Sequential Classification Under Uncertainty: A Mathematical Toolkit for Decision Makers [Funder: Carleton University COVID-19 Rapid Research Response Grants]

Problems individuals face during a pandemic include whether a patient should present themselves for testing, whether a healthcare professional should recommend testing, whether a clinician should declare a test positive, and whether a policy maker should recommend relaxing social distancing. In each instance, the decision maker has to classify signals (the symptoms, the test result, the number of confirmed cases) in a binary manner (suitable/unsuitable for testing, infected/uninfected, discontinue/continue social distancing), but with uncertainty about their correct choice and different associated payoffs. The standard framework for these classificatory problems is signal detection theory (SDT). However, in an epidemiological setting, several discriminative decisions are made sequentially. Our project will apply sequential SDT to classification problems in epidemiology to directly estimate parameters of greatest value to decision makers. By combining this approach with classical SIR models, we will provide tools for signal classification that will serve decision makers dealing with COVID-19.