Abstract
BackgroundWe evaluated the reliability and validity of the short form household food security scale in a different setting from the one in which it was developed.MethodsThe scale was interview administered to 531 subjects from 286 households in north central Trinidad in Trinidad and Tobago, West Indies. We evaluated the six items by fitting item response theory models to estimate item thresholds, estimating agreement among respondents in the same households and estimating the slope index of income-related inequality (SII) after adjusting for age, sex and ethnicity.ResultsItem-score correlations ranged from 0.52 to 0.79 and Cronbach's alpha was 0.87. Item responses gave within-household correlation coefficients ranging from 0.70 to 0.78. Estimated item thresholds (standard errors) from the Rasch model ranged from -2.027 (0.063) for the 'balanced meal' item to 2.251 (0.116) for the 'hungry' item. The 'balanced meal' item had the lowest threshold in each ethnic group even though there was evidence of differential functioning for this item by ethnicity. Relative thresholds of other items were generally consistent with US data. Estimation of the SII, comparing those at the bottom with those at the top of the income scale, gave relative odds for an affirmative response of 3.77 (95% confidence interval 1.40 to 10.2) for the lowest severity item, and 20.8 (2.67 to 162.5) for highest severity item. Food insecurity was associated with reduced consumption of green vegetables after additionally adjusting for income and education (0.52, 0.28 to 0.96).ConclusionsThe household food security scale gives reliable and valid responses in this setting. Differing relative item thresholds compared with US data do not require alteration to the cut-points for classification of 'food insecurity without hunger' or 'food insecurity with hunger'. The data provide further evidence that re-evaluation of the 'balanced meal' item is required.
Highlights
We evaluated the reliability and validity of the short form household food security scale in a different setting from the one in which it was developed
We recently explored experiences of food insecurity in a population survey in Trinidad and Tobago using the short form of the household food security scale [7]
While the consistency of these associations with those of other reports pointed to the validity of the scale, our results showed an unexpectedly higher proportion of affirmative responses to the item concerning 'balanced meals' [13], and an unexpected difference in the prevalence of food insecurity across ethnic groups
Summary
Subjects and questionnaire The sampling design for the study and data collection methods have been described previously [7]. Analysis Each item in the household food security scale was reduced to the categories of affirmative or not as described by Blumberg et al [4] (Table 1). A one parameter logistic (Rasch) model was fitted Item thresholds, their standard errors and item fit Pvalues are reported. Item thresholds are estimated parameters which indicate the location of the item in relation to the latent construct In this context, the threshold provides a measure of the relative severity of food insecurity associated with affirmative responses to an item. The combined data from the three ethnic groups were fitted to a oneparameter logistic model, but with separate item thresholds estimated for each ethnic group. We used logistic regression to estimate associations between item responses, as dependent variables, and income. The severity rankings of the other items were consistent with US data
Sign-in/Register to view highlights & summary Sign-in/Register to access full text options 
Free) 
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Schedule a call
