E. Adu Bonsu, D. Ebo, D. Doku
Objectives: To examine the association between maternal media exposure and childhood anemia in three sub-Saharan African countries, and to develop a two-stage multiple imputation approach for substantial hemoglobin outcome missingness in Demographic and Health Survey (DHS) data. Design: Multi-country cross-sectional analysis of nationally representative DHS data. Setting: Community-based household surveys in Ghana (2022), Nigeria (2023-24), and Tanzania (2022). Participants: 42,944 children aged 6-59 months across 2,621 clusters. Primary and secondary outcome measures: The primary outcome was anemia (altitude-adjusted hemoglobin <11.0 g/dL, WHO criteria); the primary exposure was a composite maternal media score (0-4 channels). Given 61.5% hemoglobin missingness, we imputed altitude-adjusted hemoglobin using multilevel Gaussian models and derived anemia deterministically, then estimated associations using population-averaged generalized estimating equations (pre-specified primary analysis), with six sensitivity analyses and exploratory mediation analysis. Results: Media exposure was significantly associated with reduced anemia odds (OR 0.946, 95% CI 0.917 to 0.976, p<0.001; FMI 32.5%), a 5.4% reduction per additional weekly channel. Estimates were consistent across six sensitivity analyses (OR range 0.914-0.961), with no heterogeneity by country (I2=0%, p=0.482). Media exposure predicted antenatal care attendance and iron supplementation (both p<0.001), but neither mediated the anemia association. Conclusions: Maternal media exposure was consistently associated with reduced childhood anemia across three sub-Saharan African settings. A two-stage multiple imputation approach addressed substantial (61.5%) outcome missingness while maintaining stable estimates, offering a transferable framework for DHS-based analyses. Prospective and intervention studies are needed before media-based communication can be recommended as a complementary anemia-control strategy. Keywords: Anemia; Child Health; Mass Media; Health Surveys; Data Interpretation, Statistical