PURPOSE: The COVID-19 pandemic showed preparing for future challenges is difficult, yet eminently necessary. Predicting uncertain future trends and planning for alternative scenarios is the new normal. This workshop will present modeling techniques and approaches that systematically deal with uncertainty around population health outcomes, treatment uptake, and health-care policy, and will discuss how these approaches can guide health care decision-making under uncertainty. Participants will be introduced to expert elicitation methods and modeling techniques used to generate actionable insights from expert assessments, and learn about the importance of high-quality input data and limitations of extrapolation methods. Participants will gain a broader understanding of the types of uncertainties to address in the context of COVID-19, such as estimated survival or vaccine uptake.

DESCRIPTION: This workshop will introduce both quantitative and qualitative techniques used to address different sources of uncertainties faced in health-care decision-making. Dr. Graf and Ms. Cope will provide an introduction and overview, along with an overview of uncertainties most commonly encountered (5 minutes). Prof. O’Hagan will provide an overview of expert elicitation methods like the Sheffield Expert Elicitation Framework (SHELF) and the Delphi method, and discuss how these methods can be used to inform key questions related to COVID-19 (20 minutes). Dr. Vardavas will discuss mathematical dynamic models, including complex population- and agent-based models as well as infectious disease transmission models, and will explore ways to address uncertainties around key modeling parameters for COVID-19 (20 minutes). Ms. Cope will review model applications for health economics (10 minutes), and will then, together with Dr. Graf, facilitate an interactive estimation exercise, where workshop participants will estimate key parameters such as COVID-19 survival rates, and also quantify their degree of uncertainty, using Zoom’s polling tool. Once submitted, audience estimates will be collected, visualized through an R-Shiny application, and discussed in detail (15 minutes).