A geostatistical approach to large-scale disease mapping with temporal misalignment.

Area of Science:

• Spatial statistics
• Geostatistics
• Disease mapping

Background:

• Temporal boundary misalignment complicates analysis of spatial trends using area-level data.
• Existing methods for temporally misaligned data often fail to account for spatial random effects over time.
• Large area-level datasets with shifting boundaries are increasingly prevalent in research.

Purpose of the Study:

• To develop a robust geostatistical model for aggregate count data that overcomes temporal boundary misalignment.
• To provide a flexible framework for analyzing spatial trends in disease incidence across changing geographic areas.
• To facilitate efficient model fitting for large-scale spatial epidemiology studies.

Main Methods:

• A geostatistical model is constructed assuming an underlying continuous risk surface that induces spatial correlation.
• The model is implemented within a generalized linear mixed model framework using radial basis splines.
• Penalized quasilikelihood approximation to maximum likelihood estimation is used for efficient model fitting.

Main Results:

• The proposed geostatistical approach effectively handles temporal boundary misalignment, rendering it a nonissue.
• The disease-mapping framework allows for fast and easy model fitting.
• The method is applicable to large disease-mapping datasets where fully Bayesian approaches may be computationally infeasible.

Conclusions:

• The developed geostatistical model provides a powerful solution for analyzing spatial trends in the presence of temporal boundary misalignment.
• This approach simplifies disease mapping and trend analysis for large area-level datasets.
• The method was successfully applied to assess socioeconomic trends in breast cancer incidence in Los Angeles.