Spatial epidemiology of parasitic infections and optimal survey design

Recent years have seen a shift towards integrated control of a number of tropical diseases. Such a strategy, however, relies on an understanding of the spatial distribution and overlap of different diseases. Using a combination of fieldwork, spatial and economic analyses and computerized simulations, optimal survey designs were explored for soil-transmitted helminths (STHs), Schistosoma mansoni and Plasmodium /alciparum in East Africa, and the potential of an integrated survey approach was evaluated. For STH, analysis indicated that hookworm clusters over larger scales than Ascaris lumbricoides and Trichuris trichiura, and that surveying small numbers of children, from four to five schools per district, provides a rapid and cost-effective approach to target treatment at district levels. For S. mansoni, Lot Quality Assurance Sampling (LQAS) was compared to a geostatistical survey design that allows spatial prediction at unsurveyed locations based on a subset of schools. Results showed that targeted treatment was more cost-effective than presumptive treatment and that, whilst LQAS correctly classified a higher proportion of schools requiring treatment, a geostatistical design proved more cost-effective. An investigation into the optimal spatial scale to conduct surveys for STH, S. mansoni and P. /alciparum in Kenya found that, over various cost scenarios, surveying fifty children from three randomly selected sites per sub-district provided a balance of performance and cost-effectiveness for all species. In sub-districts of low S. manson; and P.alciparum prevalence, LQAS should be used to target treatment. This thesis has shown that species-specific differences in spatial heterogeneity of infection and the costs of both mapping surveys and programme intervention have important implications for the optimal design of surveys. A two stage framework for integrated surveys is proposed allowing for a flexible approach to mapping. Similar studies in different settings are crucial and would help to assess whether changes in survey strategy are required as transmission drops due to control activities.