(P19-040-23) Utilization of a Limited FNDDS Dataset in the PortionSizeTM App: A Pilot Study

Objectives: PortionSize™ is a food photography-based smartphone app that is used to assess dietary intake. The app currently contains ~1,150 food items with associated food codes and nutrition values from the Food and Nutrient Database for Dietary Studies (FNDDS). The objective of this study was to determine: 1) if users can correctly select food items from PortionSize’s limited nutrient database (PortionSize FNDDS) compared to trained nutrition professionals, and 2) if the resulting energy intake calculations from a simulated meal were the same between users and nutrition professionals.

Methods: Fifteen adult participants [73.3% female, mean (SD) age: 28.0 (12.2) years, mean (SD) BMI: 24.1 (6.6) kg/m²] were trained to use PortionSize and assessed a simulated meal with the app. Fifteen meals were used, consisting in 69 food items. Two nutrition professionals analyzed food images from PortionSize to identify the best FNDDS match from within the app’s limited dataset (these food codes and the resulting energy intake values were the criterion variables). Participants’ food codes were categorized as follows and reported in percentages: exact match (the food code was the same between the user and the criterion), loose match (the code reflected the correct major food group but was not an exact match), omission (the user failed to report a food in the meal), and intrusion (the user reported a food that was not in the meal). Equivalence tests using Two One-Sided T-tests (TOSTs) at 25% bounds were performed to compare the estimated energy value per 100 g of each food code.

Results: 69.6% of participants’ food codes were exact matches, 23.2% were loose matches, 5.8% were omissions, and 1.4% were intrusions. Participant-estimated energy intake (mean = 193.3, SD = 209.7 kcal) was equivalent to estimation of energy by nutrition professionals (mean = 215.6, SD = 221.3 kcal; P = 0.0195) using the PortionSize FNDDS dataset.

Conclusions: The majority of users food codes (92.8%) were exact or close matches, resulting in a 22.3 kcal (10.3%) underestimation of energy intake when compared to the criterion measures. These results indicate that agreement is sufficient, but could be improved, between users and nutrition professionals.

Funding Sources: NIDDK; Louisiana Clinical and Translational Science Center; Louisiana State University Leveraging Technology for Innovation.