Hacker News Reader

Comment by anonuser123456 | Hacker News Reader

The best model I have seen so far is:

Take deaths, multiply by 400 and that's how many cases you have. This comes from the assumption of 1% mortality and 3 weeks to die.

So CA has ~8000. A doubling time is ~5 days. 8 weeks ~= 11 doublings. So 8000 * 2048 = 16M. Not too far off for a ballpark. Obviously, the exponential starts to breakdown somewhere... but where?

I don't think it's fear mongering. It's getting the public to understand the back of the envelope worst case... which is really bad. People NEED to panic... to get them to hunker down and take this seriously. The sooner than happens, the sooner we get it under control.

I had to yell at my inlaws to stop taking the train to work and going to the gym. They are 70 and have pre-existing conditions. This action will force them to take it seriously.

View on HN

Take deaths, multiply by 400 and that's how many cases you have. This comes from the assumption of 1% mortality and 3 weeks to die.

I don't see where 400 comes from.

If we assume that people who died today now are 1% of those who got it 20 days ago, then that is a factor of 100. Project that forward through 4 doublings for the fact that we have about 20 days of doubling every 5 days and we get to a factor of 1600. You can complicate the toy model in a number of ways, but I can't get it as low as 400.

Anyways it is worth pointing out that the curve of reported cases will continue rising for a couple of weeks as people both get symptoms then get sick enough to come to the attention of hospitals. And if the measures work, it won't be obvious from death rates for about 4 weeks. And at that point the critical question is whether we managed to slow the rise, or create an exponential decline.

My bet is that we are only going to get a slow in the rise. Hopefully enough for us to get ahead of the need for emergency treatments.

“ My bet is that we are only going to get a slow in the rise. Hopefully enough for us to get ahead of the need for emergency treatments.”

Look at the curves for every where else on the planet, the US will be the same, take it seriously

> Look at the curves for every where else on the planet, the US will be the same, take it seriously

Except this is extrapolating ahead to assume no changes to treatment available. As time progresses, we find better, more effective ways to prevent death from COVID-19.

I think doubling time is a bit faster than 5 days from what I have seen elsewhere, and that might change the math to the 25.5M in that time frame.

It seems like it’s difficult to get an accurate sense of the doubling rate right now, because of the confounding variable of increased testing.

You don’t have to get a sense of doubling via unreliable test data if you look at the doubling rate of the fatalities.

People count bodies much more reliably in pandemics. I don’t trust our confirmed case telemetry data much, but I do trust the fatality data. Unfortunately it’s about a two to four week lagging indicator. But the fatality rates also currently show this to be spreading much, much faster than it ever spread in Wuhan.

Wuhan is what happens when the curve is flattened. Not what happens when you don’t.

You still don't get accurate data if you just count bodies.

The issue here is every death is attributed to COVID-19 if the virus is present even if the actual reason is some other underlying health condition. Given that most victims are elderly and had serious pre-existing conditions this leads to massive mortality rate over-estimation.

You also cannot reliably compare EU/US data with Chinese: initially most deaths and cases of pneumonia in China were attributed to other illnesses. Secondly, Chinese government is notorious for flat-out lying in their official stats.

> The issue here is every death is attributed to COVID-19 if the virus is present even if the actual reason is some other underlying health condition.

That what you claim provably doesn't happen. Even old people simply don't die that fast otherwise, and we also have the data of the people who can't breath and have to be admitted to the hospitals: there were never that many cases before, happening that fast:

https://en.wikipedia.org/wiki/File:Is_COVID-19_like_a_flu%3F...

(the numbers were compared by the Italian National Biotechnology Association, each bar is one week of time)

It feels like testing is a lagging indicator at this point. Whatever doubling happened, already happened.

Right, in some countries doubling was every 3 days.

That was the case for Italy until 4-5 days ago and I think it still is for Spain, definitely for Germany.

Have a look at this site: https://www.worldometers.info/coronavirus/country/us/

Scroll down to "Daily New Cases in the United States". No real need to extrapolate data from other countries. New cases are doubling every 2-3 days.

Those aren't cases, those are confirmed cases. Which has more to do with how much testing we have done than how many are sick.

Problem with looking at case numbers is, you need to know how many tests have been executed. If they doubled the number of tests within those 3 days, then you measure the increase in tests, not an increase in cases.

Yeah the doubling time confuses me too. All the numbers from China, Italy and the Netherlands (where I am at) show the number of deaths doubling every 2 days upto the limit.

But even our dutch government is claiming it's 4 days standing in front of a chart where it's obviously 2 days.

After you reach the limit it's hard to count because people stop going to the hospital that can't take them in anyway. At that point you should just consider any increase in deaths compared to the normal the result of Corona. Wether it's a cancer treatment that got postponed because of Corona or Corona itself.

>Take deaths, multiply by 400 and that's how many cases you have. This comes from the assumption of 1% mortality and 3 weeks to die.

So that 400 multiplier gives the number of cases as of 3 weeks prior to the death count used in the calculation?

yeah it would have to, right?

The people didn't all die on the same day though...

Worldometers shows CA with 1060 cases and 19 deaths. Where does ~8000 come from. If 8 weeks is 56 days and cases go up 20% a day as they have been then 1.2^56 * 1060 = 28.8 million cases.

It is from my understanding that 19 * 400 ~= 8000. The OP suggested that for every 1 reported death there are 400 cases.

That's not 1060 cases, it's 1060 tested positive. You can't even get a test unless you're a celebrity or high risk... if you're a nurse and your wife is positive, they won't give you a test. We're under tested so there are a lot more than 1000 cases right now.

It's true though that they aren't using the highest rate they could (the 20% per day we got before social distancing). I hope that closing the schools helps... and the ICUs don't overflow immediately.

I do agree that people have a bad tendency to trust any numbers they can get over logic and common sense. And in this case the numbers are obviously a lower bound creating a big bias towards underestimating things.

But information on testing numbers is also lagging and incomplete. The CDC was doing over 5k tests per day a week ago when the number of new cases was around 500 per day. And the number of tests were exponentially increasing too. Does anyone have any better info about how much we have ramped up testing since then? It's a big country, and there's huge "publication bias" in the media where stories about a location or person running into testing issues get the attention and stories where things are working do not. I really don't trust anyone's ability to assess the situation at the resolution of 24 hours and make anything like an accurate prediction. I would like to see them just do full-on statistical sampling by testing people at random in different cities.

If I recall, China had about 500 cases confirmed positive, when they shut down Wuhan on January 23. And by that point, the confirmed infection count ballooned up to 80,000; even after they aggressively mandated their national lockdown. And this was during the early days, when nobody knew anything about the virus. And early test kits gave a lot of false negatives.

With California at 1060, then I can only imagine where the final number will land. And New York has 5712 infected, and yet, they still haven’t shut down the state.

====

Now, the following is likely incorrect, but this may give an general idea of the magnitude. But doing a basic extrapolation, China had a 160x multiplier, from the initial lockdown point, to get to 80,000 infected.

If we use that, then at this point, California might register about 169,000 infected. And New York, if they lock down now, might register 914,000 infected.

Sobering numbers indeed..

But these numbers are highly dependent on population density, mass transportation patterns, and social congregation scenarios. Wuhan has high speed trains, but California has Southwest planes. Car travel and ride sharing usage might be similar. So it’s unlikely the numbers are similar to compute.

At some point, the medical system will fail, and everyone will just give up, and we are now left to our own fate.

Remember that scene from the Chernobyl series where they messure the radiaton and say it is only 3.6 roentgen — which oddly happens to be around the maximum their device can messure?

This is the situation with tested cases now. You can assume that there is no nation on earthq where the number of tested cases is an accurate representation of what is really going on. Deaths are far more accurate but lag behind ehat is going on by two weeks. And two weeks is a long time with something that grows exponentially.

The bad thing about this is, that this logic is flipped on it's head at times: my girlfriend had all the symptoms just like 6 of her alumni — not a single of them could get tested. And while they where still in bed their university proudly issued a press statement stating they have zero cases of Corona. Yeah — if you are not testing you will have zero confirmed cases, that is the way it is.

The dangerous thing here isn't the current number of cases, but the insecurity about it. If you have a huge number of people starting to get ill today your hospitals will fall apart in two weeks time. Doctors will have to send people home to die, like in Italy visitora can no longer be handled, people no lomger be burried — this is not a joke.

I wanted to clarify something. The parent poster wanted to know how the 25 million case number was arrived at. I just provided an example of how such a number might be arrived at. It is far too simplistic of a model though. I have also read there is confusion over the term “case”, as to whether it’s a hospital case vs total cases in general population.

Reply on news.ycombinator.com