COVID-19 and the difficulty of inferring epidemiological parameters from clinical data - Authors' reply.

We are grateful for Simon Wood and colleagues' comments on our study, which explore some important sensitivities in the data that were available early in the COVID-19 pandemic. Wood and colleagues' re-analysis puts more weight on the Diamond Princess outbreak data, arriving at an infection fatality ratio (IFR) in the range 0·23–0·65%, whereas our analysis used data from repatriation flights out of Wuhan, leading to an IFR in the range 0·39–1·33%. Both datasets are opportunistic, and neither is perfectly representative of the underlying population of interest. For example, although the Diamond Princess outbreak has a uniquely well characterised population, the transmission setting is unusual and therefore not necessarily representative of the broader populations that such estimates would be applied to. Furthermore, the health status of cruise ship passengers is not necessarily the same as the general population of a similar age, and the standard of care received by these passengers is likely to be different to that received in settings where the health system is under more strain. Given these limitations and the fact that the Diamond Princess outbreak data were incomplete at the time of our analysis (late February, 2020), we opted to focus on repatriation flight data.

Epidemics of novel diseases are inherently rapidly changing environments, which bring unique challenges from a data analysis point of view. Our position was neatly summarised by Michael Ryan, executive director of the WHO Health Emergencies Programme, who said that “perfection is the enemy of the good when it comes to emergency management. Speed trumps perfection.” Having early estimates, although imperfect, of the order of magnitude of the IFR (ie, knowing whether the IFR is nearer to 1% or 0·01%) is essential for strategic planning, and in this sense, the re-analysis by Wood and colleagues places the IFR on the same scale as our initial estimate. We also strongly support the call for appropriately designed prevalence studies, which are now urgently needed to provide direct estimates of the IFR with fewer limitations.