Deforestation and malaria in Madagascar: a new framework to explore linkages in the absence of robust health reporting infrastructure

Abstract

BackgroundLand-use changes, such as deforestation, can increase the transmission potential of a landscape by increasing the abundance of preferred microhabitats for malaria vectors. However, this relationship between malaria and deforestation is complex and depends on factors such as the type of land use, the local ecology of malaria vector species, and the health resources available to the human population. In Madagascar, despite alarmingly high rates of deforestation and concerning reports of increases in malaria, this link remains poorly studied. MethodsThe feasibility of traditional approaches to studying the link between malaria and deforestation is limited for Madagascar because (i) health systems lack capacity to collect regular and accurate malaria data, especially in many rural areas, (ii) vector habitat preferences are poorly understood, and (iii) most existing malaria infection outcome datasets lack data on important mediating factors. To circumvent inadequacies in current data streams, we discuss a new framework that integrates novel primary data collection efforts on malaria infection and vector ecology from cross-sectional studies with longitudinal land-use data from remote sensing. Incorporating these data into hierarchical models allows us to infer structural associations between macro and micro level predictors of malaria risk in different ecological contexts in Madagascar. We compare the outputs from these models to analyses that can be conducted with robust data from the Brazilian health system. FindingsWe report data on malaria burden, deforestation, its covariates, and vector larval habitat preference for four distinct, previously understudied, ecoregions of Madagascar. Risk factors of malaria infection vary in sign and magnitude of effect across ecological contexts. InterpretationWhile data such as Brazil's allow for predictive modelling and assessment of temporally dynamic systems, our proposed modelling framework demonstrates variations in risk factors regionally in the meantime. This approach can help to mitigate current data gaps and better inform existing malaria control efforts in Madagascar during health system development. FundingThe Ren Che Foundation, Henry David Thoreau Foundation, Catholic Relief Services through a grand from United States Agency for International Development, and Madagascar Health and Environmental Research (MAHERY).