Fig. 5: Spatial distribution of the transported effects of MDA.
a, Percentage reduction in malaria incidence (1 − the ratio of cases in the MDA versus control arm) during the transmission season of the intervention year (2021), estimated using doubly robust transportability models. Covariates included precipitation, temperature, EVI, population density and the percentage of the population <10 years old. The covariate list included in the final models varied for each commune-month, following screening for collinearity, data sparsity, association with the malaria case count and feature selection using elastic net regression. Transportability analyses were restricted to communes where SMC was routinely offered during the trial period, that had a population size of <152 people per 100 m2 and that had a predicted probability of trial selection (P(S)) of >0.75. Separate analyses were performed for each commune-month (n = 366 per commune). b, Map indicating whether the 95% CIs for transported estimates include the null. c, Uncertainty—depicted as half the width of the 95% CI—estimated using a non-parametric bootstrapping procedure that resampled commune-months within trial datasets and months within the non-trial commune 1,000 times with replacement. Polygons indicate communes. Boundaries for Niokolo-Koba National Park were obtained from the World Database on Protected Areas via Protected Planet (www.protectedplanet.net). Basemap from GADM (https://gadm.org/).
