HomeLatest ThreadsGreatest ThreadsForums & GroupsMy SubscriptionsMy Posts
DU Home » Latest Threads » Forums & Groups » Main » General Discussion (Forum) » Big numbers

Fri Apr 3, 2020, 01:01 PM

Big numbers

I've been trying various ways to figure out, when all is said and done, how many people world-wide are likely to die from this COVID scourge. I've been hampered by the fact that statistics was my worst subject in university, even though I'm endlessly fascinated by the behaviour of datasets. That makes me just knowledgeable enough to be dangerous. I accept that my conclusions are endlessly debatable, but right now all such results are debatable, so I don't feel too bad.

Based on the global statistics, it looks to me like the eventual Case Fatality Rate for identified cases and deaths will be around 6%. I derived this from the published global case and death data going back to mid February. I calculated ten sets of ratios, one based on the each day's deaths divided by the same day's cases, all taken from worldometer. Then, based on the idea that one day's deaths may be better correlated with some previous day's cases (because it take a while for identified cases to die), I created 9 more series by dividing each day's deaths by the cases on each of the nine previous days.

I graphed all ten series of CFR percentages, and got a set of nested curves, all of which rose and then turned back down. I put a well-fitting trend line on each curve, extended it out about 2 weeks, and found something very interesting. The curves all crossed at the 6% point, except the one using same-day data. I'm not enough of a statistician to tell you why that intersection might have happened, but my pattern-recognition circuits all lit up when I saw it. The graph is below.

Ass a sort of cross-check, I extrapolated the global curves for identified cases and deaths out 14 days, and used the end-points of those two projections to calculate the projected CFR six weeks from now. Again, the answer was about 6%. (graphs below)

So, absent any more reliable estimate at this point in the pandemic, I've adopted 6% as my tentative CFR. I'll watch the behaviour of the data as time goes on, to see of the estimate changes.

Applying this to the future state of the world, assuming an identified infection rate of 10%, a CFR of 6% implies about 48 million deaths worldwide.

As far as the USA is concerned, a 10% infection rate and a 6% CFR projects 2 million deaths, which is within the range of the high-level projections revealed by Dr. Birx.

Obviously these are little more than wet-finger estimates because it's still early days. They will obviously change as time goes on and the results of social distancing and lockdowns are felt (or not) in different places. But it does hint at the death tolls we should be prepared for.

This so sucks.






7 replies, 801 views

Reply to this thread

Back to top Alert abuse

Always highlight: 10 newest replies | Replies posted after I mark a forum
Replies to this discussion thread

Response to The_jackalope (Original post)

Fri Apr 3, 2020, 02:28 PM

1. I would say you are very good at this. I actually can understand it and that is saying something.

Reply to this post

Back to top Alert abuse Link here Permalink


Response to monmouth4 (Reply #1)

Fri Apr 3, 2020, 02:46 PM

4. Thanks! nt

Reply to this post

Back to top Alert abuse Link here Permalink


Response to The_jackalope (Original post)

Fri Apr 3, 2020, 02:33 PM

2. This study in Lancet has the CFR at 1.38 percent. And I would bet that goes down with more data.

Reply to this post

Back to top Alert abuse Link here Permalink


Response to blitzen (Reply #2)

Fri Apr 3, 2020, 02:45 PM

3. I have no doubt the Lancet has way better statisticians than me.

I just don't trust low aggregate numbers for this disease, even when they're announced by people wearing lab coats. Not given the death rates we're already seeing, and the fact that it's only 3 months and a bit since the very first cases. Low numbers sound like "don't spook the horses" PR to me at this point.

I'll stick with my own numbers for now, until I see a reason not to.

Reply to this post

Back to top Alert abuse Link here Permalink


Response to The_jackalope (Reply #3)

Fri Apr 3, 2020, 03:02 PM

6. Yes, I think the official case numbers are way low (lack of testing, etc.), and so the CFR will...

ultimately be lower (maybe a lot lower). The site that I look at a lot (link below) has the CFR at 6%. I would imagine that the deaths number is somewhere in the ballpark, but the total cases number is probably way too low.

What scares me about the Lancet data is the fatality rates for the older age groups: 4% for age 60-64, 8.6% for 70-74, 13.4% for 80+

https://ncov2019.live/data

Reply to this post

Back to top Alert abuse Link here Permalink


Response to blitzen (Reply #6)

Fri Apr 3, 2020, 03:32 PM

7. The site I use has the CFR for closed cases at 20%

https://www.worldometers.info/coronavirus/

But the number of closed (recovered + died) cases is only 25% of the total number of cases. If the same ratio were to apply to the active cases that has shown up in the closed cases, the CFR for the disease would be 20%.

That of course would be driven down by more testing, but could also be raised somewhat by counting deaths that occur outside hospitals that may not currently be recorded.

Being elderly, the CFR numbers for the elderly make we want to stay indoors as much as humanly possible. Lucky I'm a misanthropic introvert!

Reply to this post

Back to top Alert abuse Link here Permalink


Response to blitzen (Reply #2)

Fri Apr 3, 2020, 02:47 PM

5. I disagree as a high number of people died of CFR and were not counted IMHO. We know this happened

in Italy for sure and also in China where the numbers are suspect...I think it will be between 3% and 6 %...could be higher if our hospitals are overwhelmed and treatment is poor. There will also be deaths of people who because of CFR could not receive treatment of other life threatening illness in time because of the hospitals being unable to treat them.

Reply to this post

Back to top Alert abuse Link here Permalink

Reply to this thread