Spatial and temporal distribution of severe malaria in western Kenya highlands: the effects of climatic fluctuations and landscape characteristics

Abstract

From the 1980s, series of malaria outbreaks rated as severe have occurred in the highly populated areas of western Kenya highlands at altitudes above 1300m. The residents here have low or no immunity to malaria because they have not been previously exposed. Such outbreaks can be epidemic and with unclear spatial and temporal distribution patterns. The linkages between the severe malaria distribution and the major factors that influence its spatial or temporal unclear distribution patterns have not been explicitly clarified. A clear demonstration of when and where such cases occur should precede any successful and economic severe malaria intervention or control program. Severe malaria cases reported between 1990 and 2004 to St. Elizabeth Hospital Mukumu were used in this work. Through Cullen’s epidemic model, epidemic and non-epidemic months and years were detected during this period. Cases in one selected epidemic year (1997) and one non-epidemic year (2004) that were within a 30km travelling distance from the hospital were mapped using a handheld Global Positioning Systems (GPS) unit and also through on-screen digitizing of georeferenced maps using Arcview program in Geographic Information Systems (GIS). Geostatistical tools in ArcGIS 9.3 were used for Point Pattern Analysis (PPA) which illustrated spatial homogeneity and heterogeneity of the cases in the two years. Lastly, Geospatial techniques were used to relate the distribution patterns of the cases and factors that majorly influence malaria outbreaks. These were climatic data obtained from local meteorological stations and environmental data captured through Remote Sensing techniques and cartographic skills. The influence of changes in temperature, rainfall, altitude, land cover and distance to hospital were tested for the two years using multiple linear regression. Relative risk of infections per 1000 persons in adjusted population projections and population structure and malaria cases were also calculated. XVII In the selected epidemic year(1997), higher total rainfall amount (1754mm) and higher average maximum temperature (27.98 °C ) resulted into a 1.7 fold increase in severe malaria cases (n = 3118) compared to the selected nonepidemic year (2004) (n= 1860) when total rainfall amount received was 1549mm and the average maximum temperature was 27.52 °C. Only 56 (31%) out of 180 months were epidemic between 1990 and 2004. No global clustering was detected in both years but there were highly dispersed local clusters, especially during rainy seasons (R2= - 0.02, Z score = 3.2, p > 0.01 in 1997 and R2= 0.005, Z-score 4.9, p = 0.01 in 2004). Distance to hospital was found to influence clustering within 7.5km radius from the hospital (z = 16.2 km) and this declined with increased distance from the hospital. Age structures compared to adjusted populations showed that most were children below 5 years because of their numbers. Malaria cases were found to be more clustered near water surfaces and farmlands and less in forested areas and shrublands. Number of malaria cases was on the rise in the newly deforested areas and reclaimed swamps. Heterogeneity of the severe malaria cases can be better explained using detailed epidemiological data that are tested through a battery of existing Geospatial techniques. They can elucidate spatial and temporal malaria distribution, improving the much needed targeting and lower the cost of severe malaria intervention and management.