Community health centres (CHCs) are pivotal for delivering agricultural and nutritional interventions in sub-Saharan Africa, yet robust methodologies for evaluating their impact on crop yield improvements are underdeveloped. This brief report aims to methodologically evaluate the use of multilevel regression modelling for assessing the effectiveness of CHC systems in improving agricultural yields within an integrated health and agriculture programme. We analysed anonymised programme data from multiple CHCs using a three-level linear mixed model. The model, specified as y₈₉₊ = ₀ + ₁ X₈₉₊ + u₉ + v₊ + ₈₉₊, where u₉ and v₊ are random intercepts for CHC and district, was fitted with restricted maximum likelihood estimation and robust standard errors. The multilevel approach effectively partitioned variance, revealing that 32% of the variation in maize yield was attributable to differences between districts. A positive, statistically significant association was found between the frequency of CHC-led farmer visits and yield (p < 0. 01, 95% CI: 0. 15 to 0. 42 kg/ha per visit). Multilevel regression provides a statistically sound framework for evaluating the hierarchical impact of community-based health systems on agricultural outcomes, controlling for clustered data structures. Programme evaluators should adopt multilevel modelling to account for clustering in community-based interventions. Further research should integrate longitudinal designs to assess causal pathways. multilevel modelling, programme evaluation, agricultural extension, health systems, sub-Saharan Africa This report provides a novel methodological demonstration of how multilevel regression can be applied to quantify the specific contribution of community health centres to agricultural productivity within integrated development programmes.
Kato et al. (Sun,) studied this question.