An introduction to bayesian spatial smoothing methods for disease mapping: modeling county firearm suicide mortality...

Area of Science:

• Biostatistics
• Spatial Epidemiology
• Public Health

Background:

• Disease mapping often faces challenges with small area estimation due to limited data.
• Firearm suicide data presents high variability in raw rates, especially in counties with few deaths.
• Existing methods struggle to provide stable mortality estimates for small geographic areas.

Purpose of the Study:

• Introduce Bayesian spatial smoothing models for disease mapping to public health professionals.
• Apply Besag, York, and Mollié (BYM) models to firearm suicide data (2014-2018).
• Demonstrate the utility of these models for evaluating spatial health disparities.

Main Methods:

• Utilized Bayesian spatial and space-time smoothing models (BYM Poisson).
• Fitted models to county-level firearm suicide counts from 2014-2018.
• Employed new estimation techniques in R software for accessibility.

Main Results:

• Raw county firearm suicide death rates varied widely (0-24.81 per 10,000), with many counties suppressed due to low counts.
• Spatially smoothed estimates ranged from 0.06 to 4.05 per 10,000, providing estimates for all counties.
• Space-time models yielded similar estimates with narrower credible intervals, enhancing precision.

Conclusions:

• Bayesian spatial smoothing effectively addresses highly variable rate estimates in small geographies.
• These methods are valuable for assessing spatial health disparities, particularly for rare events.
• Improved accessibility of these models in R facilitates their use by researchers and public health professionals.