The test-retest reliability and validity of food photography and food diary analyses

ABSTRACT:
Aims: To assess test-retest reliability of both food photography and food diary methods and validity of these data against known values derived from food labels.
Methods: Test-retest reliability analyses of food diary and food photography were compared using single foodstuffs using intra-class correlation coefficients, coefficients of variation and limits of agreement. For food diaries, 24-h test-retest reliability was also examined. Validity was assessed against weighed analyses. As part of habitual intake, a single foodstuff (randomly allocated from 14 common foods) were consumed by 26 participants over 24-h. On two occasions (14 days apart), single-blind dietary analyses allowed estimation of foodstuff-specific energy and macronutrient content, and 24-h intakes.
Results: For food diaries, test-retest reliability was acceptable (weight, energy, carbohydrate, protein, fat: all intraclass correlation coefficients >0.990, coefficient of variation percentage: <0.1%, limits of agreements: <0.1 to <0.1, p>0.05, effect size: <0.01). For food photography, test-retest reliability was acceptable for weight, energy, carbohydrate, and protein (all intraclass correlation coefficients >0.898, coefficient of variation percentage: 3.6% - 6.2%, limits of agreements: 1.1 to – 44.9, effect size: 0.01 – 0.12). Food photography validity was worse than food diaries for all variables (percentage difference: 8.8% - 15.3%, coefficient of variation percentage: 7.5% - 13.8%, all; p≤0.05, effect size: 0.001 – 0.11).
Conclusions: Greater reliability and validity occurred in food diaries versus food photography; findings which may suggest that using food photography may lead to an under-estimation of energy and macronutrient content, which may have implications for dietary interventions and nutritional strategies.