Relative Validity and Reproducibility of a Food Frequency Questionnaire for Assessing Dietary Patterns and Food Group Intake in Older New Zealand Adults: The Researching Eating, Activity, and Cognitive Health Study

Abstract

Background: Dietary pattern analysis considers the overall dietary intake and combinations of foods eaten. Valid and reproducible tools for determining dietary patterns are necessary to assess diet-disease relationships.

Objective: This study evaluated the relative validity and reproducibility of the Researching Eating, Activity, and Cognitive Health (REACH) Study food frequency questionnaire (FFQ) specifically designed to identify dietary patterns in older adults.

Design: A subset of participants from the REACH study completed two identical 109-item FFQs 1 month apart (FFQ1 and FFQ2) to assess reproducibility and a 4-day food record between FFQ administrations to assess relative validity. Foods from each dietary assessment tool were assigned to 57 food groups. Principal component analysis was applied to the food group consumption reported in each dietary assessment tool to derive dietary patterns.

Participants and setting: Dietary data were collected (2018 and 2019) from a subset of the REACH study (n = 294, 37% men) aged 65 to 74 years, living in Auckland, New Zealand.

Main outcome measures: Daily intakes of 57 food groups and dietary patterns of older adults participating in REACH living in New Zealand.

Statistical analysis: Agreement of dietary pattern loadings were assessed using Tucker's congruence coefficient. Agreement of dietary pattern scores and food group intakes were assessed using Spearman correlation coefficients (acceptable correlation rho = 0.20 to 0.49), weighted kappa statistic (acceptable statistic κ_{w =} 0.20 to 0.60), and Bland-Altman analysis, including mean difference, limits of agreement, plots, and slope of bias.

Results: Three similar dietary patterns were identified from each dietary assessment tool: Mediterranean style, Western, and prudent. Congruence coefficients between factor loadings ranged from 0.54 to 0.80. Correlations of dietary pattern scores ranged from 0.47 to 0.59 (reproducibility) and 0.33 to 0.43 (validity) (all P values < 0.001); weighted kappa scores from 0.40 to 0.48 (reproducibility) and 0.27 to 0.37 (validity); limits of agreement from ± 1.79 to ± 2.09 (reproducibility) and ± 2.09 to ± 2.27 (validity); a negative slope of bias was seen in the prudent pattern for reproducibility and validity (P < 0.001).

Conclusions: The REACH FFQ generated dietary patterns with acceptable reproducibility and relative validity and therefore can be used to examine associations between dietary patterns and health outcomes in older New Zealand adults.

Keywords: Congruence coefficient; Dietary assessment; Dietary patterns; Older adults; Reliability.