Quantifying spatial heterogeneity of vulnerability to short-term PM_2.5 exposure with data fusion framework

Area of Science:

• Environmental Health
• Epidemiology
• Data Science

Background:

• Estimates of disease burden from PM2.5 exposure are often biased by varying vulnerabilities across urbanization levels and reliance on non-local risk data.
• Accurate assessment requires localized risk evaluation and consideration of spatial heterogeneity in exposure and health impacts.

Purpose of the Study:

• To develop a data fusion framework for estimating PM2.5 exposure disease burden using local data.
• To evaluate local PM2.5 exposure risks and quantify their spatial heterogeneity, relationship to land-use, and associated uncertainties.
• To apply the framework to Tainan City, Taiwan, for a case study.

Main Methods:

• Utilized six local databases for PM2.5 exposure risk and disease burden (death, cardiovascular disease (CVD), respiratory disease (RD)) data.
• Applied a data fusion framework integrating emergency department visit data (2006-2016), air quality monitoring, and land-use characteristics.
• Conducted sensitivity analysis to assess uncertainties in disease burden estimations based on data sources and risk values.

Main Results:

• Identified higher risks of CVD and RD in highly urbanized areas and death in rural areas, ranging from 1.20 to 1.57 times the average.
• Quantified uncertainties in disease burden estimations, with PM2.5 exposure data contributing 20-32% and risk values 0-86%, particularly in urbanized regions.
• Demonstrated significant spatial heterogeneity in PM2.5-related health risks.

Conclusions:

• The developed data fusion framework effectively estimates localized PM2.5 exposure disease burden and associated uncertainties.
• The approach offers a scalable solution for developing and overpopulated countries to support local air quality and health management.
• Findings highlight the need for localized data in assessing environmental health impacts.