Identifiability of local transmissibility parameters in agent-based pandemic simulation

Agent-based simulation is one of the approaches that can be applied to simulate the transmission of infectious disease such as influenza within a city. Several types of agents with different behaviours are allocated to a model city and the transmission between city residents is stochastically solved locally. Simulations corresponding to specific intervention measures are carried out. However, due to the large number of parameters in the simulation, which cannot be fully constrained by surveillance evidence and epidemiological knowledge, one sometimes judges candidate intervention measures from simulation results with arbitrarily fixed parameters. In the present study, we have conducted numerical experiments to estimate reproduction numbers (transmissibility parameters) in workplaces and in homes from pseudo-observation time-course data generated by a simulation run. This pseudo-observation is generated under the assumption that transmission in workplaces is more effective than in homes. The ratio of these numbers are considered to affect the response to intervention. Our experiments indicate that a profile consisting of the total number of patients is insufficient; rather, a role-specific profile is needed to reconstruct the assumption on the ratio of reproduction numbers.