Estimating the patterns and consequences of malaria transmission dynamics on fine spatial scales

Plasmodium falciparum is the leading cause of malaria infection and a major cause of morbidity and mortality across the globe, particularly in the African region. The burden of malaria is unevenly distributed, with some countries, districts or even households within villages harboring a disproportionally higher burden. There is an intricate relationship between the mosquito vector, humans and the parasites they carry, and how they interact with the environment. Small movements on a fine-scale lead to the patterns observed in the community. Quantifying transmission dynamics on a fine-scale, how malaria infections spread locally and the processes leading to the observed spatial and temporal distribution patterns is important for many aspects of malaria epidemiology, in particular, the design of targeted interventions against malaria, the design of studies to evaluate the effectiveness of vector control in the field, and the parameterization of mathematical models to predict the likely impact of interventions for settings where data is not available.
Mathematical and statistical models have been developed to quantify fine scale malaria transmission dynamics and investigate the effects of interventions. Since data on the spread of vectors and parasites is challenging to collect, it is not available from many endemic settings for analytic methods to provide estimates, or to validate model predictions. Due to variability between settings, findings from one setting cannot be easily generalized. There is thus a need to involve methods that can extract information from imperfect but available datasets, to make the most of the existing data sources from settings with a variety of characteristics.
The overall aim of this thesis was to use statistical and mathematical modelling approaches to characterize fine scale malaria transmission dynamics and their consequences on the measurement of heterogeneity on a local scale for targeted interventions.
Chapter 2 used an established comprehensive simulator of malaria epidemiology developed at the Swiss Tropical and Public Health Institute (Swiss TPH) to predict the proportion of malaria infections that are in mosquitoes and humans and how this varies by setting specific characteristics. A substantial proportion of infections was predicted to be in mosquitoes, to vary with setting specific characteristics, and in response to interventions. The predictions also highlighted the role of the dynamics of infections in humans and mosquitoes following the introduction or scale-up of interventions.
In Chapter 3, a statistical model which takes into account movement between houses in a village to estimate how far and where mosquitoes fly to in the presence of spatial repellents was developed. This was a secondary use of data on mosquito densities. The method evaluation using simulation showed that the model could be used as a potential tool to gain information on mosquito movement, estimating the distance between the houses the mosquitoes were repelled from and the houses they move to, the proportion of mosquitoes repelled, and the proportion of repelled mosquitoes moving to another house as opposed to somewhere outside. However, the trial data needs to contain sufficient information to be able to disentangle the effects of the underlying processes and provide accurate estimates for all the parameters. We found that additional data on the total number of mosquitoes and sufficient numbers of mosquitoes repelled were required in the case of the motivating trial. Findings from the simulations could inform the design of studies and help quantify criteria for trial settings.
In Chapter 4, a simulation method was developed and applied to data on parasite genotypes from Kilifi County, Kenya. A previous study found an interaction between time and geographical distance on the genetic difference between pairs of parasite genotypes: genetic differences were lower for pairs of parasite genotypes which were evaluated within a shorter time interval and found within a shorter geographic distance apart. A stochastic individual-based model of malaria infections, people and homesteads was developed and fitted to the genetic differences in order to investigate hypotheses and parameter values consistent with the observed interaction.
The observed interaction could be reproduced by the individual-based model. Although hypothesis about immunity to previously seen genotypes, and or a limit on the number of current infections per individual could not be ruled out, they were not necessary to account for the observed interaction. The mean geographical distance between parent and offspring infections was estimated to be 0.40km (95%CI 0.24 – 1.20), in the base model. This was the first modeling study that we know of which has attempted to estimate parameter values and test hypotheses from malaria genotyping data with a low coverage of infections in a setting with moderate transmission. The findings glean some insights on how simulation can be used in quantifying factors driving transmission, and in estimating unknown parameters when analytic methods are limited.
The work in Chapter 5 uses the simulation model developed in Chapter 4 to investigate how the method chosen, local seasonality and movement of infections influence the detection of areas of higher transmission on fine spatial scales for targeted interventions. Our findings show that the identification of hotspots was less accurate when there was a gentle decay in risk from the hotspot boundary, the hotspot was irregularly shaped, there was seasonality in the area or when the mean distance between parent and offspring infections was longer. The findings highlight the importance of setting characteristics, the choice of outcome, and method of detection on the accuracy of identifying areas of higher transmission for targeted interventions. The underlying fine scale transmission dynamics should be taken into account when performing and interpreting analyses of heterogeneity for targeted interventions.
Taken as a whole, this thesis provides information on the characteristics of transmission dynamics on a fine scale. It highlights that a substantial proportion of malaria infections are in mosquitoes, and places emphasis on the role that vectors, and humans play in the spread of infections and the implications of fine scale movement for the measurement of heterogeneity for targeted interventions. The estimates have implications for the design and evaluation of malaria control and elimination interventions.