Spatiotemporal methods and applications in neglected tropical diseases

Abstract

In nature, diseases show spatial and temporal variations. The probability of acquiring diseases decline along with increasing distances and time lags. Spatial, temporal and spatiotemporal predictive models are useful to make recommendations in low-resource settings when disease registries are either non-existent or geographically incomplete. The National Snakebite Survey (NSS) was an island-wide, community based survey conducted over 11 months. The survey reported snakebites that occurred in the preceding 12 months, and the number of people surveyed in each month varied. Snakebite showed a spatial correlation between events. Spatial variation of snakebites at cluster level was assessed using purely spatial models after collapsing the time dimension and accounting for spatial correlation between events. Temporal variation of snakebites at each month was assessed using purely temporal models after collapsing the spatial dimension and accounting for recall bias and survey effort. Subsequently, the spatial and temporal models were combined to explain spatiotemporal variation incoiporating individual-level data. Overall, snakebites and envenoming bites showed clear spatial and temporal variations in Sri Lanka. Snakebites were higher in intermediate and wet zones. However, envenomings were higher in the dry zone. At the national level, the highest snakebite incidence was observed from November to December, whilst the highest envenoming incidence occurred from March to April. Snakebite and envenoming hotspots showed dynamic changes throughout the year. The average monthly snakebite and envenoming incidence in Sri Lanka were 39 and 19 per 100,000 respectively. This translates into 110,000 snakebites and 45,000 envenomings each year. The models that were developed were then used to demonstrate the residual confbunding effect and explain the health-seeking behaviour of snakebite victims. Health seeking behaviour of snakebite victims depended on probable envenoming caused by bites. Victims in high endemic areas fbr envenoming sought allopathic treatment over traditional treatment and vice versa. The thesis also presents a case study to illustrate spatial modelling in the absence of spatial correlation between events. The thesis presents some solutions to problems encountered in spatial, temporal and spatiotemporal epidemiology. These results would facilitate informed healthcare decision making in Sri Lanka, and the methods described are replicable for studies of a similar nature.