Linearized forms of individual-level models for large-scale spatial infectious disease systems

Area of Science:

• Epidemiology
• Computational Biology
• Statistical Modeling

Background:

• Individual-level models (ILMs) offer flexibility in infectious disease modeling by incorporating population heterogeneity through covariates.
• Spatial spread of diseases is often modeled using ILMs with geometric distance kernels, but these are computationally demanding for large populations.
• Bayesian Markov Chain Monte Carlo (MCMC) frameworks are commonly used for fitting these complex models.

Purpose of the Study:

• To investigate methods for speeding up likelihood calculations in spatial infectious disease models.
• To examine techniques for linearizing distance kernels within ILMs.
• To compare the performance of different linearization methods.

Main Methods:

• Development and application of linearization techniques for geometric distance kernels in ILMs.
• Bayesian MCMC framework utilized for model fitting and comparison.
• Comparative analysis of computational performance across different linearization strategies.

Main Results:

• Linearization of distance kernels significantly reduces computational time for ILMs.
• Different linearization methods exhibit varying degrees of performance and accuracy.
• The study identifies effective strategies for accelerating spatial disease spread simulations.

Conclusions:

• Linearization offers a viable approach to overcome computational bottlenecks in spatial ILMs.
• Careful selection of linearization techniques is crucial for balancing speed and accuracy.
• These optimized models can facilitate more efficient analysis of infectious disease dynamics in large populations.