What can we learn from super-wide uncertainty intervals?

This question comes up a lot, in one form or another. Here’s a topical version, from Luigi Leone:

I am writing after three weeks of lockdown.

I would like to put to your attention this Imperial College report (issued on monday, I believe).

The report estimates 9.8% of the Italian population (thus, 6 mil) and 15% of the Spanish population (thus about 7 mil people) as already infected. Their estimation is based on Bayesian models of which I do not know a thing, while you know a lot. Hence, I cannot judge. But on a practical note, I was impressed by the credibility intervals: for Italy between 1.9 mil and 15.2 mil, and for Spain between 1.7 mil and 19 mil! What could a normal person do of these estimates that imply opposite conclusions (for instance for the mortality rate, which could oscillate between the Spanish flu at one end and the regular flu at the other end of the interval)? It is also strange for me, that the wider credibility intervals are found for the countries with more data (tests, positives, deaths), not for those with less data.

My reply: When you get this sort of wide interval, the appropriate response is to call for more data. The wide intervals are helpful in telling you that more information will be needed if you want to make an informed decision.

As noted above, this comes up all the time. When we say to accept uncertainty and embrace variation, the point is not that uncertainty (or certainty) is a good in itself but rather guide our actions. Certainty, or the approximation of certainty, can help in our understanding. Uncertainty can inform our decision making.

1. Brent Hutto says:

Sometimes a perfectly valid and useful statistical result is a finding of the form, “We can not tell from the available data how many people will die from COVD-19”. As Andrew says, it implies that more (and better!) data is required from making any firm conclusions.

I think the current state of the data is firmly within that realm at the moment. We can only hope that a) those in authority begin gathering the necessary data immediately and b) not too much damage is done by decisions made based on conclusions arising from incomplete and erroneous data.

• Anoneuoid says:

Half the important info we need is just local testing criteria which is determined by the local authorities.

Like heres someone saying the NY dept of health changed the policy to only testing in-patients “last friday” (the 20th I think). Since then the % positive has been rising in that state while tests/day have levelled out. Explaining that rise is key to epidemiological models.

2. Ben Goodrich says:

Perhaps more noteworthy is that the Imperial College crew now seems to be using Stan to produce these estimates

https://github.com/ImperialCollegeLondon/covid19model/blob/v1.0/stan-models/base.stan

• Rahul says:

One of the silver linings post Covid is going to be the huge rise in status and applications of Bayesian modelling a la Stan.

3. Rahul says:

Andrew, more data is fine. But you didn’t address one specific part of the question which is interesting too:

“that the wider credibility intervals are found for the countries with more data (tests, positives, deaths), not for those with less data”

Any thoughts on this? Why would this be so?

• Andrew says:

Rahul:

Unless I’m missing something in my quick scan of these graphs, my impression is that there’s more uncertainty on the absolute scale when numbers are higher. On the log scale, I’d expect more uncertainty for the cases where numbers are lower.

• Zhou Fang says:

Haven’t look deeply into the models but generally speaking that’s a consequence of a log (or similar) parameterisation being used. Exp(A +/- B) produces wider intervals when back-converted into the original scale.

More theoretically if you have something like a Poisson distribution you’d see a relationship between variability and mean.

• Carlos Ungil says:

As others have suggested, it seems less surprising when you look at the results as “number of people infected for each death observed so far”.

In aggregate, they estimate 400 to 2400. I have not calculated them all, but for the countries where they give absolute numbers (and rounding a bit to make our host happy): Italy 200 to 1700; Spain 400 to 4000; Germany 700 to 4600.

Keep also in mind that the model is quite simplistic and doesn’t even consider the population of each country even though some of those countries are ten times as populated as others. The regional distribution of deaths (and interventions) is ignored as well.

Extrapolating naively from the multiples in the previous paragraph, in Lombardy (60% of deaths in Italy) between 10% and 90% of the population would be infected while in Madrid (50% of deaths in Spain) between 15% and 150% of the population would be infected!

4. jim says:

Nothing against Stan or Fred or whatever tools have been deployed, but as they (used to) say (a lot more often), right now it’s all academic. No one has produced a useful model and more data isn’t helpful if the data is bad, which it all probably is.

WA state is claiming 246 total fatalities. Over 2-3 weeks that is surprisingly slow growth rate. But OTOH every level of government is gearing up all over the metro area, leasing out entire downtown hotels and setting up numerous football-field-sized temporary hospitals.

I conclude that WA’s data aren’t accurate, and probably the same for everywhere else.

so right now, saying “we need more data” really means “we need an approach that can work with the shitty data we have”. And that means zooming out and making qualitative estimates and deploying those as best they can be deployed. I just don’t think modelling is doing anything useful.

• Brent Hutto says:

Speaking of “shitty data”…

There’s a big furor at the moment in my state (South Carolina) over the fact that COVD-19 infections among health care workers are not being tracked. The local health department says that is not within their tracking obligation.

Apparently, in most cases even the hospitals where the COVD-19 patients are being treated are only testing their health care workers when they meet the same criteria (symptoms) as the general public. No routine testing of even the doctors, nurses and techs who are literally hands-on with the sickest patients (in some cases without sufficient supplies of personal protective gear).

What the heck is up with our entire public health and health care system’s aversion to screening people for this virus?

• Rahul says:

Isn’t it just about the resources? Do we have enough testing kits, swabs etc.? Even when we have them do we have them at the right spot. Do we have the people to deliver them to the labs. The IT to track it?

I think the answers to a lot of it lie in how quickly can we change and set up new systems.

In principle there’s no problem with these proposals. The details are the issue. Who sets this up rapidly and how. Who are the people we have that are lying idle and yet have the talents to do this.

• jim says:

Rahul:

Yes, IMO people have lately been fooled about how quickly things can be set up and changed by computer technology. It’s really easy and fast to assemble and deploy digital systems. It’s a lot harder to assemble and deploy physical assets.

And with respect medical equipment, people are dreamin’ if they think this stuff is going to magically appear. It’s going to take months to get ventilator production rolling. Ford and GM are manufacturing experts but the FDA and medical equipment is a whole new ballgame for them.

• Rahul says:

I’m not very sure whether the FDA is right now a part of the solution or a big part of the problem. What may make sense is to temporarily castrate the FDA and give legal green lights to the manufacturers to go ahead based on internal risk assessments maybe supplemented by an ad hoc panel of external academics or experts.

Some executive orders are needed going ahead….

• Anoneuoid says:

This ER doctor says the ventilators are killing the patients because everyone assumes they should be treating ARDS but the patients have something more similar to high altitude illness:

• Anoneuoid says:

These guys say the same thing. The intubation is making things worse:
https://podcasts.apple.com/us/podcast/emcrit-podcast-critical-care-and-resuscitation/id314020330?i=1000469992639

• Navigator says:

I have heard (anecdotally) that in this situation where medical personnel are at risk of infection, they prefer to intubate because patients cannot talk, cough or otherwise spread the virus further.

• Anoneuoid says:

I have heard (anecdotally) that in this situation where medical personnel are at risk of infection, they prefer to intubate because patients cannot talk, cough or otherwise spread the virus further.

I think the vast majority of people working at the hospitals are good people who mean for the best, but there is something seriously wrong with those coming up with guidelines for what gets taught, published and allowed/recommended as a treatment…

I’ve been saying this for years and left medical research because of it.

• Anoneuoid says:

Yep, all the symptoms are the same as HAPE:
https://www.ncbi.nlm.nih.gov/pmc/articles/PMC7096066/

The ARDS was being caused by the ventilators…

• Martha (Smith) says:

In case anyone else clicking the link wonders: “Ground-glass opacity (GGO) is a radiological finding in computed tomography (CT) consisting of a hazy opacity that does not obscure the underlying bronchial structures or pulmonary vessels”

• Anonymous says:

https://us19.campaign-archive.com/?u=ef98149bee3f299584374540a&id=48d2c0484f

It has been alleged that COVID-19 associated respiratory failure is the result of a novel pathophysiology and is therefore not ARDS.

Such sentiments are frequently based on a misunderstanding of the definition of ARDS, its spectrum of disease, optimal treatment, and the physiology of hypoxemia.

Almost all commentators, even those who argue it is not ARDS, appear to agree that large distending pressures in the inflamed lung can be injurious and lead to worsening of gas exchange and pulmonary mechanics. This is the essential pathophysiology of ARDS.

COVID-19 associated respiratory failure is characterized clinically by hypoxemia and bilateral infiltrates, pathologically by diffuse alveolar damage, and is optimally treated by a strategy that minimizes distending alveolar pressure and lung stretch. On this basis, COVID-19 associated respiratory failure is ARDS.

• Anoneuoid says:

It has been alleged that COVID-19 associated respiratory failure is the result of a novel pathophysiology and is therefore not ARDS.

Such sentiments are frequently based on a misunderstanding of the definition of ARDS, its spectrum of disease, optimal treatment, and the physiology of hypoxemia.

Almost all commentators, even those who argue it is not ARDS, appear to agree that large distending pressures in the inflamed lung can be injurious and lead to worsening of gas exchange and pulmonary mechanics. This is the essential pathophysiology of ARDS.

COVID-19 associated respiratory failure is characterized clinically by hypoxemia and bilateral infiltrates, pathologically by diffuse alveolar damage, and is optimally treated by a strategy that minimizes distending alveolar pressure and lung stretch. On this basis, COVID-19 associated respiratory failure is ARDS.

The problem is that what academics are saying isn’t agreeing with what the doctors on the front line are seeing and listening to the academics is leading to 80%+ mortality rate. So something is wrong.

And those same people are probably still saying smoking is a risk factor…

• Anonymous says:

While hypothesis generating, such a mechanism would completely fail to explain the decreased P:F in COVID-19. The P:F is the ratio of PaO2, the partial pressure of arterial blood, to the fraction of inspired oxygen. The partial pressure of arterial blood is the amount of oxygen directly dissolved in the blood – i.e. not bound to hemoglobin. No amount of interference with the hemoglobin molecule will affect the amount of dissolved oxygen.

• Anoneuoid says:

I accidentally posted the above. It was supposed to include a note that there is no reason the hemoglobinopathy itself needs to be responsible for lower arterial blood oxygen. It could be something downstream like hypoxia leading to the microvascular thrombosis people have proposed, etc.

• Terry says:

“It’s really easy and fast to assemble and deploy digital systems. It’s a lot harder to assemble and deploy physical assets.”

Hardware is hard.

• Brent Hutto says:

But what could be a higher priority than knowing which hands-on health care providers are currently shedding virus?

• Rahul says:

Higher priority? Just to treat the patient in front of you!

Did we have the luxury of PCR testing in the Spanish flu pandemic?

I’m not saying we should not test health care providers! By all means, if you can. But I strain to think of that as the highest priority!

• Brent Hutto says:

I meant that testing front-line health care workers should be the highest priority for allocating COVD-19 *tests*. Not that testing is the highest prioirity of those workers.

• Rahul says:

So suppose I test NYC frontline workers. And I discover a positive. What’s the course of action?

Do we have the luxury to keep frontline workers home right now just based on a positive test?

• Brent Hutto says:

I don’t live in NY state and I’m not taking about what happens after a local health care system is overwhelmed. I’m talking about half empty hospitals with a handful COVD-19 cases in a state where the effects of the pandemic are still small. There is no excuse for not screening health care workers for the virus.

• jim says:

“What the heck is up with our entire public health and health care system’s aversion to screening people for this virus?”

I suspect two things:

1) lack of capacity
2) lack of replacements

• Martha (Smith) says:

“COVD-19 infections among health care workers are not being tracked. The local health department says that is not within their tracking obligation.”

Aargh! The old, “that’s not my job” excuse.

• Noah Motion says:

I think (royal) we can make the case that “we need more data” means (and has meant for some time) something like “we need careful, systematic, maximally-informative data collection practices.”

Yes, an approach that works with the shitty data we have would be useful, but, as you say, an essentially qualitative approach to resource allocation is probably about as good as we can do right now. It seems to me that we should put a lot more time and energy into ensuring that the data we collect is accurate and useful than we do trying to draw inferences from shitty data.

• zbicyclist says:

Yes, it’s not just MORE data, but data with less bias (or known, measured biases). We know there’s bias, but it’s hard to know how much and in what directions for these many national and state testing results — or, for that matter, death statistics where patients had multiple conditions.

To change Emerson a bit: “A foolish accuracy is the hobgoblin of small minds”.

or Rumsfeld: “this is a known poorly-known”

• jim says:

I doubt better data are forthcoming. The hospitals are overwhelmed. Data is no longer a priority, and possibly not even an afterthought.

• Brent Hutto says:

Of course in Lombardy or NYC it’s impractical to demand better data from hospitals and treatment providers.

But in the vast majority of this country, the COVD-19 cases aren’t yet swamping the hospitals. There is plenty of time and resources available in those places to do actual public health boots on the ground work before the tidal wave hits. But as it has been ever since news of COVD-19 first emerged from China, there’s no sign that anyone is gathering population data.

• Rahul says:

Maybe we are overdoing the data and the models? Right now the basic features of the problem seem pretty clear.

Models are really useful if they can lead to successful interventions. Forget successful, but what are some interventions or changes that have come out of the most sophisticated Covid models in the last week or so?

I’m not against models. But I feel we have too many people making models and too few using models +insights to come up with actionable interventions! We need more emphasis on solutions, execution,rolling things out and mobilization. That’s why I am amazed by the BadgerShield guys; simple idea; not much of a model needed. But now they are churning out face shield PPEs by the hundreds of thousands already. And their open source drawings are leading to so many copycat deployments across the world.

Take pooled testing. Excellent idea! I think the modelling and advantages are clear. The reason why it’s not rolling out so fast is we have too few people focusing on the execution and rolling out on the ground.

Execution is rarely glamorous work. And often puts you on the front-lines of risk. And needs skills of networking, persuasion, pressure, diplomacy, lobbying and maybe even lies which most of us look down derisively upon!

We need more Raoul Wallenberg’s.

• Anoneuoid says:

Afaict most hospitals in NYC are not swamped, but the ones that are swamped are not transferring the excess patients elsewhere for some reason.

• Brent Hutto says:

I’ve been on the looking for any insight I might gain into the transferring-patients situation. One obvious feature of this pandemic so far has been the extreme geographical and temporal concentrations that overwhelm LOCAL hospitals for periods of weeks to months.

Maybe there is some fundamental reason that it’s impossible to treat a COVD-19 patient who lives in a putative “war zone” at another less overwhelmed facility a couple hundred miles away. But I have not seen that sort of thing explained or even alluded to anywhere, yet.

• Anonymous says:

NY Post: “Cardiac arrest victims whose hearts cannot be restarted at the scene are now being left there — rather than being brought to coronavirus-strained hospitals for further revival attempts, according to a new guidance for medical responders.” Maybe the Regional Emergency Medical Services Council of New York has not realized that most hospitals in NYC are not swamped?

• Anoneuoid says:

Maybe there is some fundamental reason that it’s impossible to treat a COVD-19 patient who lives in a putative “war zone” at another less overwhelmed facility a couple hundred miles away. But I have not seen that sort of thing explained or even alluded to anywhere, yet.

I think it is as simple as patients = money (usually from insurance but I am sure hospitals are going to be paid by the federal government for each one marked “COVID-19”), with a lot of bureaucratic mumbo jumbo put in place as an excuse.

• Anoneuoid says:

NY Post: “Cardiac arrest victims whose hearts cannot be restarted at the scene are now being left there — rather than being brought to coronavirus-strained hospitals for further revival attempts, according to a new guidance for medical responders.” Maybe the Regional Emergency Medical Services Council of New York has not realized that most hospitals in NYC are not swamped?

How does a changed guideline mean that most hospitals are swamped? Maybe they are anticipating being swamped? Maybe they are taking advantage of these extraordinary circumstances to push off less profitable patients onto the NYPD?

Go watch videos of the hospitals looking less busy than ever before, which makes perfect sense since a hospital is the last place you want to be right now and they stopped all elective procedures, etc. I personally have had a dentist appointment delayed until mid-May.

People are jumping to crazy conclusions.

• Carlos Ungil says:

Brent, apart from the logistic difficulties involved in moving seriously ill patients hundreds of miles away that could easily contribute to the spread of the infection. According to some people like Giorgio Palu that’s one of the factors that may explain that the development of the epidemic in Lombardy has been quite worse than in Veneto.

• jim says:

Incidentally, with respect to admitting mistakes, I’ve been more or less willing to accept the official body counts up until now. But reality everywhere else is outdistancing the official numbers.

It’s dubious to base one’s opinions on Reddit photos and internet memes about urns in China. But it does seem clear after watching what happened in Italy and Spain – and given that Wuhan is a huge, densely populated city, completely unprepared in any way for this outbreak – that China’s numbers are low, probably way too low – who knows? perhaps by an order of magnitude.

Daniel Lakeland has been saying that all along. Now I think he’s right.

• No matter what one does with statistics, there’s modeling involved. That modeling may be implicit and unacknowledged, or it may be explicit. But it’s always there. Even reporting raw data requires some analytical steps to understand what the raw data represents as other comments in this thread have pointed out.

5. Jon Baron says:

A general comment from the perspective of decision making, not statistics. Sometimes we have to make decisions, even tentative decisions subject to revision later, on the basis of insufficient data and wide uncertainty intervals. The parameter being estimated (number of fugure deaths, global temperature in 20 years, if nothing further is done, whatever) is relevant to the decision, and the relative expected utility (EU) of different options. When uncertainty intervals are small, we can usually assume that EU of each option is a linear function of the parameter and thus use the best single estimate to decide what to do. When the uncertainty interval is large, we need to ask about the utility function itself. If, for example, the best estimate of warming is 2 degrees, but the uncertainty interval is from 1 to 3, we may decide that the EU difference between 1 and 2 is smaller than that between 2 and 3. Thus we would want to err on the side of the “worst case scenario”.

• Dale Lehman says:

Even more important becomes the issue of irreversibility. Reversible decisions do not pose a problem, but when they can’t be reversed, that fact must weight heavily in decision-making. You can’t bring someone back from death though you can put someone back to work. You can’t undo a sea level rise, but you can ramp up fossil fuel use if the environmental damages turn out to be less than expected. Of course, it is not this simple – there are irreversible costs attached to all decisions, even the apparent “safe” alternatives. But it becomes important to specify the extent to which alternative courses of action can or cannot be reversed. In the presence of irreversibility, there is option value – a value attached to keeping options open.

• Martha (Smith) says:

“But it becomes important to specify the extent to which alternative courses of action can or cannot be reversed. In the presence of irreversibility, there is option value – a value attached to keeping options open.”

+1

• +1 on relating this to needing to make decisions now. Good thing I scrolled down before writing a paraphrase of this response!

I wanted to add that in a lot of cases, wide uncertainty is the best we can do given available knowledge (aka theory) and data.

• Shravan says:

This is the situation in (psycho)linguistics. There is never enough data. We can act like we learnt something from an expt but we have to stay humble. Unfortunately ppl don’t understand that. A linguist once said she was going to “shock and awe” me with data and showed me 10 null results. See? Definitive evidence of no effect.

6. Mike Rivard says:

As a novice at torturing numbers, i can only ask: Have you excluded the theory that there is more than one strain of virus at play thus confounding the analysis? This may also be time based .. where strain one informed the model in january and strain 2 kicked in in march

so many questions ….

• Rahul says:

I thought that’s pretty well addressed? Aren’t various locations at intervals re-sequencing the virus? So far I thought that the consensus was that although there are some changes in the genome they aren’t very substantial.

I could be wrong!

• Mike Rivard says:

I agree that they can mechanically measure mutations …. to rephrase my question: Are there clues in the data that suggests either two process co-mingled or a shift in process/outcome?

Are there two or more groups within the larger data set; thus explaining the super-wide intervals?

• Rahul says:

…..the bigger question seems changes in genotype of the human hosts based on ethnicity. Those could also be a big factor at play explaining the heterogeneity of responses?

• Martha (Smith) says:

“…..the bigger question seems changes in genotype of the human hosts based on ethnicity. Those could also be a big factor at play explaining the heterogeneity of responses?”

I think it would be more accurate to say, “… heterogeneity in genotype of the human hosts based on ethnicity. Those could also be a big factor at play explaining the heterogeneity of responses?”

7. Carlos Ungil says:

> The wide intervals are helpful in telling you that more information will be needed if you want to make an informed decision.

You can always make an informed decision with the information you have. Of course, if you had more data you could make a more informed decision.

• Brent Hutto says:

Politicians tend to make decisions “informed” by cherry picked data presented to them with maximum spin by “experts” trying to achieve their desired result. And of course a given politician’s “experts” must be politically in-group for that specific politician.

• Martha (Smith) says:

“You can always make an informed decision with the information you have. Of course, if you had more data you could make a more informed decision.”

+1

8. James Annan says:

The basic problem is that the only data that are pretty reliable are deaths. So it may be a huge epidemic of mostly mild cases with a low death rate, or a small epidemic of a rather nastier disease. This isn’t a problem of the models per se, it’s a problem of the data.

On the topic of Bayesian estimation and model fitting….

https://julesandjames.blogspot.com/2020/04/what-can-we-learn-from-wuhan.html

Sorry not Stan though. But my efforts were sufficient to indicate that the IC group were using very optimistic parameter values in their modelling for the govt policy support…

• Anoneuoid says:

I don’t see how people are quantitatively modelling this without taking into account the testing rate (which appears to be unavailable for most places).

Also, heart attack rate increases 3-4x during soccer matches (apparently from excitement/stress). So don’t rule out panic attacks causing shortness of breath and then heart attacks from watching/reading the news.

Viewing a stressful soccer match more than doubles the risk of an acute cardiovascular event. In view of this excess risk, particularly in men with known coronary heart disease, preventive measures are urgently needed.

https://www.ncbi.nlm.nih.gov/pubmed/18234752

Myocardial injury associated with the SARS-CoV-2 occurred in 5 of the first 41 patients diagnosed with COVID-19 in Wuhan
[…]
In another report of 138 patients with COVID-19 in Wuhan, 36 patients with severe symptoms were treated in the ICU1. The levels of biomarkers of myocardial injury were significantly higher in patients treated in the ICU than in those not treated in the ICU (median creatine kinase (CK)-MB level 18 U/l versus 14 U/l, P < 0.001; hs-cTnI level 11.0 pg/ml versus 5.1 pg/ml, P = 0.004), suggesting that patients with severe symptoms often have complications involving acute myocardial injury1. In addition, among the confirmed cases of SARS-CoV-2 infection reported by the National Health Commission of China (NHC), some of the patients first went to see a doctor because of cardiovascular symptoms. The patients presented with heart palpitations and chest tightness rather than with respiratory symptoms, such as fever and cough, but were later diagnosed with COVID-19. Among the people who died from COVID-19 reported by the NHC, 11.8% of patients without underlying CVD had substantial heart damage, with elevated levels of cTnI or cardiac arrest during hospitalization. Therefore, in patients with COVID-19, the incidence of cardiovascular symptoms is high, owing to the systemic inflammatory response and immune system disorders during disease progression.

https://www.nature.com/articles/s41569-020-0360-5

The other day my younger brother signed for a package then told my mom it was coronavirus and she freaked out. It is funny, but this thing is putting people on edge.

• jim says:

” the only data that are pretty reliable are deaths. “

Are they reliable? Now I don’t think so. If someone dies having not been tested, why would overwhelmed workers waste time or resources testing them?

• jim says:

People aren’t grasping the capacity issue.

NYC had only a handful of fatalities when Cuomo imposed the shut-down order, so the fatality count is *WAY* behind what’s happening in hospitals. From the sounds of it the typical hospital has room for a few tens of serious cases. That’s like an order of magnitude below the demand.

• It has nothing to do with capacity per se. It does have a lot to do with “official capacity”.

If you want *information* then you can easily and by easily I mean *EASILY* get every major university in the US to start running PCR reactions for a few thousand dollars a day per university. There is enough PCR capacity to do easily 1 million PCR reactions a day in the US no problem. You can pool samples 50/1 pretty easily that’s now been shown in practice. Within a few days we could be doing 50 million PCR tests a day.

By the end of this week, if there were *political will to do it* we could have tested every single person in the country.

Now, collecting the samples is a bit more expensive, but the US Military could mobilize groups to do pop up sample collection. With blasting it all over the news, we could have block-by-block sampling rolling through. The Census bureau and CDC/wonder data could be telling people where the population centers are by risk factors easily and we could be prioritizing those.

The US military, specifically the Army and Marines, has ENORMOUS existing capacity to put boots on the ground ANYWHERE in the WORLD at significant scale. They are PAID to do hazardous duty and protect the country.

With a properly thought out plan, these people could have had WEEKS of preparation time teaching them about biohazards starting back say Jan 15.

We just spent 2 TRILLION on “stimulus” if we’d spend 100B on putting boots on the ground…

it’s all down to political will knowledge, and intelligence. What we need is someone like Eisenhower with knowledge of strategy and tactics, what we have is PT Barnum’s younger and less successful brother.

• we have an existing emergency alert system that blasts out things like amber alerts regionally. We could easily be informing essentially everyone of the arrival of a mobile military sample collection station at a local parking garage, with a schedule… if your street address ends in, and your SSN ends in … arrive between x and y today…

using that kind of regional randomized sampling we’d easily get prevalence info across all the major metropolitan areas across all the states…

it’s all down to a failure of imagination.

• Anoneuoid says:

I don’t think that would work out the way you are thinking… Why don’t they just test donated blood for antibodies?

• Anoneuoid says:

(April 1, 2020) — San Miguel County, CO — San Miguel County Public Health officials announced today that less than 1 percent of the COVID-19 blood tests done last Thursday and Friday were positive for antibodies and 97% of these groups tested negative.

Positive results on the first test indicate the presence of COVID-19 antibodies in the blood. This means that the individual has been exposed to COVID-19 and may or may not have ever experienced symptoms.

Another 2%, although technically considered negative, were “indeterminate,” showing a high-signal flash meaning they have an increased chance of converting to positive.

So 1-3% positive for antibodies in a small colorado town.

• That makes sense. Now that we have widespread stay home orders it’ll be hard to guess at spread rate. Antibody testing is another excellent strategy for surveillance. Hopefully we will roll that out widely for random community monitoring

• Rahul says:

Daniel, Anoneuoid:

I heard of another idea which sounds similar to pooling but seems a step forward; wondering about your thoughts:

Multiplexing. So basically instead of just pooling the samples each sample goes into a certain set of tubes and then by using the right combinations etc. you could potentially get to individual results with fewer tests (whenever you do detect a positive). I don’t have the exact details fleshed out but wondering if that would be possible?

Thoughts?

• Yes, it’s in a grant proposal a friend and I put together already ;-)

• Rahul says:

@Daniel

Isn’t it a bit late now for prevalence testing? Without any pooled testing, & given what we know, what’s your best guess of the infection rate right now? Is a more accurate prevalence test going to help much?

• Brent Hutto says:

Around 0.1% of the USA population has tested positive and 0.002% have died. Everyone expects those numbers to double roughly every 3 days.

A very important question is how long will this doubling every three days go on? That will determine the ultimate impact.

Assuming that those who have been infected will be immune for at least some extended period of time (which we don’t actually know for sure) then the doubling can only continue as long as there are plenty of never-infected individuals for it to infect.

If 30% of the population is already infected right now then by the end of this month almost everyone who is going to be sick will be sick. But if only 1% of the population is currently infected then the doubling can go on for a long time yeet. The tiny bits of evidence we have from places which have done prevalence testing *suggest* the prevalence would have been around 1% a couple week ago but that’s really not much more than an educated guess without large scale testing.

• Brent has it right. Prevalence and esp it’s rate of change affect our response. If stay at home orders are working prevalence will grow slower. If one area has high prevalence now it will have a surge in hospitals 7 to 15 days from now. Etc.

In the U.S. California put a stay at home orders Mar 19. Georgia held out til Apr 2 or so. What did that do to Georgia’s infection rate? We will find out in 12 days… That’s 12 days we have to shift hospital supplies… Where should we shift them?

Is CA order working? Perhaps overall it is but in say immigrant households not so much as they continue to participate in business activity out of fear of homelessness… I don’t know. But we could find out… Through substantial prevalence testing

• Anoneuoid says:

When Avis volunteers summoned them for blood donation, they were confident that they would find a high number of Covid positives19. Confirmation has come from the results of tests and swabs: out of 60 citizens of Castiglione D’Adda, one of the municipalities in the former red area of ​​Lodi, 40 tested positive without knowing it. All asymptomatic, escaped official statistics: they came into contact with the disease, they did not develop it, but they produced the antibodies, as if they had been vaccinated.

• Brent Hutto says:

If by some miracle, there are a few tens of millions of people in USA who are carrying antibodies (and therefore presumptively immune to reinfection at least in the short term) that would be a great pool of people to rely on for keeping society functioning as we enter the next Great Depression. Not to mention the peace of mind it would confer. I know if someone gave me an antibody test and it showed I’d already had it and never known, that’s the kind of thing I’d want to know.

And if it the prevalence is not that high today, it will in six weeks or six months or some time in the forseeable future. Knowing how has had it and who has not is as fundamental to Public Health as knowing who is sick and who isn’t right now.

I can only imagine there must be political reasons why everyone in authority in this country shuts their eyes, covers their ears and doesn’t want to hear any talk of testing asymptomatic people.

• Anoneuoid says:

See the post above about this being treated wrong because the disease mimics HAPE instead of ARDS. The mechanical ventilation is probably killing the patients, it makes perfect sense with the missing smokers and everything else seen:

https://en.wikipedia.org/wiki/Altitude_sickness#Signs_and_symptoms

So the mortality rate should drop dramatically as ICU units change their treatment strategy to be correct. My impression is they are allowed to be pretty independent, so the wrong centralized standard of care shouldn’t be able to stop it. Also:

Cigarettes as an aid to climbing Report, November 21 1922

Captain GJ Finch, who took part in the Mount Everest expedition, speaking at a meeting of the Royal Geographical Society, London, last evening on the equipment for high climbing, testified to the comfort of cigarette smoking at very high altitude. He said that he and two other members of the expedition camped at 25,000ft for over 26 hours and all that time they used no oxygen.

About half an hour after arrival he noticed in a very marked fashion that unless he kept his mind on the question of breathing, making it a voluntary process instead of an involuntary one, he suffered from lack of air. He had 30 cigarettes with him, and as a measure of desperation he lit one. After deeply inhaling the smoke he and his companions found they could take their mind off the question of breathing altogether … The effect of a cigarette lasted at least three hours, and when the supply of cigarettes was exhausted they had recourse to oxygen, which enabled them to have their first sleep at this great altitude.

https://www.theguardian.com/books/2007/oct/17/sportandleisure.sport

It is quite possible that the infected just smoking a cigarette every couple hours for a few weeks until they are immune keeps them safe. And pulse oximiters will be a very useful non-invasive screening tool.

• Carlos Ungil says:

Do you now like anecdotal evidence from villages in the worst-hit areas in Italy? Good!

For context, 51 people (over 1% of the population) died in Castiglione D’Adda in four weeks (February 23 to March 21, we don’t have more recent data yet).

In the corresponding periods in 2015-2019 the number of deaths was 6, 1, 6, 7, 4 (average 5).

The total number of deaths in the years 2012 to 2018 was 52, 46, 44, 69, 56, 60, 61 (average 55).

• Anoneuoid says:

Do you now like anecdotal evidence from villages in the worst-hit areas in Italy? Good!

Nope, I’d like much better data. But unlike you that is not the only thing I look at.

I also posted the results from the small Colorado town of <=3% in this thread.

• Brent Hutto says:

Carlos,

For my part, I place about as little credence in anecdotal “evidence” as anyone. But at this point, if someone would come to my town and administer twenty COVD-19 screening tests to randomly selected, healthy, asymptomatic individuals the results of those twenty tests would instantly, like, double my effective knowledge of this virus.

I continue to be appalled by the lack of basic, Public Health 101 level investigation into a virus that’s killing tens of thousands of people. It’s like they’re saying, “There are so many sick people it’s a waste of time trying to find out how they’re getting sick or exactly what sort of pandemic we’re dealing with”.

• Carlos Ungil says:

Brent, I’m sorry if you felt somehow attacked by my reply to someone else. I was just pointing out that, looking at that village at least, high prevalence comes hand in hand with high body count.

• Brent Hutto says:

Sorry, Carlos. I didn’t feel attacked and I knew you were responding to Anoneuoid. My defensive tone was unintentional.

I was just riffing on how I would normally turn my nose up at just a few tests here and a few there but the paucity of “real” data in this pandemic has lowered my standards considerably!

• bjs12 says:

This is not correct.

The rate limiting resource is reagents. My wife is in charge of laboratory testing (and hence, coronavirus testing) at a major hospital group in a large city. She has enough reagents to do 400 tests per day for another week or so. You could lead a battalion of troopers onto campus with one, ten, or a thousand PCR machines and not change that fact one iota. Until reagents become available at a totally different scale, your suggestions are unworkable. Maybe this will help you see where your analysis went wrong: xkcd.com/793

• This is the limiting problem for people at hospitals getting their reagents from official FDA approved sources.

My wife’s lab easily orders 50 vials of *custom* primer sets per month, there are easily 10,000 research labs ordering like that. So 500,000 vials of custom primers per month can easily be made already today. If you’re just making repeats of a single primer set, it would be inconceivable that the actual existing mfg capacity to produce COVID specific primers would be less than 100x that big. The defense production act could be used to move everyone who makes primers to making COVID primers full time 24hrs a day for weeks on end easily.

So 50M vials could be produced a month no doubt. Each vial can run hundreds of reactions. So 500M tests could be run by the end of the month, each test could be pooled 50 people, so 25B screening reactions could be done in a month in terms of reagent capacity, or at least no less than 1/100 of that.

For research labs the real expense is the TAQ polymerase because people use fancy commercial versions of it… a friend of mine produced a decade supply of TAQ as an undergraduate project back in the 90’s… it’s expensive because regulations not because it’s hard to make.

the actual technological capacity is not the issue. it’s bureaucracy and lack of imagination.

• BTW ordering a CUSTOM primer set vial for a research lab cost \$5 and is delivered 2 days later.

• bjs12 says:

“easily”, “easily”, “inconceivable”, “easily”, and “no doubt”? If those sentiments are correct, then you might be a mass murderer (going for lighthearted hyperbole). Why are you writing in all caps on a statistics blog and pestering your wife when you should be contacting every hospital lab director in the country and letting them know that their sleepless nights are over because you’ve “easily” solved the nation’s testing problem? I would be thrilled if you were right, but I suspect you are not.

• The fact is *every* biochemist I know is LIVID about the lack of testing. I have at least 3 or 4 knowledgeable ones saying the same things to me.

Here is a PhD in molecular biology quoted:

“When it comes to the current coronavirus pandemic, one of the facts that’s most baffling to me is the lack of testing capacity in the United States. This isn’t a problem in many other nations, so we can reasonably conclude it’s an issue with our policies rather than some fundamental limit due to the laws of physics. Full disclosure: I’m not an epidemiologist or a public health expert. That being said, as someone who works in the biotech industry, I’m confident I could spend around \$1,000 on supplies and have a fairly accurate diagnostic test for COVID-19 that’s scalable to 1,000 patients and compatible with standard equipment in any molecular biology lab. That’s \$1 for every person who wants to get tested. Or \$327 million for everyone in the USA. In other words, for a tiny fraction of the amount of money that’s been lost in the stock market today alone, we could easily screen EVERY man, woman, and child in the country.”

• bjs12 says:

We’ll need a tribunal for all of these people you know who aren’t fixing the actual problem. We don’t need your acquaintance to develop a test, we already have a test. We need you to get on the horn, order custom (sorry, CUSTOM) primers, and deliver them to hospitals. Which, as you know, is “easy”. If you’re not doing that, then you’re just blowing smoke.

• The point is, it’s not the technology that’s the hold up. It’s not that we can’t as a matter of technology, produce sufficient reagents to do the reactions as you asserted. It’s the POLICY. And none of the biochemists or molecular biologists have any control over policy.

• bjs12 says:

No one is too happy about how the FDA has behaved. But I don’t understand why you don’t create the test and order the custom primers for it, and then offer it to hospitals (or use the sidelined PCR machines at universities you’re always going on about). Does your argument boil down to the concern that if you solve the nation’s testing problems (easily, quickly, and cheaply) the FDA will what? shut down the operation of a national hero?

Of course, you may want to consider the other possibility, which is that it is not as easy as you want to believe it is.

• bjs12 says:

You remember when Apple asked the federal government for permission to develop the iPhone and change all of our lives? I don’t either.

• You’re just reiterating my point, which is the problem is policy. Without coordination, none of what is *perfectly possible* theoretically is going to get done.

Why do we have 100k homeless people in LA county? Is it because manufacturing housing is way too vastly expensive? Heck no. It’s because without concerted coordinated policy change, if someone wants to bulldoze a city block and put up a 4 story apartment complex, they simply can’t do it.

On the other hand, as a matter of policy CA has already decided that they can seize hotel properties and house patients in them… voila, beds potentially available to house patients quadrupled overnight.

policy. everything wrong with this whole debacle is policy. And this is after all a stats + political science blog. I’m pointing out that both stats and policy are broken here.

• Anoneuoid says:

Daniel, your recommendations for the government forcing people to do things and taking their stuff away are getting more and more extreme. Doing that type of stuff is not free, you need to consider the costs/risks.

• Universities are already forced to shutter their buildings and stop using their machines. You think they won’t welcome having the govt give them an opportunity to make some money renting out floors in their buildings and access to their machines? They literally suck down govt grants all day long. their whole purpose is to provide labor for government projects. Just give people govt grants to do this research level testing. I’m not suggesting to send troops with guns to roust lab techs, just organize a concerted effort to pay people reasonable normal lab tech wage rates to do something useful that is technologically possible today.

We’re ALREADY confiscating people’s time and the use of their machines (ie. forcing them to stay home) all my proposals here are *increasing* people’s options not decreasing them.

• Anoneuoid says:

We’re ALREADY confiscating people’s time and the use of their machines (ie. forcing them to stay home) all my proposals here are *increasing* people’s options not decreasing them.

Yes, and now that is not enough.

as a matter of policy CA has already decided that they can seize hotel properties

Do hotels usually run on government grants? I don’t know.

We could easily be informing essentially everyone of the arrival of a mobile military sample collection station at a local parking garage, with a schedule… if your street address ends in, and your SSN ends in … arrive between x and y today…

So the military sets up bases around the country and everyone is forced to go there to get a check up?

• bjs12 says:

The difference is that no one thinks solving the homelessness problem in LA is “easy”. You have literally stated that solving the coronavirus testing problem is “easy”. The only barrier you offer is policy. But policy is not in your way. You can develop a test for coronavirus. There is nothing stopping you. Research labs are free to do this. There was a push by the FDA a year or so ago to forbid labs from developing novel tests without FDA approval, but luckily that didn’t happen. You (not me) said it was easy to order custom primers. The FDA may restrict manufacturing of primers for the CDC test, but you aren’t running that test. You’re making your own. Order the primers. They’ll be here in 2 days, I hear. If you’re really anxious, label your test to be for budlightvirus for research purposes and then dare them to come after you once you’ve solved the national testing shortage.

I am not arguing that FDA policies have been good or neutral, they’ve been a disaster. And surely overturning them would be a wonderful way to improve things. But it doesn’t imply that it would *solve* everything. And it certainly doesn’t imply that that is the *only* way to improve things. You could also just go right ahead a produce a custom test and custom primers without any check-in with the FDA at all. The question is why hasn’t anyone done so?

The answer has little to do with policy and everything to do with the fact that setting up valid testing at a massive scale is really hard. Simple example: we’ve been hearing for a week or more about a test kit that will take 15min. Everyone asks my wife, “why don’t we just get these test kits, they seem great?” And her answer, shockingly, is NOT, “why golly, I was too stupid to think of doing the easy thing, I was just going to keep on doing the old hard way that’s killing people.” Her actual answer is, “the company won’t send us any of those kits.” That means the company is talking out of one mouth to the media but can’t actually back up their promises. Or, put another way, it is really hard to bring this process up at scale.

Similarly, you can point to individual biochemists who can technically do the test, but somehow none of them have solved the national problem. They work in small labs, with small machines, simple processes, tight control over their staff, etc. They don’t actually know what it means to bring up lab testing at scale. You know who does? The lab directors who are working on this problem. They know that most of the work is not the testing. It is the collection, processing, transportation, coordination, testing, staffing, reporting, etc. that make the process difficult. And each one of those is only a heading. The complexities within each one are manifold. And, of course, the interactions between all of the steps tend to multiply the difficulty, not just add.

You obviously know something about the science involved (thank god, can you imagine this discussion with a cretin from the depths of the internet?), but just as clearly don’t know anything about scaling the lab testing process out of the lab into the real world. I’ve enjoyed this discussion very much, and I’m happy to bring your ideas to the one person I can who is positioned to do something. I asked her yesterday about pooling samples. She said it was considered and rejected. Based on this thread, I’m concerned that you’ll take that to mean that people just won’t listen to you even when you’re obviously correct. But, again, I would like to at least offer the possibility that you may not know as much about this area as the people who do it for a living. Keep the ideas coming, but be prepared for them to not be implemented. And maybe entertain the possibility that the reason they are not implemented is not the same as the only reason you can think of. There might be deeper reasons.

Now I’m off to do the only thing I, personally, can do to help right now: keep the household afloat of someone working diligently to bring up as much quality testing as is humanly possible through any and all means available. Wish me luck!

• I didn’t suggest the CA hotel seizures, just mentioned that it was an example of how policy changes the situation. Let someone else argue about whether or how much the hotels should get paid. The point is it was *illegal for them to operate as a hotel* and the facility was shuttered, and then… suddenly by change of policy, the facility was usable.

you can argue that everyone should be able to run hotels and have big conferences and let everyone who goes get coronavirus and get sick… I doubt many will fall in line with that argument.

You can’t argue that it’s ok to force hotels to shutter, but not ok to pay them some moderate amount of money to use their beds. THat makes no sense.

As for the military, I’m not suggesting forcing people to give samples, I’m suggesting using highly mobile resources to allow sample collection at many pop-up locations. You think people won’t queue up at 6ft distances to give up their sample? Of course they would. They’re *paying* people already for the privilege of getting their sample tested as it is. Instead, give them a test free and use the data to move resources into place.

I’m fine with let’s do the minimal amount of forcing people to do stuff… That’s literally what I’m suggesting here, give people more options than just “stay home and go bankrupt”. Options where we move much more quickly towards a day where they don’t have to stay home.

• When I said easy, I meant technologically easy to run reactions at scale that are vastly less expensive than the \$50-\$100 per sample cost of commercial stuff that hospitals run.

Every single bio lab does these kinds of tests every day. So in say November, there were undoubtedly 10000 labs at least in this country running a few hundred reactions a day just to support their research. This is how they genotype mice, or the like.

So technologically, running 1 Million reactions a day was *already going on* in November. It literally was “business as usual” to run 1 million reactions a day in this country just as a side-line to support figuring out which baby mice have which genes.

In CA they ran ~2000 COVID tests yesterday. Consider the difference in scale between “get bio labs to do this at scale the way they already know how” which we have good evidence could easily be vastly higher than 1 Million reactions a day (because most people don’t run reactions all day long, they do one plate a day to figure out something incidental to their research or similar).

Why are we not doing this… I’ll give you an example, Univ of Washington *was* doing this and had government agents come and shut them down.

The university where my wife works has had meetings where all the professors have said things like “lets get testing going, let’s do our own testing” and the universities have said that they’ve been told to shutter their buildings and no one can enter them, literally there’s a list of like 10 people allowed to enter their buildings right now.

So, once again, why can’t I just go out and create a test and start running it… It’s illegal, people will be arrested, universities want nothing to do with it because of policy and financial risk, it’s all about law, politics, etc.

Now, put out a grant saying “any university willing to run these tests can get paid \$15 a reaction” and you’ll see millions of reactions a day.

• jim says:

“it is really hard to bring this process up at scale.”

Thank you, this is the problem. Shit doesn’t just suddenly materialize out of nowhere. This is ***ESPECIALLY TRUE*** for health care equipment and processes, which are highly regulated for good reasons.

US healthcare suppliers don’t have extra production capacity just sitting around waiting for a pandemic to break out and for demand to rise 10-fold. They’re already maxed out. They handle peak demand with OT hours, not by opening new production lines and hiring and training more staff. So they’re short both physical production capacity and trained staff to man that capacity.

So I don’t know about reagents and chemicals. But for medical equipment and supplies that are produced in a regulated environment, the production capacity ain’t there and it won’t for a long time.

• Jim, you’re absolutely right about say making ventilators…

When it comes to running PCR reactions we DO just have a million reaction a day capacity sitting around. It’s just not hospital related, it’s research related.

There is no reason why we need “hospital grade” PCR reactions right now. The way those work is that they’re kits that make it easy for people to do a couple one-off reactions without having to ramp up to efficient production type testing… Normally you might need to test a couple hundred people a week for say norovirus… So on any given day you’re doing ~ 10 reactions maybe. And you’re doing 10 reactions for some other pathogen, and 10 reactions for a third pathogen… So you need a kit where you can open up the kit, have all the stuff to do 10 reactions kind of ready to go, and then you can close it up and be done for the day, and maybe open it tomorrow or the next day.

When it comes to biotech, they might have 30,000 cages of mice at some campus and need to genotype 3000 mice a day for the same genotype over and over and over… They’re set up to do bulk testing of the same stuff over and over. They don’t open up some \$500 kit that handles 100 reactions and get out 10 aliquots, they open up a case of reagents and pull out tube after tube filling up 486 well plates and running them on 10 different machines simultaneously. 4860 reactions in an hour no problem.

Researchers know this, but apparently everyone else doesn’t. This is why researchers are fuming about lack of testing.

• jim says:

Daniel,

Yeah, OK, I accept that but our system doesn’t have the distinction of different “grades” for human care, so even though the capacity might be there for it the regulatory system doesn’t allow it.

The regulatory system has been tightly constrained intentionally. It’s whole purposed is to anticipate and counteract the exact kinds of shortcuts you’re talking about. And probably a lot of the same people who are complaining now about its inflexibility were complaining six months ago that it allowed too much leeway for companies. So it’s going to take a while to get that kind of stuff through.

• The wapo now has an expose and it confirms literally everything I’ve been saying…

https://www.washingtonpost.com/investigations/2020/04/03/coronavirus-cdc-test-kits-public-health-labs/?arc404=true

• Anoneuoid says:

Thanks for clarifying, that sounds much more reasonable.

• Anoneuoid says:
• jim says:

The question is: is the fatality count accurate? The answer is: no. There’s not enough staff to meet demand, and that includes the time to do testing and paperwork on fatalities. Whatever they do they do but chances are good that some % of COVID fatalities aren’t getting reported that way because of lack of both staff and physical resources.

• James Annan says:

The numbers aren’t perfect but are good enough and certainly better than anything else available. People dying rapidly of viral pneumonia get noticed pretty quickly in most European countries anyway. I don’t know about the USA situation.

• Carlos Ungil says:

Dead people is noticed pretty quickly but those deaths are usually not included in the “official” COVID-19 tally if no test is done (as it happens very often in most European countries due to lack of capacity or lack of interest). As Statistical Offices are starting to provide mortality estimates the undercount is becoming apparent.

• jim says:

Carlos said:

‘Dead people is noticed pretty quickly but those deaths are usually not included in the “official” COVID-19 tally’

That’s gotta be the case. I’m willing to bet once a cause of death goes on a document it has serious legal implications so it can’t be guessed at or made in haste.

• Anoneuoid says:

Breaking: Pence just said americans don’t have to worry about the cost of testing or treatment. Hospitals will be compensated directly for any treatment of coronavirus patients.

It is interesting to think about what incentives this creates.

• zbicyclist says:

First hurdle is believing the administration will actually carry through on that policy. It’s not really a policy until it is officially issued. Not that I don’t trust the government, but I don’t.

• mort says:

jim, Actually my understanding is that cause of death is actually very poorly and imprecisely recorded, which is a global public health problem in general, because it makes tracking changes in causes of mortality needlessly difficult (Jon Oliver even did a funny show about this, you can watch it on youtube I’m sure). So you are right that deaths due to coronavirus are not accurately being recorded as deaths due to coronavirus, and we will be forced to look at surplus deaths later to estimate the true toll of coronavirus, but the reason that these deaths are not being given a cause of coronavirus is not because people are so careful about assigning cause of death.

9. Terry says:

Excess mortality is the only reliable statistic I can think of. Death certificates are probably still highly reliable as to the number of deaths (although not reliable as to cause of death).

• mort says:

Exactly. Death certificates are very unreliable as to the cause of death, which is quite unfortunate. Therefore we will have to look at excess mortality (which unfortunately will be a function of both coronavirus and of ALL effects of social distancing- perhaps fewer deaths due to car accidents, and perhaps more deaths due to insufficient medical care for non-Covid-19 diseases e.g.)

• Fortunately car accidents are pretty easy to identify and relatively reliable cause of death info. But, yes this is true overall.

• Martha (Smith) says:

“Death certificates are very unreliable as to the cause of death”

Yup. When in doubt, give COD as “heart failure”.

• Terry says:

I wouldn’t mind sweeping deaths (or reductions in deaths) from other causes into the calculation.

When an influenza sufferer dies because they can’t get a hospital bed, it is reasonable to attribute that to CV.

When someone doesn’t die in a traffic accident because they are locked down, it also seems reasonable to attribute that to the side effects of CV.

In any event, these factors seem modest, as opposed to the absurdly large uncertainties from estimates based on testing.

• jim says:

“In any event, these factors seem modest, as opposed to the absurdly large uncertainties from estimates based on testing.”

Yup.

You work with the noisy signal.

• Brent Hutto says:

You work with the noisy signal and, at least in the case of the USA public health system, you resist tooth and nail any suggestion that you reduce that noise by doing systematic testing of the asymptomatic.

• jim says:

I’m not saying that. I’m saying until you get better data you gotta live with what you got.

Everyone keeps talking about testing testing testing but so far its a bunch of bullshit. If the materials to do the job really are available, then a group of academic labs could and probably would coordinate and launch the project in every state in a few days.

So until someone can figure out why that is, the testing unicorn will stay out in the pasture galloping along rainbows.

• If it weren’t for the fact that when they actually did that back in February a U Washington, actual federal agents showed up and shut them down because as soon as it was declared an international emergency it became illegal for them to do such things.

Also around the country county and state officials have called universities and demanded that they shut all their buildings, and provide a tiny list of people who are allowed to enter those buildings to do maintenance and security functions only, and no one else is allowed in the buildings.

At USC where my wife works, at least people doing direct COVID research are allowed to continue working. But of course no-one is connecting them up to actual field collection agents… So they’re sitting there with capacity to test, and do various things, but no-one is utilizing their resources.

So, that’s why. TOTAL LACK OF LEADERSHIP.

The washington post has a nice expose: https://www.washingtonpost.com/investigations/2020/04/03/coronavirus-cdc-test-kits-public-health-labs/?arc404=true

• Also, my wife has read that in Germany, where Angela Merkel is a PhD in quantum chemistry with an additional 20+ years of government experience… she immediately made every effort to rope in all the universities, and they are in fact doing exactly this idea, using univ. resources for widespread testing as ordered by the federal government. Unsurprisingly, they have a far better knowledge of their pandemic, and their CFR is low because they don’t have a ridiculously biased estimate of the case number, because they don’t just test seriously ill people either.

https://time.com/5812555/germany-coronavirus-deaths/

Above article says on Mar 20 they were testing 160k tests a week…

They report being able to do 500k tests a week.

Per million people Germany is far ahead of the US:

https://ourworldindata.org/grapher/full-list-cumulative-total-tests-per-million

• Anoneuoid says:

Consider what people are going to think if widespread antibody testing shows most people already had this before the shutdown with society barely noticing. And most of the deaths were actually from misapplying ventilators to treat ARDS when it should have been hypoxia… And further that a simple noninvasive pulse oximeter that every family should have anyway like a thermometer was all that was needed for screening instead of getting your throat swabbed by someone in full PPE.

The cherry on top is the 17 years they had to study SARS and no one figured any of this out.

This could be it for the credibility of medical research, I don’t think people will forget a folly of this magnitude.

• Anoneuoid says:

And of course they would have figured the hypoxia thing out back in 2003 if they hadn’t covered up the key hint that smokers were rarely getting sick.