Forster, Hannah, Walsh, Marianne C., O'Donovan, Clare B., Woolhead, Clara, McGirr, Caroline, Daly, E. J., O'Riordan, Richard, Celis-Morales, Carlos, Fallaize, Rosalind, Macready, Anna L., Marsaux, Cyril F.M., Navas-Carretero, Santiago, San-Cristobal, Rodrigo, Kolossa, S., Hartwig, K., Mavrogianni, C., Tsirigoti, L., Lambrinou, C. P., Godlewska, M., Surwillo, A., Gjelstad, I. M. F., Drevon, C. A., Manios, Y., Traczyk, I., Martinez, J. A., Saris, W. H. M., Daniel, H., Lovegrove, Julie A., Mathers, J. C., Gibney, M. J., Gibney, E. R., and Brenan, L.
Background: Despite numerous healthy eating campaigns, the prevalence of diets high in saturated fatty acids, sugar, and salt and low in fiber, fruit, and vegetables remains high. With more people than ever accessing the Internet, Web-based dietary assessment instruments have the potential to promote healthier dietary behaviors via personalized dietary advice.\ud Objective: The objectives of this study were to develop a dietary feedback system for the delivery of consistent personalized dietary advice in a multicenter study and to examine the impact of automating the advice system.\ud Methods: The development of the dietary feedback system included 4 components: (1) designing a system for categorizing\ud nutritional intakes; (2) creating a method for prioritizing 3 nutrient-related goals for subsequent targeted dietary advice; (3)\ud constructing decision tree algorithms linking data on nutritional intake to feedback messages; and (4) developing personal feedback\ud reports. The system was used manually by researchers to provide personalized nutrition advice based on dietary assessment to 369 participants during the Food4Me randomized controlled trial, with an automated version developed on completion of the study.\ud Results: Saturated fatty acid, salt, and dietary fiber were most frequently selected as nutrient-related goals across the 7 centers.\ud Average agreement between the manual and automated systems, in selecting 3 nutrient-related goals for personalized dietary\ud advice across the centers, was highest for nutrient-related goals 1 and 2 and lower for goal 3, averaging at 92%, 87%, and 63%,\ud respectively. Complete agreement between the 2 systems for feedback advice message selection averaged at 87% across the centers.\ud Conclusions: The dietary feedback system was used to deliver personalized dietary advice within a multi-country study. Overall, there was good agreement between the manual and automated feedback systems, giving promise to the use of automated systems\ud for personalizing dietary advice.