Machine Learning-Aided Causal Inference Framework for Environmental Data Analysis: A COVID-19 Case Study.

Area of Science:

• Environmental Science
• Epidemiology
• Data Science

Background:

• Studies suggest links between environmental conditions and COVID-19 severity.
• Existing data analysis methods struggle with complex, multidimensional environmental-COVID-19 relationships.

Purpose of the Study:

• To investigate potential causal relationships between COVID-19 severity and 10 environmental factors.
• To apply a novel causal inference framework using machine learning.

Main Methods:

• Utilized a structural causal model with machine learning for causal inference.
• Analyzed time-series data from 166 Chinese cities across different pandemic phases and socio-economic clusters.
• Examined 10 environmental factors: NO2, O3, PM2.5, PM10, SO2, CO, temperature, pressure, humidity, and wind speed.

Main Results:

• Robustness checks invalidated 89 out of 90 potential causal relationships.
• Only average air temperature showed a minor causal effect (0.041) under specific cluster-phase conditions.
• Environmental factors were found unlikely to significantly aggravate COVID-19 severity.

Conclusions:

• Environmental factors generally do not cause noticeable increases in COVID-19 severity.
• The proposed causal inference method is valuable for analyzing observational data in environmental and other fields.
• Highlights the need for advanced analytical techniques in environmental epidemiology.