This is a problem of insufficient data. We don't have a denominator, so there's no way to tell if it's more widespread than we thought. In fact there's no good way to tell how widespread it is at all. We have numerators: number of people sick enough to show to up to an ER, or number dead, or whatever, take your pick. But in every fraction we might want, we lack a denominator: total number of people infected.
There's a stuningly simple way to get one: sample randomly from a population of interest in large enough numbers (say e.g. Cook County, home of Chicago, or hell, the entire US) and test everyone for the virus. Voila! You have a representative chunk of your population, an estimate of virus prevalence, and a quantifiable degree of precision about that estimate (SE's for a confidence interval, say). With enough people in your sample, you can have extremely accurate estimates of the spread of Covid in your population of interest. You repeat this process over and over, through time, to track the numbers.
I'm not trivializing this kind of effort: it takes rigorous sampling designs and dozens or hundreds of field workers, among other things. It's intense but very straightforward. Political pollsters (and thousands of researchers in different fields) do it every day.
No one in the US has done this, and no one with any visibility from Fauci on down has even suggested it (that I know of; please correct me). I work in public health, and this is first year, first semester of grad school stuff.
This matters because the public health responses to coronavirus when there's 1% or 10% or 25% or 75% of a population infected all look very different. In short, we may be over- or underreacting to the situation with the measures we currently have in place.
I've been going half crazy wondering what's happening about this for a while now. After the initial fuckup, we now have tests. We have money. We can do this. Why don't we?
> I'm not trivializing this kind of effort: it takes rigorous sampling designs and dozens or hundreds of field workers, among other things. It's intense but very straightforward. Political pollsters (and thousands of researchers in different fields) do it every day.
I am trivializing it, when compared to the cost of stopping down there economy as we've done.
If there a chance the results can restart the economy a bit earlier, it almost doesn't matter how much it will cost.
A Stanford MD PhD who had a recent article in the Wall St Journal was interviewed on video[1]. In the video he refers to this very denominator-related test being forthcoming. There are two things I don't understand. First, why did this have to wait for some approved(?) antibody test, rather than being tried with a presumably research-grade test to start?? Second (physicist talking here), to get a rough estimate (to within a factor) whether exposure has already silently _significantly_ spread, why not relax some rigor in the sampling design and just sample 100-1000 geographically-diverse people to get _some_ estimate that could at least speak to the "is exposure already widespread?" question?
My understanding is this hasn't been done yet because we still don't have enough tests for all the people who have symptoms or might have the disease due to circumstances. But the number of available tests is going up very rapidly, so maybe pretty soon we will be able to do that sort of study.
Of course, our stable genius president has been insisting for months that there is no shortage of tests.
By the way, another way to go at this question is to test for antibodies, and that is also being developed very rapidly.
% of visits to primary care provider is a _terrible_ metric. People have been actively advised to stay away from the Doctor’s office and hospitals for routine reasons (and reschedule for later). Obviously the % visiting because they think they have Covid is going up, even if they don’t have it!
I dont understand how would such highly contagious virus suppress and contain itself if it started in CA months ago. As if Californians have no outbound travel. Impossible scenario.
I don’t think anyone is claiming that it suppressed and contained itself. They’re claiming it’s more widespread and more mild. There’s a good graphic in the economist article showing two different peaks that look the same early on.
I was pretty sick in late February, and ended up in the urgent care at my local Kaiser. It was packed at the time too. Seemed like everyone was sick with the flu.
I was terribly sick in early February, in CA. I went to my urgent care multiple times as I got worse.
The doctor tested me for influenza, but told me that he was seeing hundreds of patients, all day every day, with "an unusual flu" where they were mildly sick for a few days, recover, and then suddenly got much worse with high fevers and chest infections -- and most of those patients were turning up with negative influenza tests. But influenza tests have a high false negative rate, so just because I was negative on the test didn't mean anything. The doc said that this year, the influenza test only seemed to show positive for folks that had a high fever the first day of their illness (which, between you and me, is very common with influenza and very uncommon for COVID)
I was out of work for over a week, and only got out of bed when I had no choice the entire time.
None of my doctors asked me about travel or contact with recent travelers, and nobody ever even suggested getting a coronavirus test.
When I got back to work a week later, I learned that nearly 10% of my office had gotten sick either the same day as me or while I was out.
Did we all have COVID-19?
maybe. Maybe not.
None of us were tested. None of us will turn up in the COVID statistics.
But in early/mid February, there definitely was some coronavirus here -- we had a few known travel cases, and we had tons of travelers between all regions of China and all parts of CA. It is entirely possible for folks to have unknowingly brought the virus here in moderate numbers.
I've also read that one of the reasons that California has relatively low infection numbers given it's circumstances, may be do to the fact that Covid-19 had been spreading earlier in the year. Many people may have been unknowingly infected, and either have recovered and have immunity, or unfortunately may have died from it.
Although, it seems apparent that there were not millions of cases before we started looking, given that hospitals here were not overrun like they were in New York.
About a month ago, I did some poking around with the CDC FluView statistics (in-season estimates of influenza and all-cause-pneumonia deaths)
The conclusion that I came to was an upper bound on undetected COVID cases before 3/1 of 300k.
More than that, and the hospitalization and death rates could not have hidden in the flu and pneumonia data.
300k extra recovered cases, even if they were all in CA, would not be material in terms of new immunity.
Unless the Asia-origin strain we had was much less dangerous than the European-origin strain in NY (in which case more cases might hide in the data) -- but that is a pretty tall set of assumptions.
My take the rate of serious illness and death due to COVID19 is a bounded parameter with a linear range. The infection rate is currently unbounded and logarithmic.
Ain't got much wiggle room to work with there. You have to pretend the morbidity rate is 20 to 50 times lower than the known lower bound (3-5%). So doesn't square.
> I was pretty sick in late February, and ended up in the urgent care at my local Kaiser. It was packed at the time too. Seemed like everyone was sick with the flu.
The 2019-2020 seasonal flu has also been relatively bad compared to many other years.
Was this data not publicly available ? Were there not army of analysts? I thought with Machine Learning being all the rage and hype, we’d have more people and machine keeping tabs on all kinds of trends, this included. Good that we have these retroactive inspections, but the whole appeal of ML is its predictive power is it not? Or did the hedge funds just keep this knowledge to themselves?
Could it be because AI is an overhyped rehash of the AI craze that swept academia and industry in the 1980s, only to collapse in the 'AI winter' as everyone realised the truth? We use AI/ML for trivial purposes right now. Multivariate statistical analysis beats AI in many domains. Covid is a real-world, grown-up problem. It isn't identifying pictures of cats, or creating obscene chatbots (remember Tay?). So AI has been shown up to be a duff non-solution.
> "I think the 3.4 percent is really a false number — and this is just my hunch — but based on a lot of conversations with a lot of people that do this, because a lot of people will have this and it's very mild, they'll get better very rapidly. They don't even see a doctor. They don't even call a doctor. You never hear about those people,..."
At the time, the person who said that was called an idiot.
He's still an idiot. He chose to believe that not because of research or statistics or the advice of people with expertise, but because it was convenient. This was at the beginning of March when there had only been 11 deaths. He was making the argument that this is no worse than the normal seasonal flu, because he was loathe to take any drastic preventative action that might disrupt his precious stock market.
Well, here we are now, thousands have died, with thousands more on the way, and the economic damage has already been done. That's what happens when you let an idiot run the country.
I'm loathe to defend the current US president, but almost no Western country took this seriously until the last possible minute. There is clearly a deeper ideological issue infecting the entire West.
It's absolutely true that most western countries underestimated this. East-Asian countries were better prepared and more willing to take drastic action to do what's necessary. Western countries assumed that, like SARS, Asia would solve this. Turns out it's way more contagious than SARS.
Still, once the seriousness became clear, most countries did take action. Trump continued to deny and spread misinformation, even to this day.
Who do we believe a bullshit artist at the economist or Bob Wachter Chair, UCSF Dept of Medicine.
> @Bob_Wachter
> 3/ Another interesting #: @ucsf, we only test pts w/ symptoms, who all think "I have Covid.” But only ~4% of our tests are +, meaning most folks who think they have it, don’t.
Also, if you don’t trust the WSJ either (I wouldn’t blame you) here is an article from Nature Medicine that has a comparison of nasal vs stool tests for covid, and Figure 1b really drives home the high rate of false negatives for nasal swab tests compared to stool tests (and throat swabs are even worse than nasal swabs): https://www.nature.com/articles/s41591-020-0817-4?fbclid=IwA...
Yep. In fact, false negatives of current test are a function of time since the onset of symptoms:
Day 1? ~7% false negative
Day 10? 40% false negative
Day 20? 90% false negative
That's really high, and it seems anecdotally likely that a majority of tests have been administered to people who are 7-10 days or more post-symptom-onset.
30 days ago many HN posters were saying this. I guess they were all idiots too.
They don't say it much any more, though... why get downvoted? The fact that this has already spread through the US will become self-evident soon enough. ("The peak is coming in CA any day now! Just wait...")
That article is comparing NY to CA with the expectation that they should be more similar.
CA is a very car-centric state, the public transportation situation is a joke. I'd expect this alone to tamp down the spread relative to NY, considering the density of NYC and how ubiquitous public transportation is there.
>He does not advocate lifting social distancing rules
I believe the "idiots" referred to above are the people who blindly assumed without evidence the experimental hypothesis of this study simply as an excuse to avoid social distancing and self isolation.
There's a stuningly simple way to get one: sample randomly from a population of interest in large enough numbers (say e.g. Cook County, home of Chicago, or hell, the entire US) and test everyone for the virus. Voila! You have a representative chunk of your population, an estimate of virus prevalence, and a quantifiable degree of precision about that estimate (SE's for a confidence interval, say). With enough people in your sample, you can have extremely accurate estimates of the spread of Covid in your population of interest. You repeat this process over and over, through time, to track the numbers.
I'm not trivializing this kind of effort: it takes rigorous sampling designs and dozens or hundreds of field workers, among other things. It's intense but very straightforward. Political pollsters (and thousands of researchers in different fields) do it every day.
No one in the US has done this, and no one with any visibility from Fauci on down has even suggested it (that I know of; please correct me). I work in public health, and this is first year, first semester of grad school stuff.
This matters because the public health responses to coronavirus when there's 1% or 10% or 25% or 75% of a population infected all look very different. In short, we may be over- or underreacting to the situation with the measures we currently have in place.
I've been going half crazy wondering what's happening about this for a while now. After the initial fuckup, we now have tests. We have money. We can do this. Why don't we?
(Iceland has done this!)