Spatial measurement error and correction by spatial SIMEX in linear regression models when using predicted air...

Area of Science:

• Environmental Epidemiology
• Geostatistics
• Spatial Statistics

Background:

• Spatial modeling is crucial for air pollution epidemiology, aiming to enhance exposure assessment accuracy.
• Key uncertainties in air pollution exposure models arise from estimation errors and model misspecification.
• Existing methods often struggle to fully account for these sources of error.

Purpose of the Study:

• To investigate the impact of estimation error and model misspecification in spatial air pollution models on health effect estimates.
• To develop and evaluate a novel statistical procedure for correcting bias in air pollution epidemiology studies.
• To apply the proposed method to real-world data examining air pollution and birth outcomes.

Main Methods:

• Utilized a universal Kriging framework for predicting air pollution exposures, incorporating land-use regression terms and spatial covariance structures.
• Derived analytical expressions for bias in health effect estimates resulting from exposure model estimation error and misspecification.
• Developed and implemented a spatial simulation extrapolation (SIMEX) procedure to correct for asymptotic bias.

Main Results:

• Demonstrated that a misspecified air pollution exposure model can lead to asymptotic bias in estimated health effects.
• The proposed spatial SIMEX procedure effectively corrected for this induced asymptotic bias.
• The method's performance was validated through application to a study on air pollution and birthweight in Massachusetts.

Conclusions:

• Spatial SIMEX offers a robust approach to mitigate bias in air pollution epidemiology stemming from exposure model uncertainties.
• Accurate exposure assessment is vital for reliable health effect estimation in environmental health research.
• This methodology can improve the validity of findings in studies linking environmental exposures to health outcomes.