Use of generalized linear mixed models in the spatial analysis of small-area malaria incidence rates in KwaZulu Natal, South Africa

Spatial statistical analysis of 1994-1995 small-area malaria incidence rate s in the population of the northernmost districts of KwaZulu Natal, South A frica, was undertaken to identify factors that might explain very strong he terogeneity in the rates. In this paper, the authors describe a method of a djusting the regression analysis results for strong spatial correlation in the rates by using generalized linear mixed models and variograms. The resu lts of the spatially adjusted, multiple regression analysis showed that mal aria incidence was significantly positively associated with higher winter r ainfall and a higher average maximum temperature and was significantly nega tively associated with increasing distance from water bodies. The statistic al model was used to produce a map of predicted malaria incidence in the ar ea, taking into account local variation from the model prediction if this v ariation was supported by the data. The predictor variables showed that eve n small differences in climate can have very marked effects on the intensit y of malaria transmission, even in areas subject to malaria control for man y years. The results of this study have important implications for malaria control programs in the area.