Session Description:
Dr. Lewis will provide a broad overview of the challenges posed in attempting to create epidemiological models of COVID-19 spread in Los Angeles County, building on his work as the Director of Covid-19 Demand Modeling for the Los Angeles County Department of Health Services. The approach to be described uses a standard epidemic compartment model; however, a fully Bayesian approach was taken to facilitate the rigorous incorporation of uncertainty and thus to avoid making precise but meaningless predictions. The sequence of weekly predictions and the actual trajectory of the epidemic will provide a foundation for the presentation. The work was conducted by a multidisciplinary team who used statistical, epidemiological, medical, machine-learning and related strategies in developing epidemiological the prediction to aid in hospital preparedness and response across Los Angeles County, informing public health decision making and impacting a population of 10 million persons. Through this work, the team identified solutions to a number of outstanding epidemic modeling challenges and, at the same time, identified others for which no solution could be found because of a fundamental lack of valid data or a scientific foundation.
Session Objectives:
- To understand basic epidemic compartment models and their application in epidemic prediction modeling
- To understand conceptually the Bayesian formulation of the Susceptible, Exposed, Infected, Recovered (SEIR) model and how it was used for epidemic prediction modeling in Los Angeles County
- To understand the limitations of available case data in informing estimates of the incidence, severity, and spread of COVID-19
- To understand how the model can be used to estimate projected needs for hospital beds, intensive care, mechanical ventilation, and morgue services
- To understand the predictive performance of the Bayesian SEIR model in predicting the trajectory of the epidemic since its beginning;
- To understand how critical gaps in understanding true community infection burden and the extent and duration of naturally-acquired immunity limit the ability to create accurate epidemic models going forward.
Expect to Learn:
Attendees should expect to acquire an understanding of the mechanics, limitations, and value of epidemic prediction models, improving their ability to interpret model results that they see presented in scientific publications or the lay press.
Prerequisites:
None
Course Type: Lecture