5. Discussions
Previous studies focused on the disease prevalence, incidence rate, and mortality rate as regression outcome variables for identifying spatial risk factors (Hu, Clements, Williams, Tong, & Mengersen, 2012; Naish et al. 2014). However, these models ignored the spatial and temporal interactions among areas. This study constructed OD pairs as the unit of analysis, instead of using administration units, to explore the spatial and temporal relationships of epidemic processes. We investigated two important epidemic characteristics of dengue fever: the risk of diffusion and the transmission speed. Our significant results indicated that (1) the risk of diffusion is correlated with the socioeconomic status of townships where imported dengue cases emerge, and the risk of diffusion to geographically neighboring townships is higher than to remote townships; (2) temporal variability may influence the speed of transmission from imported townships to local townships, and there is a significant relationship between the time lags from imported cases to local emergence in late spring; and (3) the urbanization levels also influence the transmission speed from imported townships to local townships, showing that shorter time lags were concentrated in highly urbanized townships.