Transmission interval estimates suggest pre-symptomatic spread of COVID-19

Lauren Tindale; Michelle Coombe; Jessica E Stockdale; Emma Garlock; Wing Yin Venus Lau; Manu Saraswat; Yen-Hsiang Brian Lee; Louxin Zhang; Dongxuan Chen; Jacco Wallinga; Caroline Colijn

Serial intervals were estimated with the expectation-maximization approach described in Vink et al [17] . Briefly, this approach assigns the case with earliest symptom onset in the cluster as a "putative index" (PI) status, and uses the time difference between symptom onset of subsequent cases in the cluster and the putative index as "index case to case" (ICC) intervals for putative index cases in small clusters. The ICC intervals are the time differences between the symptom onset in the putative index (PI) and the others, so the collection of times t j − t pi for each j in the cluster. These intervals are not samples of the serial interval distribution, because it need not be the case that the PI infected the others. Vink et al [17] used a mixture model in which ICC intervals t j − t pi can arise in four ways: (1) an outside case infects PI and j; (2) PI infects j; (3) PI infects an unknown who infects j and (4) PI infects unknown 1 who infects unknown 2 who infects j.