Deriving risk maps from epidemiological models of vector borne diseases: State-of-the-art and suggestions for best practice.

Epidemiological models (EMs) are widely used to predict the temporal outbreak risk of vector-borne diseases (VBDs). EMs typically use the basic reproduction number (R), a threshold quantity, to indicate risk. To provide an overall view of the risk, these model outputs can be transformed into spatial risk maps, using various aggregation methods (e.g. average R over time, cumulative number of days with R > 1). However, there is no standardized methodology available for this. Depending on the specific aggregation methods used, the yielded spatial risk maps may have considerably different interpretations. Additionally, the method used to visualize the aggregated data also affects the perceived spatial patterns. In this review, we compare commonly used aggregation and visualization methods and discuss the respective interpretation of risk maps. Research publications using epidemiological modelling methods were drawn from Web of Science. Only publications containing maps of R transformed from EMs were considered for the analysis. An example EM was applied to illustrate how aggregation and visualization methods affect the final presentations of risk maps. Risk maps can be generated to show duration, intensity and spatio-temporal dynamics of potential outbreak risk of VBDs. We show that 1) different temporal aggregation methods lead to different interpretations; 2) similar spatial patterns do not necessarily bear the same meaning; 3) visualization methods considerably affect how results are perceived, and thus should be applied with caution. We recommend mapping both intensity and duration of the VBD outbreak risk, using small time-steps to show spatio-temporal dynamics when possible.