Comparison of epidemic control strategies using agent-based simulations

A simulation of the dynamics of a small population is used to assess the impact of different confinement and testing strategies in the control of an epidemic. The simulation considers individuals as agents moving randomly across the habitat according to predefined urban patterns. Agents carry a simple tracing device that identifies signals emitted by other agents, recording the position and time of the encounter. The information of every device is propagated daily to an epidemic observatory based on an online graph database. Infections are simulated as stochastic processes depending on the proximity among individuals. Different epidemic control strategies are tested with and without the information of the tracing device under several scenarios. We observe that the success of the strategies strongly depends on the duration of the period of infectiousness before the presence of symptoms and the fraction of asymptomatic agents. If these values are high, strategies based on the presence of symptoms or on testing campaigns can hardly contain the epidemic. Strategies using massive confinement of the agents are able to control the epidemic at the cost of sending a large fraction of the population into quarantine. In cases with moderate and low values for these parameters, the tracing devices can provide a slightly better performance but only if a large fraction of the agents carry the device. Otherwise, the impact of these devices is found to be negligible in comparison with other strategies not using them. Finally, we provide a methodology allowing to use the information of the graph database to estimate basic parameters of the disease such as the infection probabilityWe acknowledge the support of the Particle Physics group (CMS Tier-2 project with reference FPA2016-78727-R) and the Advanced Computing and e-Science group at the Institute of Physics of Cantabria (IFCA-CSIC-UC). We also acknowledge the Ramon y Cajal program of the Spanish Ministry of Science. Finally, we want to thank the company Muon Systems for their computational resources and Prof. Francisco Matorras and Arturo Medela Ceballos for their support and fruitful discussions.N