Prediction of adult diarrhea disease in Shanghai using meteorological factors and a web search index
- VernacularTitle:基于气象因素和某互联网搜索指数的上海市成人腹泻疾病预测方法研究
- Author:
Sixu YANG
1
;
Li PENG
1
;
Huanyu WU
2
;
Jian CHEN
3
;
Xiaofang YE
1
;
Xuefei ZHANG
1
;
Dandan YANG
1
;
Xiaohuan GONG
4
;
Sheng LIN
3
Author Information
- Publication Type:Investigation
- Keywords: diarrhea; meteorological factors; web search index; distributed lag non-linear model; prediction
- From: Journal of Environmental and Occupational Medicine 2026;43(8):951-958
- CountryChina
- Language:Chinese
- Abstract: Background Diarrhea disease is a common intestinal infectious disease, and its incidence is affected by meteorological conditions. A better understanding of its epidemiological patterns and influencing factors, together with the construction of reliable prediction models, is of great significance for precise public health prevention and control. Objective To clarify the epidemic characteristics of adult diarrhea disease in Shanghai, analyze the associations of meteorological factors and a web search index with adult diarrhea disease, and develop and compare forecasting models to support precise regional prevention and control. Methods Weekly surveillance data of adult diarrhea disease cases from the Shanghai Comprehensive Surveillance Information System for Diarrhea Diseases, together with concurrent meteorological observation data and web search index (Baidu index) data from 2014 to 2019, were collected. A distributed lag non-linear model (DLNM) was adopted to analyze the associations of multiple meteorological factors and the web search index with the number of diarrhea disease cases. By integrating meteorological factors and web search index data, three types of forecasting models were developed, including autoregressive integrated moving average (ARIMA), Random Forest, and extreme gradient boosting (Xgboost), and their predictive performances were evaluated. Results Adult diarrhea disease in Shanghai exhibited seasonal variation, with an major incidence peak in summer and winter peaks in some years. The number of cases declined annually after 2015. Mean temperature was significantly associated with the risk of diarrhea disease, and both low and high temperature exposures were associated with increased risks. The highest risk was observed at 32.8°C (RR=2.04, 95%CI: 1.62, 2.55), while the strongest effect of low temperature was observed at 0.9 °C (RR=1.53, 95%CI: 1.25, 1.88). When relative humidity exceeded 69%, the risk of diarrhea disease increased with relative humidity, reaching a peak at 81% (RR=1.20, 95%CI: 1.07, 1.35). When weekly cumulative precipitation exceeded 16 mm, the risk also increased with increasing precipitation, reaching a maximum at 105 mm (RR=1.23, 95%CI: 1.07, 1.42). The web search index was positively associated with the risk of diarrhea disease. Model prediction indicated that both the Random Forest model and the Xgboost model adequately captured the overall trend in diarrhea disease cases, with R2 values generally exceeding 0.7. Notably, the Xgboost model demonstrated greater accuracy in capturing peak intensities. Conclusion Meteorological factors are associated with adult diarrhea disease in Shanghai. The web search index may serve as an auxiliary indicator for diarrhea forecasting. Machine learning models, with advantages in integrating multisource data, may provide effective predictive tools for the prevention and control of diarrhea disease.
