The Analysis of Covid-19 Surveillance Data: What can we Learn from Limited Information?
What can we learn about the transmission process of SARS-CoV2 from daily counts of confirmed cases and tests as the pandemic unfolds? In this paper we use classic econometric techniques to filter away stochastic innovations that occur in epidemiological surveillance data that are orthogonal to the infection process. We then compute the effective reproduction number Rt and the test positivity rate t , and propose an analysis of joint trajectories over the (Rt; t) space to have a more precise assessment of the current status of the infection. We test our method using an agent-based model with an underlying SIERD infection process and a testing layer that produces confirmed positives. We find that epidemiological indicators estimates are systematically biased, but our method allows us to reduce the root mean squared error and the probability of type II errors, particularly when testing is concentrated among symptomatic patients. The joint analysis of Rt and t manages to reduce the probability of type I and type II classification errors to below 0.5%. Our country analysis shows that daily counts of cases and tests exhibit strong seasonal and atypical components which, if left untreated, can produce spurious dynamics of epidemiological indicators.