered, 175-item food frequency questionnaire (FFQ) in free-living older people and to assess whether these are influenced by cognitive function. Participants and setting: 189 free-living people aged 64-80y were recruited from participants in a previous study. Design: To assess repeatability, 102 (52M, 50F) participants completed the FFQ on two occasions three months apart. To assess validity, another 87 participants (44 M, 43 F) completed the FFQ and a four-day weighed diet record three months later. 25 nutrients were studied. Results: For repeatability, Spearman rank correlation coefficients were above 0.35 (p<0.05) for all nutrients. Cohen’s weighted Kappa was above 0.4 for all nutrients except starch, riboflavin, retinol, β-carotene, and calcium. There were no substantial differences in correlation coefficients between sub-groups divided by short-term memory test score. There was no clear pattern for correlation coefficients in sub-groups divided by executive function test score. For validity, the Spearman rank correlation coefficients were above 0.2 (p<0.05) for all nutrients except fat, mono-unsaturated fatty acids, niacin equivalents and vitamin D, and Cohen’s weighted kappa was above 0.4 for alcohol and was above 0.2 for 13 other nutrients. Participants in the lowest-score groups of short-term memory and executive function had the lowest median Spearman correlation coefficient. Conclusions: The FFQ had reasonable repeatability and validity in ranking nutrient intakes in this population though the results varied between nutrients. Poor short-term memory or executive function may affect FFQ validity in ranking nutrient intakes.