The role of seasonal population movements in malaria transmission and control in sub-Saharan Africa

Abstract

Despite considerable efforts, malaria control programmes are struggling to reach important mile- stones on the path to global elimination. Human mobility poses an important challenge to malaria control and is increasingly being accounted for when planning interventions. However, the effect of seasonality of human travel on malaria dynamics and interventions has not yet been adequately evaluated. This thesis aims to quantify the seasonality of human mobility, investigate its impact on the spread of malaria and identify potential spatio-temporal interven- tion targets. In addition, we investigate the suitability of mobile phone data and mathematical mobility models to describe and synthesise seasonal travel data.

We analysed a Namibian mobile phone dataset and extracted spatio-temporally detailed mo- bility data via a series of different methods. We then fitted mobility models, namely a gravity model and various radiation models, at different spatial scales, to spatio-temporally aggregated mobility data that was freely available for Namibia and Kenya. Next, we fitted mobility mod- els to Namibian movement data stratified by season and trip duration. For the first time, we successfully adapted mobility models to simultaneously fit trip counts and durations. Finally, we quantified travel-related parasite movement accounting for the spatio-temporal dynamics of travel and malaria.

The findings presented in this thesis confirm that Namibian travel patterns are highly seasonal; moreover, we found trip duration and distance to follow power-law distributions. Although peak travel periods do not overlap with the main transmission period in Namibia, this might be different in other settings. Accounting for seasonal mobility patterns and trip duration in models of malaria transmission proved to be practical and highlighted several potential targets for spatio-temporal interventions in Namibia. Mobility models have proved to be sufficiently flexible to be adapted to a range of practical applications; however, the effect of different input data on parameter estimates has to be further investigated.