Bayesian Spatial Functional Data Clustering: Applications in Disease Surveillance.

Area of Science:

• Spatial statistics
• Epidemiology
• Computational statistics

Background:

• Accurate disease risk mapping is vital for public health.
• Existing spatial clustering methods have limitations in flexibility and computational efficiency.
• Understanding disease evolution in contiguous regions is key for targeted interventions.

Purpose of the Study:

• To develop a novel spatial functional clustering model for disease risk mapping.
• To identify spatially contiguous clusters with similar latent disease risk functions.
• To enhance Bayesian inference for complex spatial models.

Main Methods:

• Utilizing random spanning trees for spatial partitioning.
• Employing latent Gaussian models for within-cluster structure.
• Extending random spanning trees to exponential family response variables.
• Developing a Bayesian inference algorithm using composition sampling and integrated nested Laplace approximation (INLA).

Main Results:

• The model successfully identifies spatially contiguous clusters with similar latent functions (trends, seasonality, etc.).
• The proposed Bayesian inference method improves computational feasibility and model mixing.
• The approach is effective for non-Gaussian likelihoods and cluster-specific parameters.
• Demonstrated effectiveness in simulation studies and real-world disease mapping (COVID-19, dengue fever).

Conclusions:

• The novel spatial functional clustering model provides a flexible and computationally feasible approach for disease risk mapping.
• The method uncovers meaningful spatial and temporal disease outbreak patterns.
• Findings offer valuable insights for public health decision-making and resource allocation during epidemics.