Optimizing non-pharmaceutical intervention strategies against COVID-19 using artificial intelligence.

Area of Science:

• Epidemiology
• Computational modeling
• Public health policy

Background:

• Planning non-pharmaceutical interventions (NPIs) during the COVID-19 pandemic was crucial for controlling spread while minimizing societal and economic impact.
• Increasing data availability enabled modeling of infection trends and intervention costs, framing intervention planning as an optimization challenge.

Purpose of the Study:

• To propose a computational framework to assist policymakers in planning and adapting NPIs over time.
• To optimize the selection and timing of NPIs to effectively manage the pandemic.

Main Methods:

• Developed a hybrid machine-learning epidemiological model for infection forecasting.
• Aggregated socio-economic costs from existing literature and expert knowledge.
• Employed a multi-objective optimization algorithm to identify and assess intervention plans.

Main Results:

• The framework, trained and tested globally, demonstrated modularity and adaptability to real-world scenarios.
• Proposed intervention plans generally surpassed real-world strategies in reducing infections and intervention costs.

Conclusions:

• The developed framework provides a robust, data-driven approach to NPI planning.
• This optimization strategy offers a valuable tool for public health policy during pandemics.