Agreement between online and interviewer-administered assessments of nutrition and physical activity in Vietnamese youth

Online surveys offer scalable and cost-efficient approaches for collecting nutrition and lifestyle data. However, evidence on their agreement with interviewer-administered methods in low- and middle-income settings remains limited. This study aimed to evaluate method agreement between online and interviewer-administered assessments among Vietnamese youth.

Cross-sectional study.

The study was conducted among 734 individuals aged 16–25 years from Northern, Central, and Southern Vietnam. Participants completed both online and interviewer-administered surveys on the same day. Collected data included demographics, anthropometrics (weight, height, waist, and hip circumferences), dietary habits (e.g., breakfast, emotional eating), physical activity (frequency and duration), and a 24-hour dietary recall. Method agreement was evaluated using Spearman’s rank correlation coefficient (ρ), Concordance Correlation Coefficient (CCC), Cohen’s kappa for categorical variables, Wilcoxon signed-rank tests for paired comparisons, and Bland–Altman analysis. Proportional bias was assessed for energy intake by regressing the between-method difference on the mean of the two methods.

Demographic variables showed almost perfect agreement (κ ≥ 0.90). Age and weight showed high agreement (ρ ≥ 0.97; CCC ≥ 0.95), while height and eating frequency showed moderate concordance. Dietary habits showed substantial agreement (Kappa 0.63–0.80). Physical activity frequency had moderate agreement (CCC up to 0.69), while duration-based measures, including sitting time and sleep, showed lower concordance (CCC ≤ 0.62). The 24-hour dietary recall demonstrated poor agreement, with substantial mean energy overestimation (+ 1,475 kcal) and very low concordance (CCC ≈ 0.03). Regression analysis confirmed significant proportional bias (p < 0.001), with increasing discrepancy at higher intake levels.

Online self-administered surveys demonstrated good agreement for demographic characteristics and structured behavioral measures among Vietnamese youth. However, agreement for quantitative 24-hour dietary intake was poor, with substantial systematic and proportional bias. These findings suggest that online tools are appropriate for large-scale behavioral data collection but require methodological refinement and enhanced digital features before being used for accurate energy and nutrient estimation in low-resource settings.