Optimal, near-optimal, and robust epidemic control

Abstract In the absence of drugs and vaccines, policymakers use non-pharmaceutical interventions such as social distancing to decrease rates disease-causing contact, with aim reducing or delaying epidemic peak. These measures carry economic costs, so societies may be unable maintain them for more than a short period time. Intervention policy design often relies on numerical simulations models, but comparing policies assessing their robustness demands clear principles that apply across strategies. Here we derive theoretically optimal strategy using time-limited intervention reduce peak prevalence novel disease in classic Susceptible-Infectious-Recovered model. We show broad classes easier-to-implement strategies can perform nearly well strategy. But neither nor any these near-optimal is robust implementation error: small errors timing produce large increases prevalence. Our results reveal fundamental control expose potential fragility. For control, an must strong, early, ideally sustained.