Childhood malaria case incidence in Malawi between 2004 and 2017: spatio-temporal modelling of climate and non-climate factors.

BACKGROUND: Malaria transmission is influenced by a complex interplay of factors including climate, socio-economic, environmental factors and interventions. Malaria control efforts across Africa have shown a mixed impact. Climate driven factors may play an increasing role with climate change. Efforts to strengthen routine facility-based monthly malaria data collection across Africa create an increasingly valuable data source to interpret burden trends and monitor control programme progress. A better understanding of the association with other climatic and non-climatic drivers of malaria incidence over time and space may help guide and interpret the impact of interventions. METHODS: Routine monthly paediatric outpatient clinical malaria case data were compiled from 27 districts in Malawi between 2004 and 2017, and analysed in combination with data on climatic, environmental, socio-economic and interventional factors and district level population estimates. A spatio-temporal generalized linear mixed model was fitted using Bayesian inference, in order to quantify the strength of association of the various risk factors with district-level variation in clinical malaria rates in Malawi, and visualized using maps. RESULTS: Between 2004 and 2017 reported childhood clinical malaria case rates showed a slight increase, from 50 to 53 cases per 1000 population, with considerable variation across the country between climatic zones. Climatic and environmental factors, including average monthly air temperature and rainfall anomalies, normalized difference vegetative index (NDVI) and RDT use for diagnosis showed a significant relationship with malaria incidence. Temperature in the current month and in each of the 3 months prior showed a significant relationship with the disease incidence unlike rainfall anomaly which was associated with malaria incidence at only three months prior. Estimated risk maps show relatively high risk along the lake and Shire valley regions of Malawi. CONCLUSION: The modelling approach can identify locations likely to have unusually high or low risk of malaria incidence across Malawi, and distinguishes between contributions to risk that can be explained by measured risk-factors and unexplained residual spatial variation. Also, spatial statistical methods applied to readily available routine data provides an alternative information source that can supplement survey data in policy development and implementation to direct surveillance and intervention efforts.