Optimal targeting of interventions uses estimated risk of infectiousness to control a pandemic with minimal collateral damage

Abstract

1AbstractIn this paper, we present a simple model that shows how to optimally target interventions based on the estimated risk of infectiousness of individuals. Our model can help policymakers decide when to use different types of interventions during a pandemic, depending on their precision, which is the fraction of positive predictions that are true positives. We show that targeted interventions, even with very low precision, can impose a much smaller overall burden on the population than non-targeted alternatives, such as lockdowns or mass testing. To illustrate this, we use data from the NHS contact tracing system in the UK to construct a risk function based on second degree contact tracing, which is similar to the strategy used by Vietnam in 2020. We find that with moderate precision (greater than 1/1000) and sufficient sensitivity (greater than 1 − 1/R0), countries can cope with a large number of imported cases without resorting to social distancing measures, while keeping the per-person probabilities of both infection and quarantine very low. We also show that targeted strategies are often orders of magnitude better than default strategies, making them robustly beneficial even under significant uncertainty about most parameters.