A Bayesian space-time model for clustering areal units based on their disease trends

Area of Science:

• Epidemiology
• Biostatistics
• Spatial Analysis

Background:

• Disease risk varies spatially and temporally across populations.
• Existing methods focus on spatial clustering of disease risk, not temporal trends.
• Identifying areas with similar temporal disease trends is crucial for public health interventions.

Purpose of the Study:

• To develop a novel Bayesian hierarchical mixture model for identifying areal units with similar temporal disease trends.
• To apply the model to analyze measles susceptibility and respiratory hospitalizations in the UK.
• To compare the Metropolis-coupled Markov chain Monte Carlo ((MC)$^3$) algorithm with standard Markov chain Monte Carlo (MCMC).

Main Methods:

• A Bayesian hierarchical mixture model was developed.
• Inference was performed using a Metropolis-coupled Markov chain Monte Carlo ((MC)$^3$) algorithm.
• The methodology was validated through a simulation study and applied to real-world case studies.

Main Results:

• The (MC)$^3$ algorithm demonstrated effectiveness compared to standard MCMC.
• The model successfully identified distinct temporal disease trends in different geographic areas.
• Case studies revealed insights into measles susceptibility post-MMR vaccination concerns and temporal changes in respiratory hospitalizations in Glasgow.

Conclusions:

• The novel Bayesian model effectively clusters areal units based on similar temporal disease trends.
• The methodology provides a valuable tool for understanding disease dynamics and informing public health strategies.
• The study highlights the importance of considering temporal patterns in disease risk analysis.