Reliability and relative validity of a diet index score for adults derived from a self-reported short food survey

Aim: Assess the reliability and relative validity of a diet index score for adults derived using a 38-item online survey. Methods: The short food survey (SFS) measured ‘usual’ intake of seven food groups, three food choice indicators and variety; and was completed by 61 adults aged 19-50 years from Adelaide, Australia. A score was applied to assess compliance with the 2013 Australian Dietary Guidelines. Reliability of the survey was measured between two administrations one week apart; and validity by comparing the first administration to the average of three 24-hour dietary recalls. Statistical analyses included paired samples t-tests, intra-class correlation coefficient (ICC), percentage agreement, Cohen's kappa coefficients and Bland-Altman plots. Results: Estimates of daily food group servings were reliable (within 0.3 servings, P > 0.05) but not valid for all foods groups. The mean total index score was approximately 70 points (out of 100) on both SFS administrations (ICC = 0.71 (95% CI 0.56:0.81). Relative validity analysis showed moderate correlation between SFS#1 and 24-hour recalls (ICC = 0.43 (0.21:0.62, P < 0.001)), with 51% agreement in allocation to tertiles of diet quality between methods (k = 0.262, P = 0.004). The survey overestimated the diet index score by an average of 12.7 points out of 100 [-20.11:42.94] in comparison to recalls. Conclusions: The survey overestimated compliance with guidelines relative to dietary recalls. It demonstrated good reliability; however, the validity of estimating intake of some food groups needs improvement. Future refinement will provide a valuable online tool to assess compliance with the Australian dietary recommendations.