Geospatial analysis of environmental risk factors of leptospirosis in East Africa. A multiscale, mixed method approach in spatial epidemiology

Leptospirosis is one of the most widespread, but neglected, zoonoses in the world. The disease is caused by infection with pathogenic Leptospira species spread by the urine of mammalian hosts. It is an acute febrile illness with a wide variety of clinical manifestations and, in recent years, has come to international attention as a major re-emerging infectious disease. Leptospirosis has a broad, almost ubiquitous, geographic distribution. However, the highest incidence of leptospirosis is observed throughout tropical, low- and middle-income regions of the world where the disease is considered a major cause of non-malarial febrile illness, especially throughout rural environments. Leptospirosis is considered endemic across East Africa, however, due to a scarcity of robust data on leptospirosis occurrence in humans or animals, awareness of the illness is relatively low and there are few disease burden estimates from the region. As such, the current status of leptospirosis in East Africa remains largely unknown. Effective targeting of leptospirosis control strategies throughout areas thought to experience a high burden of infection rely on detailed and accurate understanding of the geographical and temporal distribution of disease, the population at risk and the intensity of disease transmission. Such information can be generated using spatially explicit geostatistical methods which explore linkages between the distribution and intensity of leptospirosis occurrence and environmental risk factors known, or hypothesised, to have an influence on the distribution of the disease. The spread of the Leptospira bacteria relies on the abundance and geographic distribution of suitable animal maintenance hosts, such as rodents and livestock. Once shed into the environment, the survival of the bacteria is linked to a suite of climate, hydrological and terrestrial environmental processes. In general, the prevailing environmental conditions in East Africa are favourable for active transmission of Leptospira. However, significant knowledge gaps remain in relation to the specific epidemiology and environmental determinants of infection risk across the region at broad and local scales. In that context, the objective of this thesis is to examine the spatiotemporal patterns of human and animal exposure to leptospirosis in urban and rural communities of East Africa and explore potential environmental drivers of infection. To achieve this objective, and to extract maximal value from limited data availability, a novel mixed method approach in GIS-based spatial epidemiology was devised. Three distinct bodies of evidence-based research relevant to the spatial epidemiological characteristics of leptospirosis were carried out. Crucially, both data- and knowledge-driven methodological approaches were implemented in context specific case studies of investigation. Firstly, data from a geospatially explicit longitudinal study were used to determine the fine scale environmental covariates of leptospirosis within a canine sample population in a small informal settlement located in Nairobi, Kenya. High spatial resolution remotely sensed datasets and serological data were coupled within a multilevel mixed effect geostatistical modelling approach to examine environmental risk factors and generate hypotheses concerning the likely sources and transmission pathways of infection in dogs. Highly localised terrestrial landscape variables (e.g. slope) and broad scale climate regimes (i.e. precipitation) were found to demonstrate statistically significant influences on canine exposure to environmental Leptospira. Next, a comprehensive literature review was carried out to collect evidence to parameterise and drive a knowledge-based modelling approach within a GIS-based framework. A pragmatic spatially explicit multi-criteria decision analysis (MCDA) approach was adopted to derive predictions of the distribution of suitability for leptospirosis in an area of northern Tanzania. Outputs from the spatial MCDA indicate that large areas of the study area may be suitable for leptospirosis, however, predicted risk was highly spatially heterogeneous. Areas with high suitability for leptospirosis were estimated to occur primarily throughout rural areas dominated by agro-pastoral communities. Temporal trends in predicted relative risk were observed using monthly georeferenced climate data to allow for exploration of the seasonal patterns of disease. Finally, geospatial cluster detection techniques and an innovative multiscale cross-sectional non-spatial logistic regression analysis approach were combined to explore patterns of, and risk factors for human acute leptospirosis among patients presenting with fever in northern Kilimanjaro Region, Tanzania. Statistically significant hotspots of leptospirosis occurrence were identified, predominately in rural areas thought to be enzootic, and infection risk linked to several environmental processes, including soil properties, topographic profile and dominant land use. Crucially however, associations were not consistent between scales of analysis indicating that observed relationships are highly dependent on the choice of spatial extent used in the analysis. The truly multi-discipline approach adopted within this thesis demonstrated the importance of considering multiple lines of inquiry in order to fully address knowledge gaps concerning the environmental aetiology of leptospirosis. In sum, the findings presented throughout this thesis highlight the importance of a suite of broad and local scale environmental risk factors of leptospirosis and describe the spatiotemporal distribution of the disease throughout a resource poor region of the world where the infection is a major cause of illness. Crucially, the results of this study can be used to guide and improve the design of more effective and targeted surveillance and control strategies, reducing the burden of this neglected disease, which is otherwise likely set to increase in the face of human and livestock population growth and climate change. Perhaps most significantly, by enhancing the knowledge base regarding the spatial determinants of the distribution of leptospirosis throughout endemic areas, these findings can be used to inform future education and awareness strategies relevant to exposure risks of leptospirosis throughout the region. For example, this thesis presents leptospirosis as primarily a disease of rural communities, thus the outcomes presented here can be used to inform targeted livestock management and husbandry practices which are considered key components of the transmission cycle of leptospirosis. Moreover, these results should be used to advise community programs designed to raise greater awareness of the risks associated with exposure to contaminated environments during the wet seasons by recommending the provision and use of personal protection equipment for those engaged in rural occupations considered high risk of exposure to leptospirosis. In summary, the continued development of traditional and novel mathematical and geostatistical models is crucial in order to further enhance our understanding of Leptospira epidemiology and provide a valuable framework through which to inform veterinary and human public health practitioners and policy makers in the development of future intervention strategies