Latent conditional individual-level models for infectious disease modeling.

Area of Science:

• Epidemiology
• Mathematical Modeling
• Computational Biology

Background:

• Individual-level models (ILMs) are used for infectious disease spread, incorporating individual covariates for heterogeneity.
• Infectious disease modeling often faces challenges with incomplete or unreliable covariate data, impacting model robustness.

Purpose of the Study:

• To assess an adaptation of a spatial ILM incorporating a latent grouping structure.
• To evaluate the model's performance when precise covariate information is uncertain or unavailable.
• To compare the latent conditional ILM with a homogeneous ILM and one using explicit covariates.

Main Methods:

• Developed a latent conditional individual-level model (ILM) dependent on a discrete latent grouping variable.
• Tested the model's posterior predictive ability using simulation studies with epidemic data.
• Compared the proposed ILM against a homogeneous ILM and an ILM using explicit covariate information.
• Applied the models to real-world data from the 2001 UK foot-and-mouth disease epidemic.

Main Results:

• The latent conditional ILM demonstrated effective posterior predictive ability.
• The proposed model showed that discrete latent grouping is a viable alternative to precise covariate information.
• The model's performance was comparable to or better than the homogeneous ILM in certain scenarios.

Conclusions:

• Discrete latent grouping variables offer a robust alternative to explicit covariate information in spatial ILMs for disease spread.
• This approach enhances the reliability of infectious disease models when dealing with data uncertainties.
• The findings have implications for modeling epidemics, particularly in resource-limited settings or when data quality is a concern.