I thank
Milind Watve for a thought-provoking blogpost ‘Covid 19: We need neither vaccine nor herd immunity’ (dated 27th May 2020). The key argument
of blogpost is death rate of Covid19 is declining and in future it will be like
seasonal flu.
 |
| https://milindwatve.home.blog/2020/05/27/covid-19-we-need-neither-vaccine-nor-herd-immunity/?fbclid=IwAR1D0wMndXAMT5p41H5yyfklETgEEpnzzB72kOWNn1UykOx5nk8oaelD-ZY |
It is based on this graph that blogger concludes that there is a
downward trend in CFR. He further states, that after certain number of days (35
as per his prediction, though he has expressed caveats about it), ‘Covid 19
will remain only as dangerous as any seasonal flu’ (emphasis mine). I
would like to note that blogger has used CFR when he wants to talk about mortality
rate. He has also said that in growing pandemic CFR will be an underestimate of
mortality rate.
I am going to argue in this write-up is following:-
CFR representation of blogger is wrong and his conclusion weakens considerably when CFR is corrected.
Based on CFR, seasonal flu seems to be at least as deadly as Coronavirus. But seasonal flu has extremely low mortality rate as it does not affect the large population. As per government statistic, 1218 deaths have been caused by seasonal influenza (H1N1).
In growing pandemic, CFR will always be above mortality rate and it will converge to mortality rate as large and large section of population is affected.
Let me first explain Case Fatality rate and mortality rate. In 2019, 28798 Indian suffered due to seasonal flu and 1218 out of them died. Suppose Indian population in 2019 was 130 crore or 1300 million.
Mortality rate of a disease is deaths due to disease divided by
total population for given period, say a year. Hence, mortality rate of seasonal flu for India in 2019 is 1218 in 1300 million, which is less than 1
per million. Case fertility rate is ratio of deaths due to disease to total
infections due to disease for given period. So, CFR of seasonal flu 1218 to
28798, around 4% or in other words 4 out of 100 seasonal flu patients in 2019
died.
If one wants to calculate the CFR for the day, then ideally one
should consider people who contacted the disease in that day and track them to
see how many of them succumbed to that disease. One cannot use deaths and new
cases on the same day. Deaths in the numerator of CFR should have those cases
which have originated on the same day people who died got detected. (A good illustrative read)
For
Covid-19, the typical hospitalization is 2 weeks. (An example of research here)If we assume that person is
moved to hospital or isolated on any day, in 2 weeks’ time, her or his case
will be resolved: the person will be free of virus or person would succumb to
virus. Consider this artificial example.
Artificial Example for CFR
|
Week start |
New cases |
Deaths |
Same day CFR |
True CFR |
|
Week 1 start |
100 |
0 |
- |
- |
|
Week 2 start |
200 |
0 |
- |
- |
|
Week 3 start |
500 |
5 |
1% |
5% |
|
Week 4 start |
1500 |
10 |
0.67% |
5% |
|
Week 5 start |
5000 |
25 |
0.5% |
5% |
|
Week 6 start |
20000 |
75 |
0.0375% |
5% |
Since cases are rising rapidly, it generates an illusion of falling CFR. But we must consider total cases from past date, to get the accurate CFR. In my artificial example, true CFR is constant. Since most vulnerable people (in terms of susceptibility to disease) will get the virus first and likely to succumb rapidly. Hence even true CFR is likely to have a declining nature. But once the especially susceptible population has been ravaged, CFR will stabilize and will not fall downward.
Here, I am comparing the CFR as reported
by blogger to t-14 CFR where ratio of deaths to cases 2 weeks back is
considered. So t-14 CFR for 26th April is ratio of deaths on 26th
April to new cases on 12th April.
First, the downward nature is now
not so evident, though after 12th May, both the graph look similar.
If we go for linear trend in both
cases there is downward sloping linear trend. But linear trend is too simple an
explanation to model such complex process. If we go for somewhat advanced method
called moving average, we see following.
CFR can go
up as well as come down, as the t-14 CFR shows. In fact, after May 20, t-14 CFR
has started showing upward movement. It is not surprising. Stable patterns are
observed when diseases in evenly spread across geographies. After 1st
week of May, Coronavirus has reached to various rural parts of India due to
migrant movement. Unless there is exceptional containment, vulnerable
population from these parts will start succumbing to virus and hence CFR can
rise around end of May.
Coming
back to original blogpost, changing the CFR measure and prediction method to
more accurate ones changes the conclusion of blogger considerably. CFR cannot
be expected to have lowering trend.
The broader
point is even if CFR is eventually going to have a downward and stable (asymptotic)
nature, eventually stabilizing to a value like 1% as per many researchers, that
itself is not the reason to get complacent and say that we neither need vaccine
nor herd immunity. 1% looks small number, but we should understand this 1% adds
to low probability of death. Crude death rate in India in 2012 was 0.7%, 7 among
1000 Indian died in 2012. If widespread across the country, Covid19 will kill roughly 10 Indians more and total deaths
will rise to 17, doubling the crude death rate. We speed at which we will see people
around us dying will be more than double even at 1% CFR. We need exceptionally
low CFR, like closer to 0 or extremely low incidence of Covid19 like seasonal
flu. For both we need vaccine or immunity of some form.
So far, Covid 19 seems to be worse than seasonal flu. (here) May be over the years, the speed at which is spreads will reduce, but so far available evidence does not support that claim. Covid 19 will be same as seasonal flu only if CFR for Covid 19 goes very low as Covid 19 is much more widespread than flu. Blogger's downward trend predict that. But this downward trend relies heavily on linear method and it is very likely that trend is based on only partial diffusion of disease during lockdown period. Hence, blogger's argument at at best very optimistic and at worst, inaccurate. CFR evidence is without doubt an inaccurate one.
--
Quarantine and medical care of existing patients are ways of coping with Covid19 and they are important. But these coping mechanisms have huge economic cost, at present and in future if we do not get a preventive or curative way. Using quarantine and medical care to fight with Covd19 is to keep alive from ticking bomb by continuously standing up on the fuse so that it doesn't burst. We won't be killed by bomb, but we will be hugely deprived. If that is the way to live, we could have lived in trees and never ventured in dangerous jungles, thousand years back. Vaccine is gamble, but we must take it. We cannot have sour grape approach towards it.
We cannot get complacent
by simple but wrong arguments, like the blogpost which prompted this response. Understanding
Covid19 and its spread is a complex process and it often requires complicated
argumentation. There are experts who are trying their best to do so and many of
them are trying to communicate the same with us. I will post some links here.
You can try
to understand these arguments. It tough, but that is the way it is. Do not fall
for convincing, simple but inaccurate arguments. They give false sense of hope
against harm, which can be as harmful as harm itself. But I think damage is done. It is so easy to grasp wrong but simple argument, corrective argument requires some effort. Let's hope it is not the case.
--
Some helpful links:-
1. Data about Covid19 in India - https://www.covid19india.org/
2. Paul Romer on testing - a hour long podcast https://www.econtalk.org/paul-romer-on-the-covid-19-pandemic/
3. A podcast on second wave in India once lockdown eases - https://www.youtube.com/watch?v=3vlO3l9CuD8