Application of machine learning to predict hospital visits for respiratory diseases using meteorological and air pollution factors in Linyi, China

• DOI: 10.1007/s11356-023-28682-8
• Publication Date: 2023-07-12T20:31:11Z
• Authors: Jing Yang; Xin Xu; Xiaotian Ma; Zhaotong Wang; Qian You; Wanyue Shan; Ying Yang; Xin Bo; Chuansheng Yin
• Keywords: Air Pollutants; China; Meteorological Concepts; Pneumonia; Respiration Disorders; Hospitals; 3. Good health; Machine Learning; 13. Climate action; Air Pollution; 11. Sustainability; Humans; Particulate Matter; Respiratory Tract Infections

Abstract Urbanization and industrial development have resulted in increased air pollution, which is concerning for public health. This study evaluated the effect of meteorological factors and air pollution on hospital visits for respiratory diseases (pneumonia, acute upper respiratory infections, and chronic lower respiratory diseases). The test dataset comprised meteorological parameters, air pollutant concentrations, and outpatient hospital visits for respiratory diseases in Linyi, China from January 1, 2016 to August 20, 2022. We used support vector regression (SVR) to build regression models to analyze the effect of meteorological factors and air pollutants on the number of outpatient visits for respiratory diseases. To evaluate the model performance, 70% of the dataset was used for training and 30% was used for testing. The Spearman correlation and SVR model results indicated that NO2, PM2.5, and PM10 were correlated with the occurrence of respiratory diseases, and the strongest correlation was for pneumonia. An increase in the daily average temperature and daily relative humidity decreased the number of patients with pneumonia and chronic lower respiratory diseases but increased the number of patients with acute upper respiratory infections. The SVR modeling showed potential for predicting the number of respiratory-related hospital visits. This work demonstrated that combining machine learning with meteorological and air pollution data can be used for disease prediction and can serve as a useful tool for policymakers to take preventive measures.