Malaria is one of the world’s biggest infectious disease killers. Over 30 countries have commenced, or plan to commence, malaria elimination with support from international donors. Malaria occurs in transmission hotspots, particularly in low-endemic elimination settings, and populations at risk are spatially heterogeneous. Geospatial tools can be useful to help design, target, and monitor and evaluate elimination programmes. Rapid advances in technology and analytical methods have facilitated the spatial prediction of malaria risk and the development of spatial decision support systems, which can enhance elimination programmes through enabling more accurate and timely resource allocation. Similarly, mathematical models are useful for providing insight into the transmission dynamics of infectious diseases and for predicting the impact of different control measures. This thesis explores the application of geospatial tools and mathematical models to malaria control and elimination. The aims are to: 1) Review the epidemiology of P. vivax infections and identify current limitations and future challenges to mathematical modelling of P. vivax transmission (Chapter 3); 2) Produce baseline risk maps of P. falciparum and P. vivax malaria in Vanuatu to assist the Vanuatu elimination programme (Chapter 4); 3) Produce risk maps of P. falciparum malaria in Bangladesh to assist the national malaria control programme (Chapter 5); 4) Identify spatiotemporal patterns and major environmental drivers of malaria in Bangladesh (Chapter 6); 5) Compare the application of geospatial tools to malaria control versus elimination (Chapter 7); and 6) Make recommendations to policy makers on the use of mathematical models and geospatial tools for malaria control and elimination in the Asia-Pacific region (Chapter 7).
Chapter 3 employed standard literature review methods, including a systematic search of online databases, and a manual search of manuscript bibliographies. Data used in this thesis included prospectively-collected field survey data using rapid diagnostic tests (Bangladesh) and polymerase chain reaction (Vanuatu) for identifying infected people, as well as clinical surveillance data collected through the Bangladesh national malaria passive surveillance system. In Chapters 4 and 5, Bayesian model-based geostatistics were used for the spatial prediction of malaria risk in Tana Island (Vanuatu) and 13 malaria-endemic districts of Bangladesh. In Chapter 6, Bayesian conditional autoregressive prior models were used to model the spatiotemporal dynamics of P. falciparum and P. vivax malaria in Bangladesh. All models presented in Chapters 4–6 included covariates that described the physical environment, including elevation, distance to coastline, temperature, rainfall and forest cover. For Chapters 4 and 5, these covariates facilitated spatial prediction, and in Chapter 6 the purpose was to quantify the major environmental drivers of infection risk.
There have been few mathematical modelling studies of P. vivax malaria and such modelling studies are complicated by the lack of published data pertaining to many of the parameters defining P. vivax transmission, and the high degree of strain variability in different areas of the globe. Malaria risk was highly spatially heterogeneous in each of the three spatial epidemiological studies. In Tanna Island, Vanuatu, malaria risk was associated with proximity to the coastline and elevation, with a higher risk in low-lying coastal areas. In Bangladesh, prevalence of P. falciparum in 2- to 10-year-old children was positively associated with temperature, elevation and living in a forested area. Similarly, incidence of P. falciparum and P. vivax infections in Bangladesh was associated with rainfall and temperature, but the association between incidence and elevation differed between the two parasite species.
Malaria risk is spatially heterogeneous in control and elimination settings, providing strong justification for a spatially targeted approach to malaria interventions. The scale of heterogeneity that is relevant to interventions differs in control and elimination settings, with much higher-resolution information on the distribution of malaria risk being important for elimination compared with control. Additionally, the size of the target population in control and elimination settings differs, with elimination targeting much smaller groups of high-risk individuals. Currently, no framework exists for evaluating geospatial tools and future research is required to identify measureable indicators and quantify the impact of geospatial tools on elimination outcomes. Additionally, limited data exists to inform the development of mathematical models of P. vivax transmission. Future research should be aimed at better defining the epidemiology of this parasite, which is likely to be a major limiting factor to the ambitions for global malaria elimination.