Differences in Population-Based Dietary Intake Estimates Obtained From an Interviewer-Administered and a Self-Administered Web-Based 24-h Recall.

Web-based instruments are being increasingly used in nutrition epidemiology and surveillance. However, the extent to which dietary intake estimates derived from web-based 24-h recalls such as the R24W are consistent with data derived from more traditional interviewer-administered 24-h recalls (TRAD) remains uncertain. Our objective was to compare dietary intake estimates obtained using the R24W and a TRAD instrument in population-based samples from the province of Québec in Canada. This comparison of dietary assessment methods was based on data from two sample survey studies in adults (18-65 years). The R24W was used in a sample of 1,147 French-speaking adults from five regions of Québec as part of the PREDISE ( PRÉDicteurs Individuels, Sociaux et Environnementaux ) study. The TRAD was used in a sample of 875 French-speaking adults from the Canadian Community Health Survey 2015 located in the same five regions. Characteristics of both samples were matched through selection and weighting (language, sex, age, region, education, body mass index, weekend day, and season of survey). Mean and usual intake data of each sample were compared. The plausibility of reported energy intakes was compared using predictive equations of the Institute of Medicine. Mean servings/day from the R24W were higher than with TRAD for vegetables and fruit (+11%, P = 0.003), grain products (+7%, P = 0.06), milk and alternatives (+21%, P < 0.001), and meat and alternatives (+18%, P = 0.001). Intake of low nutritive value foods was also 28% higher with the R24W than with TRAD (mean difference +164 kcal; 95% CI, 107-222). As a result, total energy intakes from the R24W compared with TRAD were 18% higher in women (mean difference +325 kcal; 95% CI, 243-407) and 15% higher in men (mean difference +361 kcal; 95% CI 232-490). The prevalence of underreporting of energy intakes was 10% lower with the R24W than with TRAD (prevalence ratio 0.90; 95% CI, 0.86-0.94). In conclusion, differences between dietary assessment methods in the context of population-based surveys on nutrition have potentially important consequences on the quality of the data and should be carefully considered in future surveys and surveillance strategies.
(Copyright © 2020 Brassard, Laramée, Robitaille, Lemieux and Lamarche.)