Strongly heterogeneous transmission of COVID-19 in mainland China: local and regional variation

Yuke Wang; Peter F.M. Teunis

When a sufficient number of infectious contacts is known, the serial interval distribution may be estimated from outbreak data. First, the elements of the transmission probability matrix V are estimated, using (plausible) starting values for the parameters of the serial interval distribution. Then, V is fixed and the serial interval distribution parameters are estimated. Then the serial interval distribution is fixed and V is estimated. This alternating procedure can be repeated until no more improvement (in posterior probability) is found [14] . In the present analyses a special node (0) was defined, that has a uniform kernel κ i,0 : any subject can be infected by node 0 at any time, within a given time range. Outside that time range κ i,0 = 0. Such a node could represent environmental transmission, i.e. from "the Market" to anyone in contact with that environment, or contact with a pool of infectious subjects, i.e. from any infectious subjects within Wuhan/Hubei to subjects outside of Wuhan/Hubei. As the outbreak progressed, travel to Wuhan became increasingly less likely, especially after the lockdown of the city of Wuhan on 23 January 2020. However, contacts did not cease abruptly: travel records from the confirmed cases in Beijing showed a gradually decreasing probability of contact with Wuhan. Therefore the probability of linking any confirmed cases outside Hubei province to Wuhan was modelled as a logistic function of the date of confirmation