Updated infectiousness profile based on correction of the He et al. 2020 paper

[mhoehle]

As mentioned in #384 the paper by He et al. (2020) used to formulate a proxy for the infectiousness profile has been updated, because in the estimation two (Infector, Infectee) pairs were unintentionally ignored [1]. Consequently, the authors also updated their analysis script on github - see ehylau/COVID-19#2 and in particular the file Fig1c_Rscript.R. As a consequence, the infectiousness profile used in transmission_risk.pdf would move slightly to the left:

Old:

New (note different x-axis):

The day before DSO still has the highest infectiousness, but now even more days before DSO have non-negligble probability mass. This has consequence for the computation of the transmission risk score/level motivated in transmission_risk.Rmd and used as part of the CWA:

Old:

New:

As far as I understand the classification this would mean that contacts 7 days before upload now also become relevant.
If one chooses to accept this PR one would also need to update the scores TRANSMISSION_RISK_DAY_[0-7] variables in ProtoFormatConverterExtensions.kt.

There might be epidemiological reasons for not following the He et al. update, for example, because the results from the paper (time of exposure relative to DSO in Infector for 70+ specific Infector-Infectee pairs) only serves as proxy for infectiousness in the infector relative to DSO. Furthermore, He et al. provide no statement about the plausibility of the the two accidently excluded serial intervals of -3 and -4 days - it appears likely that there is variability in DSO depending on what the symptoms exactly are and how these were queried.

It would be interesting to hear the opinion of the epidemiological experts about this - the issue #384 has been open for a while.

[mhoehle]

Superseded after 46 days by the work of f03f4cd (this PR is not mentioned nor commented)