We develop a tractable non-parametric model for the time-varying reproductive rate of infectious diseases that combines the structure of a deterministic compartmental model and a stochastic model for incidence data. We use Bayesian inference to estimate, with uncertainty, the reproductive rate of the Coronavirus 2019 outbreak in the U.S. states of California, Florida, Michigan, New Mexico, New York, and Texas from January 2020 to March 2022. We use the inferred reproductive rates to estimate the posterior distribution of the time-varying reproduction numbers for each state. Compering the time-varying reproduction numbers across the states, we identify some epidemic waves, potentially driven from changes in human behavior and virus mutations.