Download (direct link):
Disease mapping has a long history (Howe,
1989), and the early studies undertaken in this
area exemplify the power of the method in defining the environmental and social aetiology of a specific disease (Snow, 1854; Palm, 1890). Yet, it is only recently that disease mapping has become integral to the study of infectious disease epidemiology and control (Mott et al., 1995). Two major technological advances underlie this development. The first is the advent of powerful and affordable computer mapping systems in the 1980s (Openshaw, 1996). Known as geographical information systems (GISs), such computer software packages permit the capture, storage, analysis and display of any and all types of geographical reference data. The second technological innovation concerns the increasing availability and ability to integrate data from remote sensors (RS) based on space platforms within GISs, which allow the investigation of disease co-distribution with environmental variables at various spatial scales (Openshaw, 1996). The value of stand-alone GIS and integrated GIS/RS-based approaches, not only to gain understanding of the spatial distribution of infection but also to aid the design and implementation of control programs, has been
demonstrated recently for a range of parasitic diseases, including malaria (Beck et al., 1994; Thomson et al., 1997; Omumbo et al., 1998; Snow et al., 1999). African trypanosomiasis (Rogers and Williams, 1993), onchocerciasis (Richards, 1993) and dracunculiasis (Clarke et al, 1991).
This chapter aims to describe the use of GIS and integrated GIS/RS approaches to understanding infectious disease distribution and control, using examples from work carried out on African trypansomiasis, and on lymphatic filariasis among helminth parasites. Readers are referred to the review by Mott and colleagues
(1995) for descriptions of applications to other tropical parasitic diseases.
The geographic approaches undertaken for this disease illustrate how predicting the distribution of vectors using remotely sensed data on associated environmental co-variables can help to define areas of vector-borne disease transmission. The main utility of these studies has been to demonstrate the potential of remotely sensed satellite data in uncovering vector-environmental relationships relevant to mapping the co-distribution and spread of vectors and the disease they cause (Hay et al., 1997). Thus, Kitron et al.
(1996) analysed tsetse fly catches from sets of traps set in the Lwambe Valley of Western Kenya during 1988-1990, and found that high resolution Landsat Thematic Mapper (TM) imagery data were able to explain most of the variance in fly catch density. In particular, wavelength band 7 of the Landsat-TM imagery, which is associated with soil-water content, was found to be consistently highly correlated, reflecting the importance of soil moisture in tsetse survival.
By contrast, Rogers and Randolph (1991,
1994) explored the utility of Global Area Coverage (GAC) normalized difference vegetation index (NDVI) data, derived from the National Oceanic and Atmospheric Administration’s (NOAA) Advanced Very High Resolution Radiometer (AVHRR), as a proxy for studying tsetse fly ecology and distribution in West Africa, since they considered the NDVI to integrate a
variety of environmental factors of importance to tsetse survival. They found an inverse relationship between monthly NDVI and fly mortality rate in the Yankari game reserve in Nigeria, and significant non-linear relationships between tsetse fly abundance and NDVI in the northern part of Cote d’Ivoire. They focused on a 700 km transect running north-south through Cote d’Ivoire and Burkina Faso. This area is of particular epidemiological interest, since sleeping sickness is found only in the central region of the transect, despite the local vector (Glossina palpalis) occurring throughout. The analysis showed that this focalized transmission was a result of differences in overall fly size. During the wet season, the NDVIs across the transect were all high and fly size was uniformly large. In the dry season, however, fly size was strongly correlated with NDVI, with flies in the drier north significantly smaller than those in the wetter south. Since mortality increases with decreasing fly size in tsetse, these data were interpreted as indicating a geographical gradient in the degree of man-fly contact, and thus trypanosome transmission potential. In the south, low mortality rates resulted in high densities of flies, but the flies were not nutritionally stressed (even seasonally) and so did not often resort to biting humans, who are not favoured hosts. Conversely, in the north, fly populations suffered too high a mortality to pose a serious health risk. Only in the central areas was there an intermediate density of sufficiently stressed flies, resulting in a regional and seasonal focus of disease transmission. This study showed that, although at relatively small spatial scales both tsetse distribution and abundance and disease incidence and prevalence could be related to the low-resolution NDVI, the interpretation of the data required a knowledge of local conditions and fly biology from ground studies.