An application of density estimation to geographical epidemiology

Area of Science:

• Environmental Epidemiology
• Spatial Statistics
• Public Health

Background:

• Assessing geographical variations in disease risk is crucial for public health.
• Traditional methods may struggle with sparse data in localized areas.
• Understanding disease clusters near industrial sites requires robust spatial analysis.

Purpose of the Study:

• To define and estimate a relative risk function across a geographical region.
• To apply and demonstrate the method using childhood leukaemia data near Sellafield.
• To explore modifications for improved spatial risk estimation.

Main Methods:

• Defined a relative risk function for spatial disease distribution.
• Employed kernel density estimation for disease cases and control samples.
• Utilized adaptive kernels and demonstrated the procedure with real-world data.

Main Results:

• Kernel density estimation effectively estimated the relative risk function.
• The method revealed a sharp peak in risk at Sellafield.
• A reasonably smooth risk surface was generated across the region despite limited case numbers.

Conclusions:

• Kernel density estimation is an effective tool for mapping disease risk.
• The study successfully identified a localized elevated risk of childhood leukaemia near Sellafield.
• The proposed methods are valuable for environmental health investigations with sparse data.