A Smartphone Food Record App Developed for the Dutch National Food Consumption Survey: Relative Validity Study (Preprint)

Abstract

BACKGROUND In the Dutch national food consumption surveys, dietary intake has been assessed since 2003 through 24-h dietary recalls using the GloboDiet software. A new self-administered smartphone food record application called DitEetIk! was developed for potential use in future surveys. OBJECTIVE This study aimed to evaluate the data collected with the DitEetIk!-app and its relative validity of food group, energy, and nutrient intake as compared to GloboDiet 24-h dietary recalls. METHODS 300 men and women aged 18–79 y were recruited from a consumer panel. Participants were asked to keep a record of their consumptions using the DitEetIk!-app on three non-consecutive days. Trained dietitians conducted a 24-h dietary recall interview by telephone using GloboDiet software and recalled one of three DitEetIk! recording days. The nutrient intake was calculated with NEVO version 2021/7.0. Participants with implausible records, defined as days with energy intake less than 0.6 BMR or more than 3.0 BMR, were excluded from the analyses. For 19 food groups and 29 nutrients, differences in median intakes were assessed using the Wilcoxon signed-rank test and Spearman rank correlation coefficients were calculated. Bland–Altman plots with mean differences and 95% limits of agreement were made for energy intake and the contribution to energy intake from fat, carbohydrates, and protein. RESULTS A total of 227 participants completed a combination of an app recording day and a 24-h dietary recall interview referring to the same day. Of this group, 211 participants, 104 men and 107 women, had a plausible recording day. Of all recorded foods, 13% were scanned and 19% text searched at the brand level. For 5 of the 16 food groups, the median intake assessed with the app was >10% lower than the median intake assessed with 24-h dietary recall (P<.05). This was also the case for 4 of the 29 nutrients (P<.05). For mean intakes differences were generally smaller. For energy intake, the mean difference and 95% limits of agreement was 14 (-1096 - 1124) kcal. Compared to expected energy intake, at group level, the mean percentage of underreporting with the app was 19.8% versus 20.1% with the 24-h dietary recall. Spearman correlation coefficients between intakes assessed with the DitEetIk!-app and the 24-h dietary recall ranged from 0.48 to 0.88 (median 0.78) for food groups and from 0.58 to 0.90 (median 0.72) for nutrients. CONCLUSIONS Compared to GloboDiet 24-h dietary recalls, the DitEetIk!-app assessed similar mean levels of energy intake, but somewhat lower median levels of intake for several food groups and nutrients. It is recommended to add functionalities to the app that stimulate a more complete food recording before applying it in the national food consumption surveys. In addition, more detailed information about consumed foods can be derived by encouraging the use of barcode scanning.