Building an epidemiological modelling toolkit for epidemic preparedness
Lead Research Organisation:
London School of Hygiene & Tropical Medicine
Department Name: Infectious and Tropical Diseases
Abstract
This proposal aims to change how we predict and manage infectious diseases through the application of advanced mathematical methods to the development and analysis of models of infectious disease transmission. Recognizing limitations in current models exposed by the COVID-19 pandemic, such as delayed response times and challenges in adapting models to a new disease, this project proposes a novel approach. It centres on creating a flexible, transparent toolkit for building transmission models that can rapidly adapt to new information, enabling real-time decision-making during health crises.
The methodology combines applied category theory (ACT) with operational modeling to simplify the construction and adaptation of complex disease models. ACT is a mathematical framework to describe complex systems in a structured and relational way, like a language for understanding how different parts of a system can fit together and interact, making it particularly useful for building and analyzing models in various fields, including epidemiology. By decomposing models into reusable components, the project intends to make the process of modeling more accessible and adaptable, fostering an environment where models can be quickly tailored to specific diseases or scenarios. Moreover, the research addresses the integration of these models with decision-making processes, acknowledging the uncertainty inherent in predicting disease spread and the effectiveness of interventions. It explores optimizing decisions under this uncertainty, aiming to provide robust support for public health strategies.
Applications of this research will be demonstrated through models of measles and SARS-CoV-2, the virus that causes COVID-19, showcasing the toolkit's ability to replicate existing models and create new ones that can inform policy decisions. These demonstrations will benefit from a large and detailed dataset on the transmission dynamics of these infections, and will be run in a trusted research environment, where detailed epidemiological information can be used in the models in a safe, secure manner. Through workshops and open-source distribution of the underlying software, the project seeks to empower modelers, policymakers, and researchers, enhancing preparedness for future pandemics. This initiative not only advances the field of epidemiological modeling but also contributes to a more informed and flexible response to public health threats, potentially saving lives and resources by enabling swift, evidence-based action.
The methodology combines applied category theory (ACT) with operational modeling to simplify the construction and adaptation of complex disease models. ACT is a mathematical framework to describe complex systems in a structured and relational way, like a language for understanding how different parts of a system can fit together and interact, making it particularly useful for building and analyzing models in various fields, including epidemiology. By decomposing models into reusable components, the project intends to make the process of modeling more accessible and adaptable, fostering an environment where models can be quickly tailored to specific diseases or scenarios. Moreover, the research addresses the integration of these models with decision-making processes, acknowledging the uncertainty inherent in predicting disease spread and the effectiveness of interventions. It explores optimizing decisions under this uncertainty, aiming to provide robust support for public health strategies.
Applications of this research will be demonstrated through models of measles and SARS-CoV-2, the virus that causes COVID-19, showcasing the toolkit's ability to replicate existing models and create new ones that can inform policy decisions. These demonstrations will benefit from a large and detailed dataset on the transmission dynamics of these infections, and will be run in a trusted research environment, where detailed epidemiological information can be used in the models in a safe, secure manner. Through workshops and open-source distribution of the underlying software, the project seeks to empower modelers, policymakers, and researchers, enhancing preparedness for future pandemics. This initiative not only advances the field of epidemiological modeling but also contributes to a more informed and flexible response to public health threats, potentially saving lives and resources by enabling swift, evidence-based action.
Technical Summary
This proposal describes an innovative approach to epidemiological modeling, leveraging Applied Category Theory (ACT) to enhance the flexibility, transparency, and adaptability of infectious disease models. The methodology outlined integrates ACT with operational modeling to decompose complex models into modular components, facilitating rapid adaptation to emerging diseases and real-time decision-making in public health crises.
Key aspects include:
1. Applied Category Theory: Utilized to formalize the structure of epidemiological models, ACT enables the decomposition of models into composable and reusable components. This approach simplifies the complexity inherent in modeling infectious diseases, allowing for more accessible modification and understanding of model dynamics. This aspect of the proposal will help researchers translate concepts used in ACT to those commonly used in epidemiological modeling. This approach also allows us to abstract model structure, in terms of the number and type of compartments and how they are connected, from modeling paradigm (ordinary differential equations, stochastic differential equations, Markov models, discrete event simulations, etc.).
2. Modular Modeling Framework: The proposed toolkit for building transmission models can be quickly tailored to specific pathogens or epidemiological scenarios. This framework supports the assembly of models from predefined modules, reducing development time and improving the responsiveness of modeling efforts to real-world needs. This approach also lends itself to better transparency in the modeling process, as it can show how complex models can be generated incrementally from simpler ones, exposing the underlying assumptions along the way, and avoiding the need to try to mentally 'reverse engineer' a model from the final code implementation.
3. Real-time Adaptation and Decision Support: Emphasizing the need for models that can adapt to new information and guide public health decisions in real-time, the research advocates for a dynamic modeling approach. This includes mechanisms for integrating real-time data, adjusting model parameters on-the-fly, and optimizing decision-making under uncertainty, which may stem from random chance (as is the case in stochastic models), as well as from uncertain knowledge of key epidemiological parameters.
4. Case Studies: To demonstrate the applicability of the proposed methodology, the documents present case studies on measles and SARS-CoV-2. These examples illustrate how the toolkit can replicate existing models and facilitate the development of new models that provide actionable insights for policy decisions. These studies will involve running the models in a trusted research environment, which provide secure access to fine-grained data, which are necessary for calibrating complex models.
5. Community Engagement and Open-Source Distribution: The project aims to engage with a broad community of modelers, policymakers, and researchers through workshops and open-source distribution of the modeling toolkit. This initiative seeks to democratize access to advanced modeling tools, enhancing the collective capacity to respond to future pandemics.
Overall, the research represents a significant step forward in epidemiological modeling, offering a robust framework for constructing adaptable, transparent, and user-friendly models that can inform public health strategies in the face of emerging and re-emerging infectious diseases.
Key aspects include:
1. Applied Category Theory: Utilized to formalize the structure of epidemiological models, ACT enables the decomposition of models into composable and reusable components. This approach simplifies the complexity inherent in modeling infectious diseases, allowing for more accessible modification and understanding of model dynamics. This aspect of the proposal will help researchers translate concepts used in ACT to those commonly used in epidemiological modeling. This approach also allows us to abstract model structure, in terms of the number and type of compartments and how they are connected, from modeling paradigm (ordinary differential equations, stochastic differential equations, Markov models, discrete event simulations, etc.).
2. Modular Modeling Framework: The proposed toolkit for building transmission models can be quickly tailored to specific pathogens or epidemiological scenarios. This framework supports the assembly of models from predefined modules, reducing development time and improving the responsiveness of modeling efforts to real-world needs. This approach also lends itself to better transparency in the modeling process, as it can show how complex models can be generated incrementally from simpler ones, exposing the underlying assumptions along the way, and avoiding the need to try to mentally 'reverse engineer' a model from the final code implementation.
3. Real-time Adaptation and Decision Support: Emphasizing the need for models that can adapt to new information and guide public health decisions in real-time, the research advocates for a dynamic modeling approach. This includes mechanisms for integrating real-time data, adjusting model parameters on-the-fly, and optimizing decision-making under uncertainty, which may stem from random chance (as is the case in stochastic models), as well as from uncertain knowledge of key epidemiological parameters.
4. Case Studies: To demonstrate the applicability of the proposed methodology, the documents present case studies on measles and SARS-CoV-2. These examples illustrate how the toolkit can replicate existing models and facilitate the development of new models that provide actionable insights for policy decisions. These studies will involve running the models in a trusted research environment, which provide secure access to fine-grained data, which are necessary for calibrating complex models.
5. Community Engagement and Open-Source Distribution: The project aims to engage with a broad community of modelers, policymakers, and researchers through workshops and open-source distribution of the modeling toolkit. This initiative seeks to democratize access to advanced modeling tools, enhancing the collective capacity to respond to future pandemics.
Overall, the research represents a significant step forward in epidemiological modeling, offering a robust framework for constructing adaptable, transparent, and user-friendly models that can inform public health strategies in the face of emerging and re-emerging infectious diseases.
