Spatial autocorrelation among automated geocoding errors and its effects on testing for disease clustering

Area of Science:

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

Background:

• Automated geocoding is crucial for spatial epidemiology, but introduces positional errors.
• These errors can reduce the effectiveness of spatial analysis and disease clustering tests.
• Previous research suggested geocoding errors might be spatially correlated, potentially mitigating negative effects.

Purpose of the Study:

• To explicitly demonstrate spatial autocorrelation in automated geocoding errors.
• To assess the impact of this spatial autocorrelation on disease cluster detection power.
• To discuss implications for geographic health data analysis and privacy.

Main Methods:

• Analyzed positional errors from geocoding over 6000 addresses in Carroll County, Iowa.
• Conducted two simulation studies on disease processes, one using the Carroll County data.
• Evaluated the power of two disease clustering tests under correlated versus independent error scenarios.

Main Results:

• Positional errors from automated geocoding in the study dataset were confirmed to be spatially autocorrelated.
• Spatial autocorrelation of geocoding errors maintained higher power for disease cluster detection tests compared to independent errors.
• Simulation results indicated that correlated errors offer a protective effect on statistical power.

Conclusions:

• Spatial autocorrelation in geocoding errors is a significant factor in spatial epidemiologic studies.
• This correlation can preserve the power of disease clustering detection, contrary to expectations for random errors.
• Findings have implications for measurement error modeling, privacy, and enhancing spatial analytic methods in health research.