Now-casting the COVID-19 epidemic: The use case of Japan, March 2020

Stephan Gloeckner; Gerard Krause; Michael Hoehle

For the COVID-19 in China, The World Health Organization (WHO) published a median reporting delay of 12 days (range 8-18 days) at the start of the outbreak and three days (range 1-7 days) in early February. 3 Disease transmission models report a reporting delay larger than ten days during the start of the novel coronavirus outbreak and four days delay after the implementation of the quarantine procedures on January 27 th . 4 A study monitoring transmission potential of COVID-19 in Singapore estimated a reporting delay of 6.9 days. 5 In contrast, the 2012 MERS-COV outbreak in Saudi-Arabia had a four day reporting delay 6 and during the 2018 Measles outbreak, Japan showed a mean reporting delay of 4.5 days. 7 Now-casting uses previous data to estimate the delay distribution while taking into account that cases with long-delays are currently not observed due to right-truncation with the aim to adjust epidemic curves for "observed-but-not-yet-reported cases". 8, 9 On the occasion of the 2011 STEC/HUS outbreak in Germany the now-casting method was further optimized for the specific situation of being in the midst of an acute outbreak with interventions to actively reduce the delay. 10 We report here on an improved concept of now-casting, that allows the adjustment of epidemic curves by taking into account that information about symptom onsets may not be available for a relevant proportion of the case reports. We demonstrate the approach with data from the COVID-19 outbreak in Japan, because of its high case counts, accessibility to data details and the particular importance of such an approach for the risk assessment for the upcoming Summer Olympic Games, hosted in Tokyo, July to August 2020. The aim of this approach is to reduce and transparently display uncertainty on the real time status of epidemic curves.