Malaria risk mapping and prediction in Cote d'Ivoire

[摘要] ENGLISH ABSTRACT: Malaria, being the most important vector-borne disease in Africa, is still leaving an ever-increasingtrail of health and economic impediments within the developing world. Sincemalaria's spatial distribution and intensity are influenced by numerous environmentalvariables, medical geography, the study of the spatial human-environmentalinterrelationship of disease, can make a significant contribution to modelling disease riskfor different geographic locations, as based on environmental determinants. Theinvestigation consequently included the development of malaria risk maps withinnorthern Cote d'Ivoire, an analysis of its spatial variation through exploratory spatial dataanalysis and Geographical Information System (GIS) functionality, and the search forpredictive models using environmental variables.Malaria transmission intensities of the Anopheles fonestus and Anopheles gambiaemosquito species have been studied as factors having a direct influence on the malariatransmission cycle, whilst indirect environmental influencing factors which were studiedincluded: wetland rice production, water bodies and their flooding pattern, temperatureand relative humidity ranges, amount of rainfall and insolation, distances between aquaticbreeding grounds and villages, and vegetation cover. Data have been obtained fromongoing surveys carried out by Institut Pierre Richet/OCCGE and the WARDA HealthResearch Consortium within a random selection of 24 villages in northern Cote d 'Ivoire,district level meteorological observations, and GIS and satellite imagery computations.The data sets have been manipulated, standardised and reduced to three malaria riskindicators, and five uncorrelated environmental measures through principal componentanalysis.The application of exploratory spatial data analysis, in which the spatial capabilities ofGIS played a significant role, involved:a) Mapping the disease's spatial distribution and extreme risk areas by means ofproportional circle and probability maps. b) Exploring and modelling first and second order spatial disease variation throughTriangulated Irregular Network (TIN) interpolation, semi-variograms, and trendsurface analysis.c) Visual and statistical detection of associations between malaria and environmentaldeterminants, by means of proportional circle maps and multiple regressionanalysis, in the search for predictive models. Residuals of final models have beenmapped to investigate possible patterns in over- and underestimation.A general increase in disease incidence and transmission levels during the study period,as well as areas of high risk have been identified. Investigation of first order variationhighlighted these high risk areas. Semi-variograms revealed a high degree of spatialvariability, with little or no spatial dependence between observations. No clear secondorder variation could therefore be identified, nor any reasonable first or second orderspatial variation model be deduced. Relative humidity and temperature measures appearvisually to be associated with malaria occurrence, and were statistically confirmed toaccount for a substantial percentage of disease variance through regression modeling,together with distance and vegetation density measures. However, the small proportionof variance explained by some models, together with the irregular spatial distribution andhigh levels of residuals, made reliable prediction impossible. It is therefore clear that therelationships between malaria morbidity or transmission and the environmental factorsare complex and often highly site specific, possibly requiring higher order polynomialfunctions and a wide spectrum of determinants, still to be identified and included insubsequent studies.

[发布机构] Stellenbosch University