EpiCURB: Learning to Derive Epidemic Control Policies

The effectiveness of an epidemic control policy relies largely on how much effort is invested in every public health measure. Unfortunately, it is seldom possible to optimally allocate funds to these measures if the isolated effect of each intervention cannot be reliably estimated. We show how this challenge can be overcome by utilizing EpiCURB, a simulation-control framework that enables us to measure the effect of both untargeted and prioritized interventions on the epidemic outcome, where the latter are guided by reinforcement learning routines that effectively rank eligible individuals.