Monthly food insecurity assessment in rural mkushi district, Zambia: a longitudinal analysis

BACKGROUND: Perception-based scales are widely used for household food insecurity (HFI) assessment but were only recently added in national surveys. The frequency of assessments needed to characterize dynamics in HFI over time is largely unknown. The study aims to examine longitudinal changes in monthly reported HFI at both population- and household-level. METHODS: A total of 157 households in rural Mkushi District whose children were enrolled in the non-intervened arm of an efficacy trial of biofortified maize were included in the analysis. HFI was assessed by a validated 8-item perception-based Likert scale on a monthly basis from October 2012 to March 2013 (6 visits), characterizing mostly the lean season. An HFI index was created by summing scores over the Likert scale, with a possible range of 0–32. The Wilcoxon matched signed-ranks test was used to compare distribution of HFI index between visits. A random effect model was fit to quantify the sources of variance in indices at household level. RESULTS: The median [IQR] HFI index was 4.5 [2, 8], 5 [1, 8], 4 [1, 7], 4 [1, 6], 3 [1, 7] and 4 [1, 6] at the six monthly visits, respectively. HFI index was significantly higher in visit 1 and 2 than visit 3–6 and on average the index decreased by 0.25 points per visit. Within- and between-household variance in the index were 10.6 and 8.8, respectively. CONCLUSIONS: The small change in mean monthly HFI index over a single lean season indicated that a seasonal HFI measure may be sufficient for monitoring purposes at population level. Yet, higher variation within households suggests that repeated assessments may be required to avoid risk of misclassification at household level and to target households with the greatest risk of food insecurity.