Household Food Insecurity Assessed by the Food Access Survey Tool: A Comparison between the Item‐wise Internal Validity in Rural Bangladesh and Rural Zambia

Abstract

Food insecurity can be accompanied by stressful experiences, which may be common across different cultures. Researchers have developed perception‐based scales that summarize stressful experiences in order to assess severity of household food insecurity (HFI). Stressful experiences evolve with progressive HFI, therefore different items can reflect different HFI severity. However, the performance of individual items on perception‐based scales in different cultural settings is unknown. We have adapted the Food Access Survey Tool, a 9‐item Likert scale, to assess HFI in the past 6 months in rural Bangladesh (n=11,992) and Zambia (n=840). We used the Rasch partial credit model (PCM) to study item‐wise internal validity and severity. Other statistical methods examined unidimensionality, monotonicity and measurement invariance assumptions. In Bangladesh, all 9 items fit well in the PCM and all assumptions were met. In Zambia, the item “frequency of square meals” displayed an out‐of‐range fit statistic on the PCM and violated model assumptions. We noted differences in the distribution of item severity: e.g., “frequency of buying staple foods” versus “worrying about food” scored lowest in item severity in Bangladesh and Zambia, respectively. “Borrowing food” and “taking food for credit” both ranked low in item severity in Bangladesh but high in Zambia. In both settings, “switching from a preferred staple to other grains” had the highest item severity. Though core experiences of HFI may be common, caution is necessary when adapting, interpreting, and comparing individual items commonly used in HFI assessment tools across cultures.