Active surveillance for malaria is required in near-elimination settings to prevent onward transmission from a small number of infected individuals. Having a robust understanding of the temporal and spatial scales of transmission is essential for developing an effective active surveillance program. Inferring transmission chains, including with the use of genetic data, can reveal the spatiotemporal dynamics of P. falciparum transmission in these settings. A previous effort did this with spatial and temporal data in the context of eSwatini, a country positioned to become the first nation in sub-Saharan Africa to achieve elimination. However, there are a number of factors, including the choice of data types and assumptions about travel history information, that could cloud estimates of malaria transmission chains. We built on this prior work by incorporating a more mechanistic understanding of the spatial and temporal scales of transmission to infer P. falciparum malaria transmission chains. Because our approach is fully Bayesian, we were able to quantify uncertainty around our estimates and propagate this uncertainty into predictions of malaria transmission patterns.
Results/Conclusions
Applying our model to eSwatini surveillance data, we find that our inferences of malaria risk are highly sensitive to the choice of data and assumptions that we feed into the model. Notably, the mean estimates of the basic reproductive number under control range from 0.43 to 1.0, and the inferred proportion of imported infections varies from 0.0022 (95% CI: 0.0015 – 0.0029) to 0.57 (0.57 – 0.57). In addition, we evaluated the performance of our approach on simulated data. We find that while our model can correctly classify over 98% of cases as imported or locally-acquired, the upper bound on the accuracy of identifying “who infected whom” is 21% (19 – 23%). This limitations of spatial and temporal data for this purpose highlight the need to incorporate genetic data into estimates of malaria transmission chains in near-elimination settings.