Description and performance evaluation of two new web-based tools developed to assess and monitor diet quality: Nova-HF and Nova-UPF screeners

Background and objectives: The quality of diets and risk of chronic diseases can be difficult to evaluate or monitor in some contexts due to the unavailability of tools that are simple and easy to collect. We developed two web-based tools to assess and monitor diet quality and described their performance against a traditional method of food consumption data collection. Methods: This analysis included part of the participants of the NutriNet-Brasil study, a web-based prospective cohort that currently counts on more than 100,000 volunteers. A total of 812 adults (18 years or older) answered a 2-minute self-administered questionnaire about the consumption of a set of foods on the day prior to the interview. Food items were based on the Nova food classification system: the Nova screener of healthy foods (Nova-HF) is comprised of 33 whole nutrient-dense foods (fruits, vegetables, legumes, whole grains, and nuts), and the Nova screener of ultra-processed foods (Nova-UPF) includes the most consumed UPF in Brazil (23 food items). Each tool generates a score, which is obtained by the sum of positive answers. We compared the Nova-HF and the Nova-UPF scores to the total energy intake (%) from whole nutrient-dense foods and ultra-processed foods, respectively, obtained through a web-based open-ended 24-hours recall developed and validated specifically for the NutriNet-Brasil study, which considers the food processing degree. We evaluated the relationship between quintiles of each score and the correspondent % of energy intake by linear regression, and the agreement between the quintiles of each method, using the Prevalence-adjusted and bias-adjusted kappa (Pabak). Results: Quintiles of each score presented a linear relationship with the correspondent % of energy intake (p-value for linear trend