CONSTRUÇÃO DE ESCALAS PARAAVALIAÇÃO DO CONSUMO ENERGÉTICO, QUALIDADE DA DIETA E CONSUMO DE ALIMENTOS ULTRAPROCESSADOS COM BASE NA TEORIA DA RESPOSTAAO ITEM

Abstract

Eating is a set of multimotivated and multidetermined behaviors that promote diverse effects on the health and nutrition conditions of individuals and collectivities. During phases in which the intensification of physiological processes and behavior such as adoles cence, the promotion of irregular eating habits can cause deleterious effects throughout the life cycle. Thus, the objective of this dissertation was to estimate the level of con sumption of ultra-processed foods (UPF) in adolescents between 18 and 19 years old through the Theory of Response to Item (TRI). This is a psychometric study with data from the third phase of the RPS birth cohort consortium (Ribeirão Preto; Pelotas; São Luís). The sample consisted of 2,515 adolescent individuals aged between 18 and 19 years from São Luís, Maranhão, Brazil. Food items were extracted from the Food Frequency Questionnaire (FFQ) adapted and validated for the adolescent population. The determi nation of ultra-processed food items occurred through exploratory factor analysis (EFA). Ultra-processed food items were determined, foods that had a factorial load > 0.30 and commonality > 0.20 for a factor that conceptually grouped foods with a high degree of industrial processing. After extraction by EFA, the items that presented nutritional sim ilarity were grouped, making 13 ultra-processed food items from the FFQ. Subsequently, the food items were categorized according to their daily consumption portions, namely: do not consume; consume less than one portion; consume one less than two servings; consume above two servings. IRT models suitable for polytomic items were tested, with the best model determined by lower AIC and BIC values, as well as good goodness of fit (CFI > 0.90; TLI > 0.90; RMSEA < 0.90; 08). The latent trait was later divided into tertiles, namely: Low consumption (θ ≤ 90); Moderate consumption (90 < θ < 120) and High consumption (θ ≥ 120). The Samejima Gradual Response Model was fitted with satisfactory goodness of fit (CFI = 0.92; TLI = 0.90; RMSEA = 0.05 [CI95% 0.05 - 0.06]. The sample showed consumption of Moderate AUP (θ = 100.64), being signif icantly higher in girls (θ = 100.73).The food items showed satisfactory discrimination values (ai > 0.65) and location. The fast food item had the highest discrimination value (ai = 1.497), thus better differentiating the individuals in the sample with higher levels of UPF consumption. Regarding the location parameter (bi), the item artificially sweet ened drinks showed the lowest values in all consumption categories, thus indicating that individuals did not need to have high consumption of UPF to ingest this food, making it easy to consume. peak in median latent traits, therefore evaluating sa satisfactorily the consumption of UPF, especially in individuals with moderate consumption of this food group. The application of IRT in estimating latent traits related to the food consumption of collectivities proved to be satisfactory, with results with good quality of fit and pro vided information. The sample of adolescents showed moderate consumption of UPF, but significantly higher in the female public. The interpretable scale obtained allows assessing UPF consumption quickly and with a smaller number of items usually present in a FFQ, making it a potential tool for assessing bad eating habits in nutritional surveillance.