|
Abstract:
|
Contact tracing is an important tool in preventing the spread of infectious diseases. However, it proved challenging to implement for a rapidly spreading disease like COVID-19. We show that it is possible to improve the efficiency of contact tracing by combining microeconomic tools that measure individual heterogeneity with ideas from machine learning about sequential optimization. First, we incorporate heterogeneity in individual infectiousness with multi-armed bandits to establish an optimal contact tracing algorithm. This suggests a simple strategy: test a small number of an infected person's contacts and only test the remaining contacts if enough of the initial pilot contacts test positive. Under this framework, testing contacts of infected persons is first a way to ascertain whether the infected person is likely to be a "high infector." Second, we demonstrate using three administrative datasets from Pakistan, Punjab, and South India from contact tracing during COVID-19 that this strategy improves efficiency. Third, we show that this is a simple, easily implementable strategy. We believe that these results are immediately transferable to contact tracing of any epidemic.
|
