Evaluating Measurement Invariance of the Household Food Insecurity Access Scale Among Youth across Age, Gender, Country, and Time

Background: Validity of food insecurity measures and its invariance (equal measures across groups and items) is critical for practice and research. The specific measure in question is used to inform program eligibility and make inferences about relationships of food insecurity with other outcomes related to living stability. However, there is scant evidence to demonstrate validity and invariance of food insecurity measures among youth across different groups. We examined construct validity and measurement invariance of the Household Food Insecurity Access Scale (HFIAS), a common measure of food access, among youth across gender, age, time, and country. We hypothesize that invariance will be demonstrated across the groups we test, as this measure is written at a fifth-grade level and asks interviewees about basic food attainment.

Methods: Sample includes 8,076 youth aged 15 to 24 years old from Ghana (78%) and South Africa (22%). The sample is 66% female, 15% are in middle adolescence, and 54% in late adolescence. We examined construct validity by country, age, gender, and time through confirmatory factor analysis (CFA). We then explored measurement invariance by sequentially testing baseline, configural, metric, and scalar invariances across gender, age, time, and country. We evaluated model fit using model chi-square, comparative fit index (CFI), Tucker-Lewis index (TLI), and standardized root mean square residual (SRMR). Analyses were conducted using Mplus 8.3.

Results: Results suggest validity of the unidimensional HFIAS when used in Ghana and South Africa. Findings also indicate configural, metric and scalar invariance across countries. HFIAS remains invariant across gender (male or female), and across time (baseline, after six months, after one year, and after two years). Additionally, models indicate configural, metric, and scalar invariance across age groups (middle adolescence, late adolescence, emerging adulthood). All models showed good fit, which included values of ≥ 0.95 for CFI and TLI, and < 0.08 for SRMR.

Conclusions: HFIAS demonstrates that the unidimensional structure applies to all groups (configural); individual HFIAS items are equally salient to the construct of food insecurity for all groups (metric); and members of different groups who experience the same true levels of food insecurity choose the same response options (scalar). In other words, HFIAS scores are comparable across youth in Ghana and South Africa, gender, age, and time. This allows for continued use of this measure, given invariance is shown across groups. When a scale like HFIAS shows invariance, it shows that the scale is measuring what it intends to measure. For example, constraining the population to inspect invariance over the specified group, like gender, means that each sub-group (male and female) is being measured in the same way as the entire group is being measured, collectively. Therefore, men and women are shown to report food insecurity in the same way the entire group is reporting food insecurity. This extends to all groups involved in this analysis, which means the HFIAS is invariant across country, gender, age, and time. Because this measure is used all over the world, establishing measurement invariance is important in understanding and validating its continued use.