A Method to Estimate Health Effects Based on Error-Prone Simulated Environmental Exposure: An Application to a...

Area of Science:

• Environmental epidemiology
• Climate science
• Biostatistics

Background:

• Earth System Models (ESMs) offer continuous environmental data but are limited by measurement error uncertainty in epidemiological studies.
• Existing methods struggle to address spatiotemporal error covariance inherent in ESMs.

Purpose of the Study:

• To introduce a novel latent-variable approach for correcting measurement errors in ESM-derived exposure data.
• To quantify the impact of fine particulate matter (PM2.5) on birthweight, accounting for model uncertainties.

Main Methods:

• Derived spatiotemporal error covariance from comparing Coupled Model Intercomparison Project Phase 6 (CMIP6) PM2.5 simulations with global station data (5,661 sites).
• Applied a latent-variable model to correct CMIP6 PM2.5 exposures for measurement error.
• Associated corrected PM2.5 exposures with birthweight data from 132 Demographic and Health Surveys.

Main Results:

• Initial correlations between CMIP6 models and observations ranged from r=0.40-0.68.
• Uncorrected effect estimates for PM2.5 on birthweight varied widely across ESMs.
• The measurement error-corrected estimate showed a consistent reduction of 3.34 g in birthweight per 10 μg/m³ increase in PM2.5.

Conclusions:

• The negative association between PM2.5 exposure and birthweight is robust, even with measurement error in ESM data.
• Correcting for measurement error in environmental epidemiology reduces bias and improves the consistency of effect estimates.
• This framework enhances the utility of ESMs for reliable environmental health research.