A dynamic ensemble model for short-term forecasting in pandemic situations.

Area of Science:

• Epidemiology
• Machine Learning
• Public Health

Background:

• The COVID-19 pandemic strained hospital capacities and necessitated dynamic public health interventions.
• Existing predictive models struggled with the rapid evolution of the pandemic, including new variants and policy changes.
• Accurate forecasting is crucial for managing healthcare resources and mitigating economic impact during epidemics.

Purpose of the Study:

• To develop adaptive ensemble models for forecasting infectious disease dynamics.
• To investigate the utility of secondary metadata, such as Google searches, in improving epidemiological models.
• To enhance preparedness for future pandemic or epidemic situations.

Main Methods:

• Utilized ensemble modeling techniques allowing for dynamic adjustments in model composition and weighting.
• Integrated secondary metadata from Google searches to inform and enhance the ensemble predictions.
• Validated the approach using surveillance data for COVID-19, Influenza, and severe acute respiratory infections (SARI).

Main Results:

• Ensemble models demonstrated greater robustness and adaptability compared to individual predictive models.
• The inclusion of Google search data positively influenced the performance of the ensemble models.
• The proposed methodology showed promise in handling the complexities of real-world epidemic data.

Conclusions:

• Adaptive ensemble models offer a more resilient approach to forecasting in dynamic epidemic environments.
• Leveraging diverse data sources, including online search trends, can significantly improve epidemiological surveillance.
• This research contributes to building more effective tools for public health preparedness and response.