Retrospectively understanding the multifaceted interplay of COVID-19 outbreak, air pollution, and sociodemographic factors through explainable AI

Hygiene and environmental health advances Pub Date : 2025-01-23 DOI:10.1016/j.heha.2025.100119

Mohmmed Talib , Kripabandhu Ghosh , Gopala Krishna Darbha

{"title":"Retrospectively understanding the multifaceted interplay of COVID-19 outbreak, air pollution, and sociodemographic factors through explainable AI","authors":"Mohmmed Talib , Kripabandhu Ghosh , Gopala Krishna Darbha","doi":"10.1016/j.heha.2025.100119","DOIUrl":null,"url":null,"abstract":"<div><div>This study aims to holistically comprehend the intricate dynamics between air pollution, socio-demographics, and COVID-19 outcomes in India. The primary objective centers on deploying explainable AI (XAI) methodologies to elucidate the intricate pathways and latent mechanisms governing these associations.</div><div>A multi-faceted approach was employed integrating ecological study, hybrid-ML, and XAI techniques to characterize the underlying dependencies and interactions driving the pandemic's spatiotemporal evolution and system dynamics. The ecological study analyzed the association between air pollution levels and COVID-19 case fatality rates (CFRs) across distinct pandemic phases. We utilized a Negative Binomial model for interpretability and implemented a stacked ensemble framework to enhance predictive performance. This stacked model was further leveraged to provide deeper insights into the underlying patterns through XAI techniques.</div><div>The ecological study identified distinct patterns in CFR across different pandemic phases of the pandemic, with higher pollution levels monotonically associated with increased CFRs. Furthermore, the stacked ensemble model consistently outperformed its base models, demonstrating the benefits of combining multiple models. Additionally, the XAI analysis identified NO<sub>2</sub> as a key driver of COVID-19 cases and mortalities, while PM<sub>10</sub> was found to be particularly influential on mortalities. The study concluded distinct COVID-19 transmission patterns across regions and pandemic phases, highlighting the influence of non-pharmaceutical interventions, viral strains, and socio-demographics in driving the pandemic.</div><div>The findings highlight the need for strong pollution controls to mitigate air pollution's impact on health. The developed hybrid model can aid in predicting COVID-19 outcomes in future respiratory outbreaks, supporting public health planning and targeted interventions.</div></div>","PeriodicalId":73269,"journal":{"name":"Hygiene and environmental health advances","volume":"13 ","pages":"Article 100119"},"PeriodicalIF":0.0000,"publicationDate":"2025-01-23","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":"0","resultStr":null,"platform":"Semanticscholar","paperid":null,"PeriodicalName":"Hygiene and environmental health advances","FirstCategoryId":"1085","ListUrlMain":"https://www.sciencedirect.com/science/article/pii/S2773049225000029","RegionNum":0,"RegionCategory":null,"ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":null,"EPubDate":"","PubModel":"","JCR":"","JCRName":"","Score":null,"Total":0}

引用次数: 0

Abstract

This study aims to holistically comprehend the intricate dynamics between air pollution, socio-demographics, and COVID-19 outcomes in India. The primary objective centers on deploying explainable AI (XAI) methodologies to elucidate the intricate pathways and latent mechanisms governing these associations.

A multi-faceted approach was employed integrating ecological study, hybrid-ML, and XAI techniques to characterize the underlying dependencies and interactions driving the pandemic's spatiotemporal evolution and system dynamics. The ecological study analyzed the association between air pollution levels and COVID-19 case fatality rates (CFRs) across distinct pandemic phases. We utilized a Negative Binomial model for interpretability and implemented a stacked ensemble framework to enhance predictive performance. This stacked model was further leveraged to provide deeper insights into the underlying patterns through XAI techniques.

The ecological study identified distinct patterns in CFR across different pandemic phases of the pandemic, with higher pollution levels monotonically associated with increased CFRs. Furthermore, the stacked ensemble model consistently outperformed its base models, demonstrating the benefits of combining multiple models. Additionally, the XAI analysis identified NO₂ as a key driver of COVID-19 cases and mortalities, while PM₁₀ was found to be particularly influential on mortalities. The study concluded distinct COVID-19 transmission patterns across regions and pandemic phases, highlighting the influence of non-pharmaceutical interventions, viral strains, and socio-demographics in driving the pandemic.

The findings highlight the need for strong pollution controls to mitigate air pollution's impact on health. The developed hybrid model can aid in predicting COVID-19 outcomes in future respiratory outbreaks, supporting public health planning and targeted interventions.

Abstract Image