Realistic power simulations compare point- and area-based disease cluster tests

Area of Science:

• Epidemiology
• Spatial Statistics
• Biostatistics

Background:

• Disease cluster detection is crucial for public health.
• Existing methods include area-based and point-based approaches.
• The relative performance of these methods requires further investigation.

Purpose of the Study:

• To compare the statistical power of two area-based disease cluster tests (Moran's I and I* (pop)) against two point-based tests (Cuzick-Edwards and Grimson's).
• To evaluate these methods using realistic disease simulation data.

Main Methods:

• Area-based methods: Moran's I, I* (pop).
• Point-based methods: Cuzick-Edwards test, Grimson's test.
• Simulations based on fox rabies, childhood leukaemia, and Lyme disease data.

Main Results:

• The area-based method I* (pop) and the point-based Cuzick-Edwards test showed higher statistical power.
• This finding challenges the assumption that point-based methods are inherently superior for complex spatial disease data.
• I* (pop) effectively utilizes inter-region variability, a feature lacking in Moran's I.

Conclusions:

• Area-based methods, particularly I* (pop), can be highly effective for disease cluster detection.
• The choice of method should consider the specific characteristics of the spatial data.
• Further research into novel area-based statistical approaches is warranted.