ABSTRACT
Epidemic spreading can be suppressed by the introduction of containment measures such as social distancing and lockdowns. Yet, when such measures are relaxed, new epidemic waves and infection cycles may occur. Here we explore this issue in compartmentalized epidemic models on graphs in presence of a feedback between the infection state of the population and the structure of its social network for the case of discontinuous control. We show that in random graphs the effect of containment measures is simply captured by a renormalization of the effective infection rate that accounts for the change in the branching ratio of the network. In our simple setting, a piece-wise mean-field approximation can be used to derive analytical formulae for the number of epidemic waves and their length. A variant of the model with imperfect information is used to model data of the recent COVID-19 epidemics in the Basque Country and Lombardy, where we estimate the extent of social network disruption during lockdowns and characterize the dynamical trajectories in the phase space.