Malaria risk in Nigeria: Bayesian geostatistical modelling of 2010 malaria indicator survey data.

ABSTRACT

BACKGROUND: In 2010, the National Malaria Control Programme with the support of Roll Back Malaria partners implemented a nationally representative Malaria Indicator Survey (MIS), which assembled malaria burden and control intervention related data. The MIS data were analysed to produce a contemporary smooth map of malaria risk and evaluate the control interventions effects on parasitaemia risk after controlling for environmental/climatic, demographic and socioeconomic characteristics. METHODS: A Bayesian geostatistical logistic regression model was fitted on the observed parasitological prevalence data. Important environmental/climatic risk factors of parasitaemia were identified by applying Bayesian variable selection within geostatistical model. The best model was employed to predict the disease risk over a grid of 4 km(2) resolution. Validation was carried out to assess model predictive performance. Various measures of control intervention coverage were derived to estimate the effects of interventions on parasitaemia risk after adjusting for environmental, socioeconomic and demographic factors. RESULTS: Normalized difference vegetation index and rainfall were identified as important environmental/climatic predictors of malaria risk. The population adjusted risk estimates ranges from 6.46% in Lagos state to 43.33% in Borno. Interventions appear to not have important effect on malaria risk. The odds of parasitaemia appears to be on downward trend with improved socioeconomic status and living in rural areas increases the odds of testing positive to malaria parasites. Older children also have elevated risk of malaria infection. CONCLUSIONS: The produced maps and estimates of parasitaemic children give an important synoptic view of current parasite prevalence in the country. Control activities will find it a useful tool in identifying priority areas for intervention.