Developing a Time Series Predictive Model for Dengue in Zhongshan, China Based on Weather and Guangzhou Dengue Surveillance Data

Abstract

Background: Dengue is a re-emerging infectious disease of humans, rapidly growing from endemic areas to dengue-free regions due to favorable conditions. In recent decades, Guangzhou has again suffered from several big outbreaks of dengue; as have its neighboring cities. This study aims to examine the impact of dengue epidemics in Guangzhou, China, and to develop a predictive model for Zhongshan based on local weather conditions and Guangzhou dengue surveillance information.

Methods: We obtained weekly dengue case data from 1st January, 2005 to 31st December, 2014 for Guangzhou and Zhongshan city from the Chinese National Disease Surveillance Reporting System. Meteorological data was collected from the Zhongshan Weather Bureau and demographic data was collected from the Zhongshan Statistical Bureau. A negative binomial regression model with a log link function was used to analyze the relationship between weekly dengue cases in Guangzhou and Zhongshan, controlling for meteorological factors. Cross-correlation functions were applied to identify the time lags of the effect of each weather factor on weekly dengue cases. Models were validated using receiver operating characteristic (ROC) curves and k-fold cross-validation.

Results: Our results showed that weekly dengue cases in Zhongshan were significantly associated with dengue cases in Guangzhou after the treatment of a 5 weeks prior moving average (Relative Risk (RR) = 2.016, 95% Confidence Interval (CI): 1.845-2.203), controlling for weather factors including minimum temperature, relative humidity, and rainfall. ROC curve analysis indicated our forecasting model performed well at different prediction thresholds, with 0.969 area under the receiver operating characteristic curve (AUC) for a threshold of 3 cases per week, 0.957 AUC for a threshold of 2 cases per week, and 0.938 AUC for a threshold of 1 case per week. Models established during k-fold cross-validation also had considerable AUC (average 0.938-0.967). The sensitivity and specificity obtained from k-fold cross-validation was 78.83% and 92.48% respectively, with a forecasting threshold of 3 cases per week; 91.17% and 91.39%, with a threshold of 2 cases; and 85.16% and 87.25% with a threshold of 1 case. The out-of-sample prediction for the epidemics in 2014 also showed satisfactory performance.

Conclusion: Our study findings suggest that the occurrence of dengue outbreaks in Guangzhou could impact dengue outbreaks in Zhongshan under suitable weather conditions. Future studies should focus on developing integrated early warning systems for dengue transmission including local weather and human movement.

Fig 1. Locations of Guangzhou, Zhongshan and the neighboring cities.

Nowadays, Guangzhou is considered a prominent commercial and business center in the PRD region. To maintain its role, major infrastructural projects were undertaken, including the construction of ring roads, highways, and railway tracks. This in turn provided easily accessible transportation between Guangzhou and other cities for the public. As Zhongshan is 86 kilometers from Guangzhou (a distance interconnected by Guangzhou-Zhuhai local train and several highways), workers and businessmen frequently travel between these two cities, and some go to work in Guangzhou during the day and return home to Zhongshan after work.

Fig 2. Weekly dengue case numbers in Zhongshan and Guangzhou, 2005 to 2014.

* *: The Y axis is displayed in natural exponential scale.

Fig 3. Cross correlations for natural logarithm transformed weekly dengue case counts of Zhongshan and Guangzhou.

After making a prior moving average at a span of 5 weeks for each meteorological factor, cross-correlations were also used to find the best lags at which meteorological factors led to Zhongshan epidemics. Fig 4 shows that each treated variable was positively correlated with natural logarithm transformed weekly dengue cases in Zhongshan at different lags (Details in S2 Table). Therefore, we selected the nearest peak as the best lag for each variable. Treated weekly average minimum temperature, relative humidity and rainfall have triggered epidemics in Zhongshan by 6 weeks (correlation coefficient = 0.304), 15 weeks (correlation coefficient = 0.274) and 7 weeks (correlation coefficient = 0.299), respectively.

Fig 4. Cross-correlation plots for natural logarithm transformed weekly dengue case counts and different weekly meteorological factors in Zhongshan.

A. Cross-correlation between natural logarithm transformed weekly dengue case counts and treated weekly average minimum temperature in Zhongshan, B. Cross-correlation between natural logarithm transformed weekly dengue case counts and treated weekly average relative humidity, C. Cross-correlation between natural logarithm transformed weekly dengue case counts and treated weekly accumulated rainfall.

Fig 5. ROC curves for forecasting dengue epidemics in Zhongshan at a threshold of (A) 3 cases in a week, (B) 2 cases in a week and (C) 1 case in a week.

To validate the robustness of our model and avoid overfitting, a k-fold cross-validation with a k value of 10 was applied. ROC curves were plotted again for each model in this section (S1 File), with average 0.938, 0.956 and 0.967 AUC at a forecasting threshold of 1, 2 and 3 cases per week, respectively. In the out-of-sample prediction (detailed results in S1 File), the sensitivity of our model remained as high as 78.83% and specificity 92.48% when we intend to forecast outbreaks involving more than 3 cases in a week in Zhongshan. Similarly, when we intend to forecast outbreaks with a smaller threshold, i.e., 1 case per week, our model had a sensitivity of 85.16% and specificity of 87.25% and at 2 cases per week, the sensitivity was 91.17% and specificity 84.00%. Such results meant that our model was rather robust and accurate.

Fig 6. Out-of-sample prediction for dengue epidemics in Zhongshan (2013/12/28–2014/12/26).