(P11-010-24) A Food and Nutrient Database for Dietary Studies (FNDDS) Tailored to Hawaii May Reduce Dietary Assessment Error

Objectives: Many commonly consumed foods (including mixed dishes) in Hawaii are not found in the Food and Nutrient Database for Dietary Studies (FNDDS). Therefore, when using FNDDS for dietary assessment in Hawaii, the closest match may need to be selected (e.g., closest match for spam musubi in FNDDS is sushi NFS), which may introduce error. Alternatively, nutrients for these foods could be estimated using recipes and then added into FNDDS. PacTrac 3 is a nutrient database developed for Hawaii and the US affiliated Pacific. The objectives of this study were to identify commonly consumed foods in Hawaii that are not in FNDDS, and to assess absolute error introduced by 1) matching foods to the closest match in FNDDS, and 2) estimating nutrients of foods using recipes, with PacTrac being the gold standard comparison method.

Methods: A dataset of 1,858 unique foods was compiled from dietary studies conducted at the University of Hawaii (UH) at Manoa and the UH Cancer Center. Using this dataset, trained researchers (CL and KA) identified 97 foods not found in FNDDS (version 17-18). These 97 foods were 1) matched to the closest match in FNDDS and 2) using recipes, broken down into ingredients and matched to FNDDS ingredients codes (version 17-18). Differences in absolute error for energy (kcal), and nutrients on Nutrition Facts Labels available from the PacTrac 3 output (total fat, saturated fat, cholesterol, sodium, total carbohydrate, dietary fiber, protein, calcium, iron, and potassium), per 100g of food, were assessed between the “closest match” method and the “recipe” method, using paired samples t-tests and Cohen’s d for effect size.

Results: For the “closest match” method there was significant error for estimates of cholesterol, total carbohydrate, and dietary fiber (all p-values < 0.05 and all Cohen’s d >0.2) and no significant error and small Cohen’s d ( < 0.2) for all other nutrients. For the “recipe” method, there was no significant error and small Cohen’s d for energy (kcal), and all nutrients assessed. Differences in error between the “closest match” and “recipe” methods were significantly different for cholesterol and total carbohydrate (p-values < 0.05), but not for dietary fiber (p=0.16).

Conclusions: Tailoring FNDDS for Hawaii may reduce dietary assessment error for cholesterol and total carbohydrate.

Funding Sources: USDA NIFA