Disease control as an optimization problem.

Area of Science:

• Epidemiology
• Computational Biology
• Public Health Policy

Background:

• Traditional disease control policy development relies on intuition and simplified models, risking the exclusion of more effective, complex strategies.
• Existing control theory approaches struggle with uncertainties in initial conditions, model parameters, and stochastic disease spread dynamics.

Purpose of the Study:

• To develop and apply advanced optimization techniques for discovering complex, data-driven disease control policies.
• To overcome limitations of traditional methods by incorporating machine learning and convex optimization for broader policy exploration.
• To demonstrate the framework's capability in optimizing policies for disease eradication, using COVID-19 as a case study.

Main Methods:

• Utilized convex optimization theory and machine learning algorithms for policy optimization.
• Developed a framework capable of handling arbitrary uncertainties in initial conditions and model parameters.
• Incorporated stochastic models and allowed for optimization over policies with continuous or discrete weekly measures (e.g., lockdowns).

Main Results:

• Successfully optimized disease control policies over hundreds of parameters, moving beyond simplified policy families.
• Demonstrated the ability to handle complex uncertainties and stochastic elements inherent in disease transmission models.
• Applied the framework to the Susceptible-Exposed-Infected-Recovered (SEIR) model to minimize COVID-19 eradication time.

Conclusions:

• Advanced optimization techniques offer a more comprehensive approach to designing disease control policies.
• The proposed framework effectively addresses uncertainties and complex policy structures, potentially leading to more efficient disease management.
• This method provides a powerful tool for optimizing public health interventions, exemplified by its application to COVID-19 eradication.