Increasing the potential for malaria elimination by targeting zoophilic vectors

Countries in the Asia Pacific region aim to eliminate malaria by 2030. A cornerstone of malaria elimination is the effective management of Anopheles mosquito vectors. Current control tools such as insecticide treated nets or indoor residual sprays target mosquitoes in human dwellings. We find in a high transmission region in India, malaria vector populations show a high propensity to feed on livestock (cattle) and rest in outdoor structures such as cattle shelters. We also find evidence for a shift in vector species complex towards increased zoophilic behavior in recent years. Using a malaria transmission model we demonstrate that in such regions dominated by zoophilic vectors, existing vector control tactics will be insufficient to achieve elimination, even if maximized. However, by increasing mortality in the zoophilic cycle, the elimination threshold can be reached. Current national vector control policy in India restricts use of residual insecticide sprays to domestic dwellings. Our study suggests substantial benefits of extending the approach to treatment of cattle sheds, or deploying other tactics that target zoophilic behavior. Optimizing use of existing tools will be essential to achieving the ambitious 2030 elimination target.


S1 Model derivation
The model is based on a standard expression for R 0 2 2 2 0 2 ma bc R e (Anderson and May eq 14.11 adjusted per text page 400 to reflect low human mortality rate relative to latent period) m = total number of Plasmodium-susceptible mosquitoes per person. a = biting rate per mosquito per day. c = proportion of bites on infectious humans producing infection in mosquito. 2 = mosquito instantaneous background daily mortality rate. b = probability that a bite from an infectious mosquito on a human host will generate a human Plasmodium infection. γ = per day instantaneous recovery rate in human host. τ = period from acquisition of Plasmodium infection to infectiousness in mosquito / days.

Model Assumptions
The model considers as its baseline a vector population subject to lethal interventions applied via human dwellings and explores the effects of varying the probabilities of taking a human feed and the effects of adding cowshed-based interventions.
Individual vectors are assumed to bite both human and livestock hosts with a given probability per feeding cycle of selecting a human host.
Vectors are assumed to feed once per feeding cycle.
Vector mortality and host choice is assumed to be unaffected by vector age. Once infectious, vectors do not recover and become non-infectious. Juvenile density dependence effects mean that changes to the adult vector population size do not affect the number of newly-mature adults joining the population per day. The human population size is not changed as a result of the interventions being considered Parameter/variable definitions m 0 = total number of susceptible mosquitoes per person with human-related intervention in place. m z = total number of susceptible mosquitoes per person which will choose to feed on a human host Z = proportion of blood meals taken on humans. a = biting rate per mosquito per day. c = proportion of bites on infectious humans producing infection in mosquito. 2 = mosquito instantaneous background daily mortality rate in absence of interventions. b = probability that a bite from an infectious mosquito on a human host will generate a human Plasmodium infection. γ = per day instantaneous recovery rate in human host. τ = time in days from acquisition of Plasmodium infection to infectiousness in vector. Δ H = increase in the average instantaneous daily mortality rate during a feeding cycle for mosquitoes attempting to feed on a human, arising from human-related intervention, as a proportion of the rate in the absence of any intervention. Δ L = increase in the average instantaneous daily mortality rate during a feeding cycle for mosquitoes attempting to feed on a non-human host, arising from cowshed-related intervention, as a proportion of the rate in the absence of any intervention.
Given an adult vector population size in the presence of a given human-feeding related intervention, m 0 , the introduction of a new source of mortality applicable to livestock-feeding vectors will change the adult population size. With an assumed constant rate of recruitment to the adult population, the average age structured survival probabilities of an individual vector also represent the age structure of the adult population.
The average per day mortality for the vector population before introducing the livestock-related intervention is 2