Multi-Model Ensembles in Infectious Disease and Public Health: Methods, Interpretation, and Implementation in R

Area of Science:

• Computational epidemiology
• Statistical modeling
• Public health informatics

Background:

• Multi-model ensembles are widely used in forecasting for performance benefits.
• Their application is growing in public health for infectious disease outbreak prediction.
• Challenges include interpreting diverse methods and lack of standardized software.

Purpose of the Study:

• To introduce the statistical foundations of probabilistic forecasting and multi-model ensembles.
• To present the hubEnsembles package as a flexible software framework.
• To provide a tutorial and case study for practical ensemble generation.

Main Methods:

• Introduction to statistical foundations of probabilistic forecasting.
• Development and presentation of the hubEnsembles software package.
• Tutorial and case study using real-world data from the FluSight Forecast Hub.

Main Results:

• Demonstrated the utility of multi-model ensembles for improved outbreak forecasting.
• Introduced a flexible framework (hubEnsembles) for practical ensemble generation.
• Provided a reproducible case study for applying ensemble methods.

Conclusions:

• Multi-model ensembles offer enhanced accuracy and reliability in public health forecasts.
• The hubEnsembles package addresses practical challenges in generating and interpreting ensemble predictions.
• Standardized tools are crucial for advancing the application of ensemble methods in epidemiology.