Disease risk near point sources: statistical issues for analyses using individual or spatially aggregated data

Area of Science:

• Environmental epidemiology
• Spatial statistics
• Biostatistics

Background:

• Assessing disease risk near environmental pollution requires handling data at multiple scales.
• Socioeconomic factors can confound the relationship between pollution and health outcomes.
• Existing statistical methods face challenges with mixed-level data.

Purpose of the Study:

• To investigate statistical challenges in analyzing disease risk near point pollution sources.
• To explore the impact of socioeconomic confounding in these analyses.
• To evaluate methods for handling data at both individual and areal levels.

Main Methods:

• A statistical review of existing methodologies.
• Examination of data aggregation issues.
• Consideration of ecological fallacy and confounding.

Main Results:

• Data aggregation to the areal level generally introduces bias in disease risk estimation.
• This bias is significant except in rare, specific scenarios.
• Mixed-level data (individual and areal) present substantial analytical hurdles.

Conclusions:

• Analyzing spatial disease risk with mixed data is challenging, with no simple solutions.
• Explicitly acknowledging statistical assumptions is crucial for interpreting results.
• Keeping data disaggregated is recommended to maximize information and minimize bias.