Can nutrition-sensitive agriculture interventions address intersectional inequalities in women’s diets? A mediation analysis using cross-sectional trial data from Odisha, India.

Background Improving nutrition for all requires understanding how interventions influence nutrition inequalities within society. Intersectionality, which considers how multiple disadvantages intersect, may offer more precise insight into the equity of these interventions. Objective Using an intersectionality-informed approach and mediation with exposure-mediator interaction, we investigated how participation in nutrition-sensitive agriculture interventions tested in the ‘UPAVAN’ trial affected inequalities in women’s diets in Odisha, India. Methods We analysed cross-sectional endline data from 3,294 mothers of children aged 0-23 months in 111 UPAVAN intervention villages. We estimated dietary inequalities as excess relative risk of minimum dietary diversity (MDD-W) according to Scheduled Tribe identity (ST, non-ST), education (≥5, <5 years), or wealth (higher, lower), and comparing intersectional groups that combine ST/non-ST with education or wealth group. We used a 4-way decomposition to estimate whether these MDD-W inequalities were affected by social group differences in: intervention participation rates (mediation only), participation benefits (interaction only), or both combined (mediated interaction). Results Intervention participation and MDD-W were greater among the more advantaged groups of non-ST, higher education, or higher wealth. Often, the more disadvantaged groups had greater participation benefits (interaction only), which narrowed MDD-W inequalities. However, intersectional groups with two disadvantaged characteristics (e.g., poorer ST) had smaller participation benefits than those with one (e.g., wealthier ST), which widened MDD-W inequalities. Differences in participation rates had negligible effects on MDD-W inequalities. Often, any marginal widening of MDD-W inequalities due to disadvantaged groups participating less (mediation only) was suppressed by their greater participation benefits (mediated interaction). Conclusion To our knowledge, this is the first intersectionality-informed analysis of nutrition interventions. UPAVAN interventions mostly had equitable impacts, reducing several inequalities in maternal diet quality. We demonstrate how intersectionality-informed analyses can help identify inequities in nutrition interventions and inform the design of inclusive interventions that reach and benefit the most marginalised groups.