The effects of seasonal climate variability on dengue annual incidence in Hong Kong: A modelling study

Abstract

In recent years, dengue has been rapidly spreading and growing in the tropics and subtropics. Located in southern China, Hong Kong's subtropical monsoon climate may favour dengue vector populations and increase the chance of disease transmissions during the rainy summer season. An increase in local dengue incidence has been observed in Hong Kong ever since the first case in 2002, with an outbreak reaching historically high case numbers in 2018. However, the effects of seasonal climate variability on recent outbreaks are unknown. As the local cases were found to be spatially clustered, we developed a Poisson generalized linear mixed model using pre-summer monthly total rainfall and mean temperature to predict annual dengue incidence (the majority of local cases occur during or after the summer months), over the period 2002-2018 in three pre-defined areas of Hong Kong. Using leave-one-out cross-validation, 5 out of 6 observations of area-specific outbreaks during the major outbreak years 2002 and 2018 were able to be predicted. 42 out of a total of 51 observations (82.4%) were within the 95% confidence interval of the annual incidence predicted by our model. Our study found that the rainfall before and during the East Asian monsoon (pre-summer) rainy season is negatively correlated with the annual incidence in Hong Kong while the temperature is positively correlated. Hence, as mosquito control measures in Hong Kong are intensified mainly when heavy rainfalls occur during or close to summer, our study suggests that a lower-than-average intensity of pre-summer rainfall should also be taken into account as an indicator of increased dengue risk.

Figure 1

The area division: New Territories - South (NTS; light blue), New Territories - North (NTN; orange), and Hong Kong & Kowloon (HKL; green) along with the selected automatic weather stations (white circles) and rainfall stations (light yellow circles). The grey lines indicate the official district borders. As Ma Wan Island (red) is located at the center, all the 17 non-blood transfusion cases on that island are assigned into the 3 areas in our study: 5 cases in NTS, 6 cases in NTN, and 5 cases in HKL. The inset provides annual local case numbers from 2002 to 2018.

Figure 2

The monthly climate data for the whole of Hong Kong from 2002 to 2018. (a) Monthly mean and monthly minimum temperature, and (b) monthly total and monthly maximum rainfall from January to August for the years 2002, 2003–2017 combined, and 2018 are shown. Climate data were retrieved from 11 weather stations and 2 automatic rainfall stations across the three pre-defined areas.

Figure 3

The number of monthly reported local dengue cases from 2002 to 2018. Different colours represent the number of dengue cases in different years.

Figure 4

The normalised climate predictors for each area using (a) the monthly mean temperature and (b) the monthly total rainfall. The areas NTS, NTN, and HKL are defined as in Fig. 1.

Figure 5

Comparison between observed and predicted number of annual dengue cases. (a–c) Observed and predicted number of annual dengue cases in each pre-defined area (NTS, NTN, and HKL) with 95% confidence intervals using leave-one-out cross-validation. The predicted values represent the mean of the Poisson distribution of the annual dengue cases in each area, estimated using a generalized linear mixed model with a restricted maximum likelihood method. (d) Observed and predicted number of annual dengue cases in the whole Hong Kong area (ALL) with 95% confidence intervals using leave-one-year-out cross-validation.