A new algorithm developed by computational epidemiologists Sandor Beregi and Kris Parag offers a fresh approach to decision-making during infectious disease outbreaks. Published in the journal PLoS Computational Biology, this algorithm adapts model predictive control (MPC) from engineering to the complexities of epidemic surveillance. By incorporating non-pharmaceutical interventions like school closures and travel restrictions into their framework, Beregi and Parag aim to enhance decision-making processes with the assistance of computer technology.
MPC involves projecting future scenarios and selecting the best sequence of actions based on these projections, while also accounting for uncertainties. This method allows for short-term planning by evaluating the outcomes of each action and adjusting decisions accordingly. By connecting model forecasts with potential interventions, the researchers introduce a control theory approach to the field of epidemiology, addressing the challenges posed by delayed and incomplete data.
The algorithm developed by Beregi and Parag simulates various outbreak scenarios and evaluates different intervention strategies to minimize infections and associated costs. Compared to traditional strategies like event-triggered feedback control and time-triggered cyclic control, MPC consistently outperformed in terms of effectiveness and adaptability. However, the study also acknowledges the limitations of data-driven control, particularly when faced with significant reporting delays.
Implementing MPC in epidemiology presents technical challenges due to the discrete nature of epidemic interventions and their delayed effects. Despite these obstacles, the researchers demonstrate the potential of bridging disciplinary gaps between control theory and epidemiology to improve decision-making processes during public health crises. Their research currently lacks formal guarantees of optimality, a challenge they openly acknowledge. However, their work showcases that the method is both feasible and practical.
The Bigger Picture
Model predictive control and other formalized decision support methods offer a unique perspective during disease outbreaks. Instead of providing a definitive answer, these approaches bring mathematical precision and transparency to decisions made under pressure and uncertainty. By integrating epidemiological models with explicit cost considerations, MPC gives decision makers a clearer understanding of the outcomes of their choices.
The study conducted by Beregi and Parag highlights the potential and limitations of technical solutions. While algorithms can enhance human decision-making, they cannot replace the complex processes of prioritization and defining societal objectives.
