Risk assessment of dengue fever in Zhongshan, China: a time-series regression tree analysis

Dengue fever (DF) is the most prevalent and rapidly spreading mosquito-borne disease globally. Control of DF is limited by barriers to vector control and integrated management approaches. This study aimed to explore the potential risk factors for autochthonous DF transmission and to estimate the threshold effects of high-order interactions among risk factors. A time-series regression tree model was applied to estimate the hierarchical relationship between reported autochthonous DF cases and the potential risk factors including the timeliness of DF surveillance systems (median time interval between symptom onset date and diagnosis date, MTIOD), mosquito density, imported cases and meteorological factors in Zhongshan, China from 2001 to 2013. We found that MTIOD was the most influential factor in autochthonous DF transmission. Monthly autochthonous DF incidence rate increased by 36.02-fold [relative risk (RR) 36.02, 95% confidence interval (CI) 25.26-46.78, compared to the average DF incidence rate during the study period] when the 2-month lagged moving average of MTIOD was >4.15 days and the 3-month lagged moving average of the mean Breteau Index (BI) was 16.57. If the 2-month lagged moving average MTIOD was between 1.11 and 4.15 days and the monthly maximum diurnal temperature range at a lag of 1 month was <9.6 degrees C, the monthly mean autochthonous DF incidence rate increased by 14.67-fold (RR 14.67, 95% CI 8.84-20.51, compared to the average DF incidence rate during the study period). This study demonstrates that the timeliness of DF surveillance systems, mosquito density and diurnal temperature range play critical roles in the autochthonous DF transmission in Zhongshan. Better assessment and prediction of the risk of DF transmission is beneficial for establishing scientific strategies for DF early warning surveillance and control.