Internal validity of the Food Access Survey Tool in assessing household food insecurity in rural Zambia

We assessed the internal validity of the Food Access Survey Tool (FAST) using data from households (n = 907) enrolled in an efficacy trial of biofortified maize in rural Zambia. This scale assesses food insecurity over a 6-month recall period. A Rasch partial credit model was used to evaluate item performance. Unidimensionality was assessed by principal component analysis, monotonicity was assessed by non-parametric methods, and differential item functioning (DIF) by several characteristics was assessed by cumulative ordinal logistic regression models. One item (frequency of consuming three square meals) did not fit the partial credit model. The remaining eight items fit in a primary single statistical dimension and item category severity increased monotonically with increasing severity of food insecurity. We identified statistically significant DIF in three subgroup comparisons, but effect sizes of total DIF were considered practically insignificant (<2 %). After excluding the item on "square meals," the FAST serves as an internally valid tool to measure household food insecurity in rural Zambia.