Twitter Thread: bit.ly/CovidMath
Why do y'all make ME do the #Math all the time? I'm so f'n #TiredOfBeingRight.
Why we SHOULD take precautions against #CoVid19 #Coronavirus and how is it NOT LIKE the yearly flu numbers - A Thread:
First, "The Flu":
Best numbers available are 2016-2017. More recent numbers are still in the 'estimate' category, so we'll use the '16-'17 numbers as baseline and reference more recent data as estimates only.
All data is US only, unless noted otherwise.
[All text in brackets are later edits.]
2016-2017 Flu data:
Symptomatic (assumed) - 29,220,523
Medical visits (confirmed) - 13,633,446
Hospitalizations - 496,912
Deaths - 38,230
https://www.cdc.gov/flu/about/burden/past-seasons.html
https://www.cdc.gov/flu/about/burden/2016-2017.html
Population estimate 2016 - 318 million
9% of the population was symptomatic
4% were confirmed medical visits
[0.15%] hospitalized
Of confirmed cases, mortality rate is 0.28%.
Of assumed cases, mortality rate is 0.13%
Second, "H1N1 2009 swine flu" (US only):
Assumed cases: 60.8 million cases
Medical visits [(confirmed)]: Unknown (like Covid19, testing data is incomplete)
Hospitalizations: 274,304
Deaths: 12,469
https://www.cdc.gov/flu/pandemic-resources/2009-h1n1-pandemic.html
Population estimate 2009 - 308 million
20% of the population was symptomatic
0.09% hospitalized
Of assumed cases, mortality rate is 0.021%.
If we assume, like seasonal flu, confirmed mortality is 2X's assumed, the baseline H1N1 "confirmed" mortality would be about 0.04%.
We can already see that H1N1 wasn't that bad compared to seasonal flu. Or was it?
To understand this, you have to understand how many strains of flu virus are out there, in "the wild" if you will. There are four basic types: A, B, C, D.
https://www.cdc.gov/flu/about/viruses/types.htm
Types A & B are your typical human flu viruses. Type C can sometimes cause mild symptoms. Type D mainly affects cattle.
Type A is split into 2 subtypes: H & N. There are 18 H types and 11 N types. Of these combinations, only 131 subtypes have been detected in nature thus far.
These 131 subtypes can be further broken down into 'groups' and 'subgroups'. But for sake of discussion, we will only focus on Type A's 131 subtypes.
Type B is split into 2 lineages: Y & V. These also can be divided by group and subgroup.
Since Type D affects cattle, we will disregard it.
Since Type C is often mild, we will disregard it.
Since Type B is rarely known to cause pandemic, even though severe, we will disregard it.
So, of the 38,230 "flu" deaths, these are attributable to those 131 strains of Type A.
So when you consider that H1N1 was "only" 32% as deadly as the seasonal flu, what you're really saying is that ONE virus is 32% as deadly as [all] 131 viruses [combined].
Or, more accurately, H1N1 (12,469 deaths) is 42X's more deadly than a single seasonal flu virus (292 deaths per virus).
Finally, what we know of the Coronavirus so far:
Symptomatic (assumed) - unknown (incomplete data)
Medical visits (confirmed) - 558,599 (as of 4/12/2020)
Hospitalizations - 93,631
Deaths - 22,154
Again, we will assume some things. One, these numbers are accurate (something we'll discuss later). And two, that symptomatic cases are roughly 2X's the confirmed cases. Each of these assumptions will be vetted later.
Population estimate 2020 - 318 million
0.35% are already symptomatic [(assumed)] at three months in.
0.18% have already been confirmed.
0.029% have already been hospitalized.
[These numbers are for only three months.] Simple extrapolation to a full year means we are looking at [0.12%], or [374,524] people being hospitalized.
So far, it looks like H1N1. Maybe even a little better, right? Not so fast.
Of the confirmed cases, the mortality rate is 3.97%.
Remember, we are assuming symptomatic cases is 2X's the confirmed. So the mortality rate of all symptomatic cases would be 1.98%.
Recap:
Seasonal flu symptomatic mortality rate: 0.13%
H1N1 symptomatic mortality rate: 0.21%
Covid19 symptomatic mortality rate: 1.98%
Do you see the difference yet?
*Elephant in the room*
The numbers.
Here we'll address assumption #1; the accuracy of the numbers.
Many have been saying the numbers are inflated. And they may well be. Some of the reasons for this belief include: inaccurate reporting, intentional skewing, and outright lying.
Inaccurate reporting:
Yes. This is a dynamic situation. We may never know 100% of all the data. Even the seasonal flu data is subject to this. The H1N1 data shows that much of the earliest numbers were full of errors and assumptions. Testing for a novel virus isn't error free.
As with any statistical analysis, you have to bake inaccuracies into the cake. You have to try to keep all of your assumptions the same across data sets. So, we assume there are inaccuracies in the Covid19 data, but we hope similar inaccuracies exist in the H1N1 and flu data.
And we know they both do have similar inaccuracies. Not everyone who gets the seasonal flu reports it. Even of those who see a doctor for it, few get tested. They are often diagnosed without confirmation.
Likewise, there are reports of diagnosis of Covid19 without testing.
Intentional skewing:
There is money to be made from inflating Covid19 numbers. But again, this is baked into the cake. However, the intentional skewing may be higher in the US and other countries because of the governments throwing money at those "confirmed" cases.
So, let's compare to other countries. We know China has bad data, so we'll throw them out from the start. So let's use several others:
Italy (worst case)
UK (similar to US)
Turkey (possible control group)
India
Malaysia
Japan
Singapore (best case)
We need to compare three things:
Confirmed cases
Confirmed rate (percent of population)
Confirmed mortality rate
Italy: 156363 - 0.26% - 12.7%
UK: 89554 - 0.13% - 12.7%
Turkey: 61049 - 0.073% - 1.96%
India: 9635 - 0.0007% (!) - 3.43%
Malaysia: 4817 - 0.015% - 1.60%
Japan: 7370 - 0.006% - 1.67%
Singapore: 2532 - 0.044% - 0.31%
Analysis:
The confirmed rate as a percent of the population is ALL OVER THE PLACE. In Italy and the UK, you see 'confirmed' cases at rates similar to the US (~0.18%). In all other countries, the rates are much lower. So this appears to confirm that the reported cases are high.
But how high? Can we tell from the other numbers?
First, we have to assume some other factors. One: Turkey, India, and Malaysia likely have terrible reporting and tracking. Their numbers are suspect. If you look at India, their reported cases are staggeringly low.
Another anomaly with India is their death rate is very high for having so few reported. I would bet that all of India's numbers are bad. So we will leave India aside for now.
What about Italy and UK? Their reported cases rate is similar to the US, but mortality is HIGH!
The death rates in Italy and UK are a factor of 10 higher than the US. It is VERY likely that there is something wrong with their numbers. We might assume they would be higher due to some outside factors, but not 10X's more.
What about Japan and Singapore?
Japan and Singapore already have a high level of social distancing and widespread mask use. It would stand to reason that their numbers would be low. And their reported rate is very low. Their reported rate is a factor of 10 lower than the US and the UK.
But their death rates are not so dissimilar to the US, Turkey, and even Malaysia. Singapore has a very low death rate; possibly the lowest in the world. Is there misreporting on the other side? Perhaps Singapore's semi-socialist governance wants to inflate their healthcare?
However, even if we take Singapore's unusually low death rate, it is still higher than the 2009 H1N1 death rate.
More likely, the correct death rate is something along the rate of Japan (which is similar enough to Turkey and Malaysia to give confidence).
[Outright lying:]
[It's absolutely possible that countries are lying about their results. China is a prime example. But the question becomes, who is lying and which direction are thy skewing the numbers? The US, UK, and Italy may be lying to make their numbers higher. Japan and Singapore may be lying to make their numbers lower. But again, this has to be backed into the cake.]
What does that mean?
A mortality rate of 1.6% [(Malaysia / Japan)] is 7.5X's the rate of the H1N1.
And it's 12X's the rate of ALL seasonal flu viruses, combined (remember, there are 131 varieties)!
As for the assumption that symptomatic cases are roughly 2X's the number of confirmed cases, we can make the assumption in any direction, and the outcome will be affected very little. Once we have determined the mortality rate of the virus, very little else matters.
For instance, let's assume the number of symptomatic cases are 5X's (instead of 2X's) the confirmed cases, and the mortality rate is that of Japan (1.67%):
US's assumed symptomatic cases: 2,792,995
Expected deaths: 46,643
That's 2X's the known deaths (which we have already assumed is inflated). If the actual deaths are lower than the reported, then the US mortality rate would be lower than Japan. Does that make sense?
Let's assume it does (which it doesn't): Then what?
If the US assumed symptomatic cases are really that high, and the actual deaths are lower (let's say by half):
11,077 deaths / 2,792,995 cases = 0.04% mortality rate
That is STILL DOUBLE the rate of the H1N1!!!
Recap:
We have QUINTUPLED the symptomatic cases, and we have halved the REPORTED deaths.
And Covid19 is still twice as deadly as anything we've dealt with in a century.
So, even if we play the game of "The numbers are wrong", we still come out with a pandemic on the scale of the 1918 Spanish Flu.
But if we take a realistic look at the numbers being reported around the world, we are looking at a 1.5% - 3% mortality rate.
If we don't act like Covid19 is a big deal (which it is), and we behave like we did under the H1N1 or Ebola pandemics, we will end up with 80 million infected in the US alone. And with a mortality rate of a conservative 1.5%, that's 1,200,000 dead in a year.