Psychometric properties of a modified US-household food security survey module in Campinas, Brazil

Objective: To assess the internal validity of a multiple-item measure of household food security in Brazil using statistical methods based on the single-parameter logistic (Rasch) measurement model. Subjects/Methods: Sample of the non-institutionalized civilian population living in the municipality of Campinas selected using stratified cluster sampling. Of the 1000 households randomly chosen, 847 responded to the interview. Responses to each of the 15 questions were coded into dichotomous items indicating whether the specific food-insecure condition had occurred (other than in just 1 or 2 days) during the 3 months before the survey. Scaling analyses were conducted separately as well as jointly for adult/household-related items and child-related items. Item-fit statistics were examined to determine the extent to which the items appear to measure the same underlying phenomenon, and item severity scores were compared with those of equivalent items in the US Current Population Survey. Results: Except for one item, infit statistics were within a range considered adequate (0.80 - 1.2), indicating a common phenomenon being measured with approximately equal discrimination. The relative severites of the items in the Campinas survey were generally similar to those of equivalent items in the US Current Population Survey. Analysis of all 15 items together indicates a higher severity level for child-related items compared with equivalent adult-related items. Conclusions: This analysis will serve as the prototype for confirming the psychometric validity of a food insecurity scale at a national level.