SIR (Susceptible, Infected or Recovered) stochastic network models are commonly used to describe the progression of epidemics inside a network. A task of interest in epidemiology is to use these models to estimate the state evolution, both at an individual as well as a population level. In this paper, we propose using continual testing to improve the state estimation at the individual level. Our testing is inspired from entropy reduction principles and requires only a small number of tests.