Epidemiology of dengue and the effect of seasonal climate variation on its dynamics: a spatio-temporal descriptive analysis in the Chao-Shan area on China's southeastern coast

Abstract

Objective: Dengue is a mosquito-transmitted virus infection that remains rampant across the tropical and subtropical areas worldwide. However, the spatial and temporal dynamics of dengue transmission are poorly understood in Chao-Shan area, one of the most densely populated regions on China's southeastern coast, limiting disease control efforts. We aimed to characterise the epidemiology of dengue and assessed the effect of seasonal climate variation on its dynamics in the area.

Design: A spatio-temporal descriptive analysis was performed in three cities including Shantou, Chaozhou and Jieyang in Chao-Shan area during the period of 2014-2017.

Setting: Data of dengue cases of three cities including Shantou, Chaozhou and Jieyang in Chao-Shan area during 2014-2017 were extracted. Data for climatic variables including mean temperature, relative humidity and rainfall were also compiled.

Methodology: The epidemiology and dynamics of dengue were initially depicted, and then the temporal dynamics related to climatic drivers was assessed by a wavelet analysis method. Furthermore, a generalised additive model for location, scale and shape model was performed to study the relationship between seasonal dynamics of dengue and climatic drivers.

Results: Among the cities, the number of notified dengue cases in Chaozhou was greatest, accounting for 78.3%. The median age for the notified cases was 43 years (IQR: 27.0-58.0 years). Two main regions located in Xixin and Chengxi streets of Chaozhou with a high risk of infection were observed, indicating that there was substantial spatial heterogeneity in intensity. We found an annual peak incidence occurred in autumn across the region, most markedly in 2015. This study reveals that periods of elevated temperatures can drive the occurrence of dengue epidemics across the region, and the risk of transmission is highest when the temperature is between 25°C and 28°C.

Conclusion: Our study contributes to a better understanding of dengue dynamics in Chao-Shan area.

Keywords: China; climate; dengue; seasonal variation; spatial; wavelet analysis.

Figure 1

Geographical location of Chao-Shan area including Shantou, Chaozhou and Jieyang cities on China’s southeastern coast.

Figure 2

Proportions of gender-specific cases of dengue fever by age groups and estimates of onset-to-diagnosis distributions of dengue fever cases in Chao-Shan area, 2014–2017. (A) Based on total cases. (B) Based on cases aged less than 35 years. (C) Based on cases aged greater than or equal to 35 years. (D) Onset-to-diagnosis distribution by city.

Figure 3

Geographical distributions of dengue cases in Chao-Shan area from the beginning of 2014 to the end of 2017. Spatial smoothing with the kernel density estimation produced dengue risk hotspots suggested that the geographical regions with the greatest number of dengue cases were around Xixin and Chengxi streets in Chaozhou city.

Figure 4

Temporal dynamics of dengue in Chao-Shan area from the beginning of 2014 to the end of 2017. (A) Weekly time series of number of probable and confirmed dengue cases in Chao-Shan area. (B) Weekly time series of number of probable and confirmed dengue cases by city in Chao-Shan area.

Figure 5

Wavelet analyses for time series of notifications of dengue and climatic variables including mean temperature, rainfall and relative humidity in Chao-Shan area, 2014–2017. Local wavelet power spectrum for dengue cases (A), mean temperature (B), rainfall (C) and relative humidity (D). Solid and bold lines indicate boundary of statistical significance.

Figure 6

Analysis of potentially non-linear effects of mean temperature on seasonal dynamics of dengue in Chao-Shan area using a generalised additive model for location, scale and shape (GAMLSS) model based on data from years 2014–2017. Initially, climatic variables including mean temperature, rainfall and relative humidity were assessed using the GAMLSS model with different df for the smooth terms during model selection. Then, the optimal model with the statistically significant variable (mean temperature) was determined. The values of partial effect function from a smoothing effect term in a GAMLSS model were extracted and plot against their predictors. This plot shows the significant non-linear partial effect of mean temperature on dengue dynamics, and the risk of dengue transmission is highest when the temperature is between 25°C and 28°C. The light blue dots denote partial residuals of the model, which were added in the plot to see how the model fits. From the dots of partial residuals, we can there are no outliers in the data.