Spatially Correlated Analysis of Infectious Disease Outcomes Based on Bayesian Functional Hierarchical Models.

Area of Science:

• Epidemiology
• Biostatistics
• Computational Biology

Background:

• Infectious diseases like COVID-19 cause significant public health challenges.
• Accurate forecasting of disease spread is crucial for effective interventions.
• Existing models struggle with temporal misalignment, spatial dependencies, and data overdispersion.

Purpose of the Study:

• To develop a novel Bayesian hierarchical model for analyzing and predicting infectious disease trajectories.
• To address limitations in current forecasting methods, specifically temporal misalignment, spatial dynamics, and overdispersion.
• To provide a tool for early prediction of case surges and inform public health resource allocation.

Main Methods:

• Developed a Bayesian hierarchical model for spatially correlated functional count data.
• Implemented curve preprocessing for temporal alignment of epidemic curves.
• Utilized Negative-Binomial distribution for overdispersion and nonparametric basis functions for temporal dynamics.
• Modeled spatial correlation using a Leroux conditional autoregressive prior.
• Employed a Gibbs sampler for posterior inference and forecasting.

Main Results:

• Simulation studies showed significant improvements in estimation accuracy and prediction precision over alternative models.
• The model successfully analyzed COVID-19 case data across U.S. states.
• Identified time-varying effects of key covariates on disease spread.
• Demonstrated capability for early prediction of case surges in regions with delayed outbreaks.

Conclusions:

• The proposed model offers a robust framework for analyzing and forecasting infectious disease dynamics.
• It effectively handles temporal misalignment, spatial dependencies, and overdispersion in count data.
• The model provides valuable insights and predictive capabilities for public health decision-making and strategic planning.