Multilevel modelling of structured intersectional inequalities in child anaemia in Peru: a MAIHDA approach

Socioeconomic exposures related to anaemia in Peruvian children have been modelled assuming additive or log-additive relationships, yet such approaches overlook the fact that illness emerges from the complex interplay of multiple, intersecting determinants. Using data from the 2017-2023 Peruvian Demographic and Health Survey, we cross-classified age, wealth index, maternal education, ethnicity and region of residence to estimate the prevalence of anaemia across their intersections and decompose the total intersectional effects into additive and interaction components on the log-odds scale.

We applied Multilevel Analysis of Individual Heterogeneity and Discriminatory Accuracy (MAIHDA) using two-level random-intercept logistic regression, with individuals (Level 1) nested within 162 intersectional strata (Level 2). The first model estimated crude between-stratum variance, while the second included individual-level covariates to assess how much of that variance they explained.

A total of 255 381 children aged 6-59 months were analysed. The estimated prevalence of anaemia ranged between 10.2% and 68.1% across strata, being higher among the youngest and most disadvantaged (Indigenous, poor, low maternal education, non-Coastal). Most between-stratum differences were captured by the additive main effects of the strata-defining variables, consistent with a modest role for interactions.

Anaemia in Peruvian children is unequally distributed across intersectional social strata, with the highest burden concentrated among the most disadvantaged groups. These estimates and interpretations rely on standard MAIHDA assumptions.