Estimating cumulative spatial risk over time with low-rank kriging multiple membership models

Area of Science:

• Environmental epidemiology
• Spatial statistics
• Biostatistics

Background:

• Health outcomes are often linked to cumulative environmental exposures.
• Traditional spatial risk studies are limited by considering only single residential locations.
• Accurate risk assessment requires accounting for residential mobility and multiple exposure points.

Purpose of the Study:

• To develop a novel Bayesian model for estimating cumulative spatial risk at a point level.
• To incorporate multiple residential locations per subject into spatial risk analysis.
• To improve the estimation of disease risk by accounting for environmental factor accumulation.

Main Methods:

• Proposed a Bayesian model embedding a multiple membership model (MMM) into a low-rank kriging (LRK) model.
• Utilized point-level data for enhanced precision compared to administrative areas.
• Compared the LRK-MMM model with existing multiple membership models using simulation studies.

Main Results:

• The LRK-MMM model demonstrated improved spatial sensitivity (0.12-0.54) and power (0.02-0.94) in detecting disease risk regions.
• The model effectively estimates cumulative spatial risk considering multiple residential histories.
• Simulations confirmed the model's efficacy and advantages over area-level models.

Conclusions:

• The proposed LRK-MMM Bayesian model offers a computationally efficient and precise method for spatial risk assessment.
• This approach enhances the ability to identify environmental risk factors for diseases.
• The model is applicable to case-control studies for estimating cumulative spatial risk with covariate adjustment.