Relative Validity and Reliability of the Remind App as an Image-Based Method to Assess Dietary Intake and Meal Timing in Young Adults

**Background:** Accurate assessment of dietary intake and meal timing is crucial for chrononutrition research. Traditional methods like pen-and-paper food records are burdensome and prone to bias. Image-based dietary records via smartphone apps offer a promising alternative, but their validity for assessing meal timing has not been rigorously tested. This study aimed to evaluate the relative validity and reliability of the Remind app as an image-based method to jointly assess dietary intake and meal timing compared to handwritten food records.

**Methods:** A 3-day cross-sectional study was conducted with 71 young adults (81.7% women; mean age 22.5 ± 2.3 years; mean BMI 22.3 ± 3.2 kg/m²) recruited from the University of Barcelona. Participants completed both a 3-day handwritten food record (reference method) and a 3-day image-based dietary record using the Remind app (test method) on the same days (two weekdays and one weekend day). A registered dietitian instructed participants on recording details. For the image-based method, participants photographed foods/beverages at a 45° angle with a fiducial marker and entered written descriptions and times. Dietary data were processed using PCN Pro 1.0 software. Relative validity was assessed using five statistical tests per Lombard et al.: Bland-Altman plots with Spearman correlation for proportional bias, % difference, paired t-test/Wilcoxon signed-rank test, Pearson/Spearman correlation coefficients, and cross-classification into tertiles. Reliability was evaluated using the intraclass correlation coefficient (ICC) with 95% confidence intervals.

**Key Results:** The average energy intake was 1485.6 ± 301.4 kcal/day. Mean meal timings were breakfast at 09:27 ± 01:00, lunch at 14:11 ± 00:34, and dinner at 21:31 ± 00:35. For energy and macronutrients, Bland-Altman plots showed most values within acceptable limits. Bland-Altman Spearman correlation indicated good outcomes (p > 0.05) for energy and most nutrients, but poor outcomes for iron (p = 0.031), vitamin E (p = 0.006), and vitamin C (p = 0.042), suggesting proportional bias. The % difference was <11% for all energy and nutrients (good). Paired t-test/Wilcoxon showed good agreement for energy, protein, fat, and fiber, but poor for carbohydrate and several micronutrients. At the individual level, correlation coefficients and cross-classification were good for all energy and nutrient intakes (e.g., energy r = 0.798, 64.8% in same tertile, 1.4% in opposite). For food groups, Bland-Altman Spearman correlation was good for most except tubers (p = 0.003) and oils/fats (p = 0.028). % difference was good for most food groups, acceptable for eggs (11.8%) and non-alcoholic drinks (−20.0%), and poor for legumes (20.6%) and oils/fats (31.8%). Cross-classification was good for all food groups except oils/fats (52.1% same tertile, 14.0% opposite). For meal timing, all statistical tests showed good agreement at both group and individual levels (e.g., breakfast: % difference 0.1%, r = 0.998, 100.0% in same tertile; lunch: % difference −0.0%, r = 1.000, 96.7% in same tertile). Reliability (ICC [95% CI]) was moderate to excellent for energy (0.793 [0.688, 0.928]), good to excellent for dietary fiber (0.906 [0.853, 0.940]) and vitamin B12 (0.917 [0.868, 0.948]), and excellent for fish intake (0.956 [0.930, 0.972]) and meal timing (breakfast ICC = 0.998 [0.997, 0.999]; lunch ICC = 1.000 [1.000, 1.000]). Oils and fats showed poor to moderate reliability (0.383 [0.166, 0.565]).

**Clinical Implications:** The Remind app demonstrates good relative validity and moderate to excellent reliability for jointly assessing dietary intake and meal timing in young adults. This is the first study to validate an image-based app specifically for meal timing, which is critical for chrononutrition research linking temporal eating patterns to health outcomes like obesity. The app reduces participant burden and improves data quality through real-time monitoring and visual documentation. However, limitations include the relative validation design (neither method is a true gold standard), a 3-day recording period that may be insufficient for micronutrients, and a sample of young adults that may not generalize to other populations. The app showed suboptimal performance for certain food groups (oils/fats, legumes) and some micronutrients, warranting caution in those specific assessments.