B1703
Title: State-space epidemiological compartmental models: Approximations, control, forecasting, and real-world complications
Authors: Logan Brooks - Carnegie Mellon University (United States) [presenting]
Abstract: Epidemiological compartmental models describe the state of a population using the amounts of individuals falling into certain predefined categories encapsulating their health status, location, age, and/or other characteristics. The evolution of the population state describes the transition of individuals between compartments as their status changes due to interactions with other individuals or the passage of time. The mean change in a compartment's membership at some time is typically a quadratic function of the population's current state. These dynamics can be difficult to analyze, and paired with an observation model, challenging to fit. Linear dynamical approximations lead to statements about herd immunity and objectives for disease control. Similarly, various manipulations of the state-space model equations motivate additional types of approximations and compartmental-model-inspired forecasting frameworks; sufficiently constrained models can also be fitted with general-purpose frameworks. However, real-world details of epidemiological surveillance systems and disease dynamics create complications for common model structures. We will discuss some of these compartmental modeling frameworks, approximations, and complications in the context of seasonal epidemics of influenza-like illness and the COVID-19 pandemic.
