[Jasper (OP)]

Recent articles in the Spectator, building on epidemiological research from Oxford, suggested that infection rates might be significantly (say, up to 50% of the population) higher than in the current Imperial models. In turn, this might mean that the death rate from Covid-19 could be significantly lower than current headlines suggest. How credible is this view and is there any hard evidence to support or challenge it?

[Petrochemicals]

Yep, stands a chance, given that we have little testing. Its all over england, all the major conurbation centres and the cases are not decreacing like other countries. That being said, the uk has a population who has harsh winters every year so genetically evolution has probably made us more resistant to these sort of illnesses. Probably why people from ethnicities of hot sunny weather historically are seeing a higher mortality rate with an infectiin rate at the same level !

[chris]

It's highly likely that infection rates are higher than we thought or think. Partick Vallance (UK chief scientific officer) suggested at one point that the ratio of deaths to actual infections might be close to 1:1000; so with 10,000 plus deaths there may well have been 10 million cases in the UK already.

What will solve this is widescale serological testing. Once we can test the UK at scale for antibody to infection and discover how many people have really had it, we can refine and reinforce our models and control measures accordingly. Personally, I suspect that there has been considerable widescale spread in the community already.

[NigelD]

At the moment the ratio of 1:1000 is right at the extreme of the range of possible deaths/infections ratio and a more generally accepted figure, based on data from places such South Korea where there was more widespread testing is about 1%.   At the moment the UK figures show a mortality rate of about 17% of the (Pillar 1) people testing positive.   This suggests that the actual number of UK cases is about 17 times the government figures.

[chris]

Thanks @NigelD

a more generally accepted figure, based on data from places such South Korea where there was more widespread testing is about 1%.

Have you got a reference for this, so we can evidence-base the discussion here, in case people want to follow up? Thanks.

[NigelD]

A reference is "Transmission potential and severity of COVID-19 in South Korea"  International Journal of Infectious Diseases
Volume 93, April 2020, Pages 339-344

[chris]

Thanks; I'll give that a read this weekend. A little light distraction!

[set fair]

14% in a study of about 200 pregnant women in New York... if they regularly attend hospital that could might make the figure unrepresentedly high.

New england Journal of Medicine

https://www.nejm.org/doi/full/10.1056/NEJMc2009316

[NigelD]

The New York data may suggest that we are significantly undercounting the total number of cases.   I have a toy mode that pretty accurately matches the UK data so far, but does not allow me to estimate the degree to which we might be undercounting.  It does however show that the degree of undercounting could make a big difference to how things might progress in the next month or so in that if the mortality rate is indeed about 1:1000 then we may be getting closer to herd immunity than we realise.
We now appear to have sufficient underused testing capacity that we could get a good idea of the UK undercount  by testing people who present at hospitals for other reasons, e.g physical injury (less biased than testing people who are pregnant), and try and arrange things so that the demographics of the people being tested are representative of the UK demographics.  A sample size of about 1000 would give a good indication of this and is now within the testing capacity, and would provide very useful data for planning what should be done next.