Spatial modelling and simulation for disease surveillance

This thesis addresses the application of spatial approaches to disease surveillance and control from three different aspects. First, temporal trends, geographical distribution and spatial pattern of cardiovascular disease admission rates were explored on a national level in Sweden. Global Moran’s I was used to explore the structure of patterns and Anselin’s local Moran’s I was applied to explore the geographical patterns of admission rates. Second, the impact of landscape structure and environmental and socioeconomic conditions on disease elements were used to create susceptibility maps. Machine learning algorithms such as artificial neural networks, logistic regression and fuzzy operators were applied to model Visceral Lesihmaiasis (VL), and generate predictive risk maps for two endemic areas in northwestern Iran and southern Caucasus. Third, the dynamicity of a complex vector-borne disease, Cutaneous Leishmaniasis (CL), were simulated based on the interactions between vectors, hosts and reservoirs using agent-based modeling approaches. A Susceptible-Exposed-Infected-Recovered (SEIR) approach together with Bayesian modeling has been applied in the model to explore the spread of CL. The model is adapted locally for Isfahan Province, an endemic area in central Iran. The results from this thesis demonstrate how spatial approaches facilitate creating road maps for policy making and resource planning processes of healthcare authorities.