Updated Wuhan Mortality Data

[lrossouw]

I wanted to see how the upwards revised Wuhan mortality data may be affecting the IFRs by age in Verity et al. (2020).

I guess the IFRs may be directly affected if, as I surmise, the IFRs are estimated broadly speaking in pseudo code as:

ifr <- deaths_in_wuhan / infections_estimated_from_flights_data

Is this correct?

It's not clear to me which scripts need to be rerun to produce updated output.

[bobverity]

Thanks for the message, Yes, we have also been looking into the change that this makes to estimates. Pete - any chance you can put those changes you made onto another branch?

[lrossouw]

Thanks that would be useful. I would also suggest a run.R or a main.R that documents the order in which the scripts are executed to produce the results of the paper.

It could look something like this:

source("R_scripts\first.one.R")
source("R_scripts\second.one.R")

I suspect it may be running the three scripts `cfr_age_x' but not certain.

Also thanks for the work!

[pwinskill]

Hi, yes you are right that it is the 3 scripts named "cfr_age_x". The 2nd has the deaths data inputs (total and age-disaggregated) and runs the fitting.

We too are very interested in the updated Wuhan death estimates, and agree that we would likely expect the IFR estimates to increase. However, we have been quite hesitant to re-run the fitting in full at present for two reasons:

1. We do not know when the additional deaths occured. Specifically, with resepct to the age-stratified study conducted before the end of the epidemic.
2. We do not know how the addtional deaths are distibuted amongst age groups. The addtional (previously unreported) deaths are likely a biased sample and therefore not representative of all deaths.

To re-run would require assumptions on 1 and 2 - we are hoping that more informaton regarding these deaths may be released in coming days

[lrossouw]

The question is I guess what proportion of the deaths would have occured in the time period you are interested in? This NY times article says it's early deaths that were wrong which means the impact on your paper would be way more than the 50% being quoted as the increase.

I wanted to start modelling it and realised it requires and disproportionate adjustment on the data you used.

[becare-rocket]

If the China NHC censored the case data in Wuhan, we cannot use it as an estimate for IFR. Instead, this should be run on Spain or Italy data where censorship is unlikely.

[pwinskill]

Thanks @becare-rocket. The team have been continuously working to improve and refine estimates in light of new data. A large component of this has and continues to involve estimates for the UK and other European countries.