Comparison of a Web‐Based Food Frequency Questionnaire with Photograph‐Assisted Food Records in Assessing Nutrient Intake of Adolescents

In assessing dietary intake of adolescents, utilizing computer and electronic media has several practical advantages over written media since this age group is more adept with modern technology. In recent years, digital dietary assessment methods have been developed in an effort to improve the accuracy of food and nutrient consumption estimates and to decrease respondent burden in quantifying food intake. We developed a web‐based food frequency questionnaire (FFQ) for adolescents and compared its performance in assessing the intake of 20 nutrients with multiple non‐consecutive photograph‐assisted food records. Data were provided by 33 female and 22 male adolescents of multi‐ethnic composition (~46% non‐Hispanic whites) with mean age of 15.3 ± 1.7 years, who met the selection and compliance criteria and consented to use their own mobile phones for the food records. Participants completed the self‐administered 151‐item semi‐quantitative web‐based FFQ and six 1‐day food records. Personal mobile phones were used to take photographs of foods to be eaten then sent together with the 1‐day food record via text messaging to a designated secure webmail address for the study. The Nutrition Data System for Research was used for nutrient composition and analysis of food records to determine intake of 20 nutrients (energy, carbohydrates, total fat, SFA, MUFA, PUFA, trans‐fats, total protein, vegetable protein, animal protein, fiber, β‐carotene, retinol, vitamins E, C, and B12, folate, calcium, iron, and zinc). Comparisons between the methods were done using Wilcoxon Signed Ranks test, correlations with energy adjustment and correction for attenuation with multiple measurements, and cross‐classification analyses. Compared to the food records, the web‐based FFQ has significantly higher absolute intake estimates for 17 nutrients except carbohydrates, trans‐fats, and vitamin E. Deattenuated correlation for energy is 0.37; deattenuated energy‐adjusted correlations ranged from 0.11 (vitamin B12) to 0.82 (animal protein), with 11 nutrients having deattenuated correlations greater than 0.50. Within‐subject variances in the multiple food records range from 0.02 (carbohydrates) to 2.12 (β‐carotene). Agreement within one quartile ranges from 61.8% (vitamin B12) to 99.0% (vegetable protein) while gross misclassification ranges from 0% (animal and vegetable proteins and vitamin E) to 12.7% (zinc). Overall, there is moderate to strong agreement ( r > 0.30) between the two methods in ranking nutrient intake estimates of adolescents. Given that photograph‐assisted food records are considered appropriate as a reference method in evaluating FFQs, our results demonstrate the usefulness of this web‐based tool in ranking the nutrient intake of this adolescent population. However, care must be taken when using absolute intake estimates as the exposure variable. The web‐based FFQ needs further improvement to reduce overestimation bias. Support or Funding Information Soy Nutrition Institute