The temporal and spatial distribution of malaria in Africa, with emphasis on Southern Africa

The three-way relationship between the Plasmodium parasite, the Anopheles mosquito vector and the human host determines the incidence of malaria disease. The three life cycles, the interactions respectively between human and parasite, human and mosquito, and mosquito and parasite, and the ultimate transmission cycle, vary in time and space. Environmental, genetic and behavioural factors influence the three life cycles and the interactions. These factors also vary in time and space. At every level the variation itself, whether random or cyclical, is not uniform but varies in frequency and magnitude. Explaining, and particularly predicting, malaria transmission rates in time and space thus becomes a difficult undertaking.
Knowing and understanding some of this variation, and its causes, is important for well-timed and well-targeted malaria interventions. In the fringe areas of malaria in Africa, which are prone to epidemics, some forewarning of unusually high incidence periods would be valuable to malaria control and management services.
This thesis investigated the temporal and spatial effects on malaria transmission of various environmental factors, particularly climate, and of non-climatic factors, particularly those relating to malaria control. Different data sets and methodological approaches were applied in seven separate studies, and malaria distribution in time and space was investigated at different scales.
At the continental scale, the distribution of malaria in Africa was modelled as a factor of climate using raster GIS techniques.
At the national scale, using prevalence data from Botswana, spatial variation in prevalence was modelled as a factor of environmental determinants, prior to comprehensive malaria control. The spatial and inter-annual variation in prevalence, in the presence of intense control, was also modelled as a factor of climate.
At the sub-national level South Africa was used as an example. Inter-annual variation in malaria incidence in the highest-risk province was explored for possible links with climatic and non-climatic factors. Finally, inter-annual and spatial variation in sub-provincial level incidence data for South Africa, were analysed with respect to climatic and non-climatic determinants, for which data were available.
The two study areas (Botswana and South Africa) both lie at the fringe of malaria distribution, experience strongly seasonal transmission and epidemics, and both benefit from intensive malaria control. The two study areas represent two slightly different scenarios: in Botswana the analysis period covered the steady introduction of comprehensive control, while in South Africa the study period covered a time when effective control was being threatened by the spread of insecticide- and drug resistance, and the general health of the population was increasingly affected by the HIV pandemic.
The main findings were the following:
• It was possible to estimate the distribution of malaria in Africa fairly successfully from long term mean climate data via simple GIS methods. The model compared well with contemporary malaria data and historical ‘expert opinion’ maps, excepting small-scale ecological anomalies. The model provided a numerical basis for further refinement and prediction of the impact of climate change on transmission. Together with population, morbidity and mortality data, it has provided a fundamental tool for strategic control of malaria.
• In Botswana the spatial variation in childhood malaria prevalence, prior to intense comprehensive control, was significantly associated with underlying environmental factors. It could be predicted and mapped using only three environmental predictors, namely summer rainfall, mean annual temperature and altitude. After starting with a long list of candidate variables, this parsimonious model was achieved by applying a systematic and repeatable staged variable exclusion procedure that included a spatial analysis. All this was accomplished using general-purpose statistical software.
• In the presence of intense control, the spatial and temporal variability in childhood malaria prevalence in Botswana could no longer be explained by variation in climate. The effects of malaria control and good access to treatment seem to have replaced climate as the main determinant of prevalence. This also suggests that prevalence, a less direct measure of transmission rate, is more prone to non-climatic effects than incidence rate.
• Total population malaria incidence in KwaZulu-Natal, the highest risk province of South Africa, remained significantly influenced by climate over a 30 year period, even in the presence of intense control. The inter-annual variation in case numbers were significantly associated with several climate variables, mainly mean annual daily temperatures and summer rainfall. However, climate factors did not explain the longer term total incidence rates.
• The longer term trends in total malaria incidence in KwaZulu-Natal province, over the same 30 years period, were significantly associated with the spread of anti-malarial drug resistance and HIV prevalence. Cross-border movements of people, agricultural activities and emergence of insecticide resistance also affected the level of malaria transmission at certain periods and to some degree, but this could not be formally quantified.
• When considering malaria incidence in three malarious provinces of South Africa at a sub-provincial level, the observed temporal and spatial variation could largely be explained by available weather, HIV prevalence and drug-resistance data. However, much of the region-specific temporal trends remained unexplained. Temporal forecasts, based on 18 years of data, predicted for six years for six regions, were not very accurate and lacked precision. It seems that the interplay of climatic and non-climatic factors in the South African context is too complex to allow forecasts that are suitable for decision-making at the provincial level.
• The findings of this thesis emphasize that in addition to shorter-term variation, which seems to be driven by climate in many cases, malaria transmission is largely determined by non-climatic factors in southern Africa. This appears to be particularly true where the natural malaria endemicity has been modified by control interventions. As the drive to control malaria in Africa continues and intensifies, the need for long-term surveillance of not merely malaria transmission, but also of the coverage and effectiveness of control interventions, will grow.