Big city, small world: Density, contact rates, and transmission of dengue across Pakistan.

Moritz U. G. Kraemer; T. Alex Perkins; Derek A.T. Cummings; Rubeena Zakar; Simon I. Hay; David L. Smith; Robert C. Reiner

To disentangle the different aspects of dengue dynamics and their drivers we used a model 218 containing only the climatological covariates and performing backwards model selection until 219 each covariate in the model was significant at a 5% level resulted in a model that explained 220 76.9% of the deviance and whose adjusted R-squared was 0.746. Amongst the yearly-221 averaged covariates, EVI and precipitation remained in the model as well as the derived Ae. To understand these differences the final model was then compared to a nested model where 235 the coefficient for Lahore was allowed to vary independently of all other districts. Deviance 236 explained increased to 77% and adjusted R-squared increased to 0.753. Further, the mixing 237 coefficient for Lahore ( = 0.74) was significantly larger than the mixing coefficient for the 238 other districts ( = 0.59, p-value=0.0068) ( Figure S1 ). The median R0 for Lahore was 239 estimated at 3.28, the highest for all districts. 240