Designing Efficient Contact Tracing Through Risk-Based Quarantining

Contact tracing for COVID-19 is especially challenging because transmission often occurs in the absence of symptoms and because a purported 20% of cases cause 80% of infections, resulting in a small risk of infection for some contacts and a high risk for others. Here, we introduce risk-based quarantine, a system for contact tracing where each cluster (a group of individuals with a common source of exposure) is observed for symptoms when tracing begins, and clusters that do not display them are released from quarantine. We show that, under our assumptions, risk-based quarantine reduces the amount of quarantine time served by more than 30%, while achieving a reduction in transmission similar to standard contact tracing policies where all contacts are quarantined for two weeks. We compare our proposed risk-based quarantine approach against test-driven release policies, which fail to achieve a comparable level of transmission reduction due to the inability of tests to detect exposed people who are not yet infectious but will eventually become so. Additionally, test-based release policies are expensive, limiting their effectiveness in low-resource environments, whereas the costs imposed by risk-based quarantine are primarily in terms of labor and organization.