On prior smoothing with discrete spatial data in the context of disease mapping

Area of Science:

• Spatial statistics
• Biostatistics
• Geographic Information Systems (GIS)

Background:

• Disease mapping uses Markov random field models to analyze health event counts in areal units.
• Spatial priors are crucial for adjusting raw risk estimates but the extent of smoothing they introduce is not well-understood.
• Existing literature lacks comparative analysis of different spatial priors' smoothing effects and parameter influence.

Purpose of the Study:

• To compare the smoothing effects of seven commonly used spatial priors in disease mapping.
• To investigate how varying prior parameters influences the degree of smoothing.
• To develop quantitative metrics for assessing and calibrating spatial smoothing.

Main Methods:

• Simulations and real-world data analyses using areal maps of peninsular Spain and England.
• Application of two datasets with associated populations at risk.
• Development of empirical and theoretical metrics to quantify and calibrate smoothing effects.

Main Results:

• Quantitative characterization of smoothing extent within and across different spatial models.
• Establishment of a link between theoretical smoothing metrics and empirical observations.
• Demonstration of how varying prior parameters impacts the resulting spatial smoothing.

Conclusions:

• This research provides a framework for understanding and quantifying spatial smoothing in disease mapping.
• The developed metrics enable objective comparison and calibration of different spatial priors.
• Findings can inform the selection of appropriate spatial priors for more accurate health event analysis.