A statistical modelling framework for mapping malaria seasonality

Many malaria-endemic areas experience seasonal fluctuations in cases because the mosquito vector's life cycle is dependent on the environment. While most existing maps of malaria seasonality use fixed thresholds of rainfall, temperature and vegetation indices to find suitable transmission months, we develop a spatiotemporal statistical model for the seasonal patterns derived directly from case data. A log-linear geostatistical model is used to estimate the monthly proportions of total annual cases and establish a consistent definition of a transmission season. Two-component von Mises distributions are also fitted to identify useful characteristics such as the transmission start and end months, the length of transmission and the associated levels of uncertainty. To provide a picture of "how seasonal" a location is compared to its neighbours, we develop a seasonality index which combines the monthly proportion estimates and existing estimates of annual case incidence. The methodology is illustrated using administrative level data from the Latin America and Caribbean region.