ABSTRACT
OBJECTIVE
To characterize the internal structure of the Food and Nutrition Surveillance System (Sisvan) form of food intake markers for individuals over 2 years of age and to investigate measurement invariance between Brazilian macro-regions, life stages and over the years.
METHODS
A parallel analysis with factor estimation was carried out, complemented with exploratory factor analysis using all Sisvan records with valid responses in the country in 2015 (n = 298,253). Only the first record per individual was considered. Next, multigroup confirmatory factor analysis was used to investigate configural, metric and scalar invariance between the five macro-regions (Midwest, Northeast, North, Southeast, South) and life stages (children, adolescents, adults, elderly) in the same reference year. Invariance was evaluated longitudinally using valid individual records from 2015 to 2019 (n = 4,578,960). The adequacy of fit indices was observed at each step.
RESULTS
Acceptable fit indices and adequate factor loadings were found for a two-dimensional model, which grouped ultra-processed foods (factor 1) and unprocessed or minimally processed foods (factor 2). The two-dimensional structure, with the respective items in each factor underlying the set of markers, was equivalent across macro-regions, life stages and longitudinally, confirming the configural invariance. The weights of each item and its scale were homogeneous for all groups of interest, confirming metric and scalar invariances.
CONCLUSIONS
The internal structure of the Sisvan form of food intake markers adequately reflected its conceptual foundation, with stability of factors related to healthy and unhealthy eating in configuration, weights and scale in the investigated categories. These findings qualify food and nutritional surveillance actions, enhancing the use of Sisvan food intake markers in research, monitoring, individual guidance, and care production in the Brazilian Unified Health System.
Eating; Food and Nutrition Surveillance; Data Reliability; Health Information Systems; Psychometrics