Posted 24 April 2020
Two posts on one day. A busy day for numbers.
(Because I compare PCR tests and Antibody tests and compare CFR to IFR, reading my post from 22 April first will give this post more meaning).
As antibody tests become more available, more and more pockets of the world are now using them to ask the key question: How many people have already had COVID-19? Because so many cases can be asymptomatic or be associated with symptoms that are like a seasonal cold, we just do not know what proportion of the population has had the virus. All we know are the case numbers for people as they are hospitalised – a very different and much smaller figure.
This is a key parameter. If only 1% of the population (say) have already had COVID-19, then ending lockdown will lead to a big rise in cases (assuming that a much higher percentage of the population is susceptible to infection).
But, if (say) 20% of the population has had COVID-19 already and the susceptibility of the population is also around (say) 20%, then ending lockdown should have little consequence for new cases.
Lots of ‘ifs’, but it gives you an idea as to why the data is so important.
So, two parameters are key: What percentage of us have had COVOID-19 and what percentage of us are susceptible.
My article a few days ago mentioned one study where antibody testing had been used in one county of the U.S. in a bid to estimate an answer to the first question. That study found that the ‘official’ cases figure was probably under-estimated by a factor of 50-85.
Just so you can understand the method here, remember there are two types of testing: PCR and Antibody. The former tells you if you have a current infection at the day of test. One of these tests could be negative today (because you are not infected) but positive next week (because you have since become infected). This is the one that front line workers can now register for and it is used to decide if you should isolate. It is the test used at hospitalisation to determine the confirmed case number. My post looking at this explains why this has to be significant understatement of the actual case number.
The antibody test is the measure of whether you have had a COVID-19 infection but are now recovered. If that test is positive today, it will be tomorrow, the day after and so on. It is the robust measure of “had the infection” and so if we have a meaningful sample of a whole population and know how many of those test positive for the antibody, we have a good estimate of the percentage of the population that have had the infection (the actual case number).
Now there is a new study of New York City (random antibody testing of a sample of the population, not just those entering hospital) that suggests 21% of the population have had (and recovered from) COVID-19.
Rolling the numbers. The population of NY City is 8.4million. 21% of that is 1.8million. There have been 11,300 deaths so far in that city, so that gives a IFR of 0.6%. A closer estimate of the chances of dying if you get the infection.
The number of confirmed cases (the subset of cases tested at point hospitalisation is only 146,000). This gives a CFR of 11300/146000 = 7.7%. The number the media often confuse to how likely you are to die if you get the infection.
Seasonal flu in a non-pandemic year has an IFR of around 0.1-0.3. This can go as high as 0.5% in a pandemic year (like winter 2017/2018.
So, maybe we are getting a clearer picture that the IFR of coronavirus is rather closer to that of seasonal flue than the early models predicted.