| Sumario: | I examine whether the effectiveness of community health worker (CHW) home visits in reducing under-five mortality varies with the population-to-CHW ratio in rural, fragile, and conflict-affected Mali. Muso, a global health NGO focused on reducing child and maternal mortality, recently signed a data use agreement with IFPRI to enable study of new research questions using data from its recent randomized controlled trial (RCT) of proactive home visits in Mali’s Bankass region. The original trial found that proactive home visits by professional CHWs did not lead to detectable reductions in child mortality relative to fixed-site care, although health infrastructure improvements in both treatment arms likely led to experienced large declines in under-five mortality across all study communities amid ongoing conflict.
In this project note, I study heterogeneous effects of the proactive home visit model by population-to-CHW ratios using nearly census level baseline population data and administrative records on CHW assignment. I test the hypothesis that home visits may operate differently when CHWs face particularly high or low caseloads. In theory, CHWs serving small populations may not be necessary for improving access because travel distances to fixed sites are already short, whereas CHWs serving very large populations may be unable to conduct sufficiently frequent and high-quality home visits. If both mechanisms operate, an optimal population-to-CHW ratio may exist.
Instead, I find evidence that CHW home visits are more effective when the population-to-CHW ratio is high. At 900 people per CHW (about the 90th percentile in the trial data), children in control clusters experience significantly higher mortality risk. This elevated risk, however, is offset in clusters assigned to proactive home visits. Additional checks indicate that this is driven by CHW coverage rather than underlying population size. These findings suggest that proactive home visits may be most beneficial in areas where fixed-site CHWs otherwise face heavy caseloads.
|
|---|