Real-time forecasting and early warning of bacillary dysentery activity in four meteorological and geographic divisions in China

• Published in: The Science of the total environment 2021-03, Vol.761, p.144093, Article 144093
• Main Authors: Wang, Shuzi, Liu, Zhidong, Tong, Michael, Xiang, Jianjun, Zhang, Ying, Gao, Qi, Zhang, Yiwen, Lu, Liang, Jiang, Baofa, Bi, Peng
• Format: Article
• Language: English
• Subjects: Bacillary dysentery; Beijing; Boosted regression tree; China - epidemiology; Cities; Dysentery, Bacillary - epidemiology; Forecast; Forecasting; Humans; Incidence; Meteorological Concepts; Meteorological factors; Outbreak alert
• ISSN: 0048-9697; 1879-1026
• DOI: 10.1016/j.scitotenv.2020.144093

Accurate and timely forecasts of bacillary dysentery (BD) incidence can be used to inform public health decision-making and response preparedness. However, our ability to detect BD dynamics and outbreaks remains limited in China. This study aims to explore the impacts of meteorological factors on BD transmission in four representative regions in China and to forecast weekly number of BD cases and outbreaks. Weekly BD and meteorological data from 2014 to 2016 were collected for Beijing (Northern China), Shenyang (Northeast China), Chongqing (Southwest China) and Shenzhen (Southern China). A boosted regression tree (BRT) model was conducted to assess the impacts of meteorological factors on BD transmission. Then a real-time forecast and early warning model based on BRT was developed to track the dynamics of BD and detect the outbreaks. The forecasting methodology was compared with generalized additive model (GAM) and seasonal autoregressive integrated moving average model (SARIMA) that have been used to model the BD case data previously. Ambient temperature was the most important meteorological factor contributing to the transmission of BD (80.81%–92.60%). A positive effect of temperature was observed when weekly mean temperature exceeded 4 °C, −3 °C, 9 °C and 16 °C in Beijing (Northern China), Shenyang (Northeast China), Chongqing (Southwest China) and Shenzhen (Southern China), respectively. BD incidence (Beijing and Shenyang) in temperate cities was more sensitive to high temperature than that in subtropical cities (Chongqing and Shenzhen). The dynamics and outbreaks of BD can be accurately forecasted and detected by the BRT model. Compared to GAM and SARIMA, BRT model showed more accurate forecasting for 1-, 2-, 3-weeks ahead forecasts in Beijing, Shenyang and Shenzhen. Temperature plays the most important role in weather-attributable BD transmission. The BRT model achieved a better performance in comparison with GAM and SARIMA in most study cities, which could be used as a more accurate tool for forecasting and outbreak alert of BD in China. [Display omitted] •Temperature had the largest contribution for weather-attributable BD transmission.•Temperate city was more sensitive to high temperature compared to subtropical city.•The dynamics of BD can be accurately forecasted by the BRT model.•The BRT model had the better performance in comparison with GAM and SARIMA.