Hacker News Reader

Comment by ggm | Hacker News Reader

I will repeat a question I would love an epidemiologist to reply to: Where is the random background testing? Not testing "I feel unwell" presentations at random, truly going into the field and testing the wider population at random, healthy (believed) or not?

You cannot model the IFR accurately from skewed samples and it feels like we aren't even doing skewed sampling.

View on HN

Epidemiologist here: you're right.

Testing resources are in short supply, so testing is being performed to guide clinical decisions (ie sick people) rather than public health/science (ie random sample).

The handful of serosurveys that have been performed have been quite valuable, but there aren't nearly enough.

Hopefully this changes as testing capacity ramps up.

When the clinical demand wanes, I hope there will be extensive antibody testing in various regions to do a comparison between how many people actually got it vs. how much damage it did. If, as I suspect, there are significant differences (ex: almost everyone in this town seems to have the antibodies, but nobody died, vs. a third of these people have the antibodies, but a lot of people died), we might be able to learn things such as which public policy decisions mattered most and maybe even which ethnicities are more vulnerable or resistant genetically. If a genetic variation is found, it might even lead to a new generation of preventative therapies that go beyond vaccines to shield people against viruses that don't even exist yet.

Not really.

There are a million reason for differences like that, without studying the problem extremely carefully, you'll never know which of those reasons was consequential. For instance, more people may die in a certain city because there is more pollution in the air, or the concentration of fluoride in the city's water supply is higher, or maybe that city houses a large meat packing plant and an abnormally large number of residents had to roll the dice and work throughout the lockdown. The number of possibilities are nearly endless, and you'd need to eliminate them prior to drawing any conclusions.

Even ethnically, you can't really draw any conclusions. Were people of this ethnicity or that ethnicity more likely to be essential Walmart stockboys than others? That one little detail can have an enormous impact on medical outcomes depending on the contagiousness of a pathogen like covid-19.

To draw any reasonable conclusions, you'd really need far more than just a comparison between regions, because the different regions have such different public health needs and environmental realities. You'd even need more than a simple comparison between ethnicities. You would need to go wayyy more deep. But I can guarantee that researchers will dig deep to find anything that can be used attack similar pathogens in the future.

> without studying the problem extremely carefully, you'll never know[...]

It seems likely to me that this problem will get studied extremely carefully.

As I said, test for antibodies. Put the "who's more likely to work at Walmart" speculations aside and start with, "Of the people who were provably infected, how likely were they to die?" question. If there are large differences, then look at age (for sure), prior health factors, gender, ethnicity, etc., to see what (if anything) made the virus not matter, and consider whether there might be other ways (besides vaccination) to make future viruses (prob some category) not matter to anyone.

I'm convinced this is my aliens theory, but I do think pollution plays a big role. My hypothesis being that long term exposure to high pollution is akin to chronic asthma.

Fits in Italy, as their elderly are retired locals exposed to their terrible air for life. In China, the hard hit had a transient population that moves away on retirement. In New York, it is hitting more men, who are proportionally more likely to with outside jobs such as construction. That is to say, all of the hard hit groups had long term exposure to pollution.

And again I stress. I view this as my conspiracy theory. Really want to not believe it.

Firefighters have an inordinately high rate of lung cancer, even though comparatively few smoke (as their job comes with physical fitness requirements). This rate of cancer is believed to be tied to regular smoke exposure.

On a lower, slower timeframe I wouldn't be surprised if pollution had the same effect re: weakening the lungs.

Follow up question: are you guys seeing a drop in norovirus, rotovirus and D&V? I would expect social distancing has reduced opportunistic infection rates: no swimming pools, less interactions at large.

I have anecdotal evidence from a hospital saying that wards usually reserved for norovirus and C. difficile at this time of year have not been required this year due to the lack of those infections.

the Swedish state epidemiologist used that measure as a way of measuring the effectiveness of the social distancing and quarantine measures.

apparently, the number of people receiving hospital treatment or dying from yearly virus infections like the flu has dropped significantly.

Do you have any insight into the validity of the Nasopharyngeal covid19 PCR test? I suspect I got a false negative. Get to have a "conversation" with HR tomorrow since I am essential.

For samples with coronavirus particles, sensitivity and specificity are around 100%. (1)

The issue is that many samples do not contain virus. As far as I know, that has not been quantified outside of China. (2)

1. https://www.fda.gov/media/136151/download

2. https://www.mdmag.com/medical-news/comparing-rt-pcr-and-ches...

Single swab tests may be inaccurate, nasal, throat and anal swabbing is usually recommended for good coverage.

The PCR test itself has near 100% sensitivity and specificity when done correctly.

Poor swabbing or contamination can result in false negatives or positives.

Manual PCR testing is known to be quite prone to human errors hence why the all in one cassette kits are quite popular.

I really appreciate this answer, and as a lay person, it helps me put some pieces together in my mind. I suspect there are at least four reasons for testing, and they all benefit from better results, but two of these need good quality unbiased general statistics, and the other two only need high confidence negative/positive results for specific cases:

- Policy makers weighing the strategic pros/cons to orders for the public

- Individuals (either frightened or skeptical) weighing their decisions

- Physicians treating patients

- Quarantining people based on contact tracing

Can surveillance testing not be done using standard laboratory reagents (ie not FDA approved)? (I suppose the rather invasive sample collection protocol might pose a bit of an issue though.)

Some studies are looking into performing tests on sewage to estimate the number of infected people in an area.

There are ways of doing population wide testing that are not very invasive.

Foolish. Short-sighted. Informed policy was, and is, more important than a couple thousand more clinical tests.

If policy had been informed, the number of people saved would have out-weighed the handful who died due to missing out on a test.

Hard to make this decisions in smaller contacts when you are potentially endanger real patients needing care right now Vs future "less dead" estimations.

At this point do we even need to test most of the serious cases. if it walks like a duck, quacks like a duck, coughs like a duck, it's a duck. Didn't China start counting cases based on CT scans at one point too?

China is not forthcoming with the truth, at all.

Health systems deal with scarcity on a daily basis. There's no room for the emotional hand-wringing you're describing. Moving scarce resources from diagnosis patients to studying the population during a pandemic will be one event in a causal chain that results in deaths. It also prevents the outbreak from spreading beyond control.

Several orders of magnitude more people have been killed by uninformed policy than would have been killed by redirecting a portion of tests. What kind of MONSTER chooses for so many more people to die?!?

If you don't test a patient that's positive and put them with other patients and unprotected health personnel you'll infect others and significantly weaken the hospital's ability to treat patients. If you don't test a patient that's negative and put them with covid patients they risk getting infected, taking up an extra ICU spot, and might die. Would you want one of your loved ones in that position?

What would change in public policy with more randomized tests?

A full nationwide shut down without waiting for individual states to shockingly have the same issues as every other state was one possibility.

By comparison many ill people are currently being told to stay home until they have difficulty breathing, making tests have minimal clinical value. Especially as the risk of false negatives are significant.

AFAIK, they are different tests:

* PCR tests to check if you are carrier of the virus right now. That's what you do with patients and personnel.

* Anti-body tests to gauge the number of people having been exposed to the virus in the past.

It takes a couple of days to develop anti-bodies so you don't want to use that one for the first use-case. And you can have anti-bodies without carrying the virus. (Well, it depends on the type of antibodies)

If you know the exposure in the general public, we would better know the real mortality, how far we are with herd-immunity, and if general quarantines do make sense, or it might be sensible to be more selective.

Do you know what's even worse for the hospital's ability to treat patients? If the city continues to operate normally until the number of cases is so large that it's obvious the outbreak is unmanageable even without testing.

Randomized tests would have told policy makes exactly how fast this spreads, even in Western cities. Lockdowns would have happened earlier. The total number of infected would be 1-2 orders of magnitude lower. Hospitals would have been better off.

>Would you want one of your loved ones in that position?

I wouldn't want one of my loved ones to die because my civilization was so short-sighted that it let a disease run rampant. Your appeal to emotion is garbage, and it doesn't even make sense, since all of our chances would be better if we'd known what was going on.

Soooo....

- SARS started out with a similar <4% estimated fatality rate and was then revised upwards to anywhere from 9-15% later.

- COVID-19 is caused by a different strain of the same virus as SARS.

- The CFR of SARS and COVID-19 appear to be very similar, and more notably, appeared by be very similar when we had around 8,000 infections which is where SARS ended. (Similar meaning the CFR hovers between 4% and 20% of closed cases.)

How are we so sure that this is any less deadly than SARS?

The above suggests to me one of the two is likely to be true:

1. There could have been many more undetected cases of SARS than we knew about, indicting an IFR much lower than it's recorded CFR.

2. COVID-19 could actually end up having an IFR that is similar to SARS (~10%).

But of course, I am no epidemiologist, so I assume there's a flaw in my logic.

Did I miss something, or is this pretty much the same disease as SARS but with a higher R0?

There's little to no data that backs up your guess.

Studies on COVID-19 estimate that the true IFR is somewhere between 0.1% to 0.39%...why? Because the more we test, the more we find asymptomatic and mild cases. And we're yet to even do the type of serological testing that would give us such a decisive sample. Yet we're ALREADY seeing data that suggests that the IFR is lower than SARS and asymptomatic cases/transmission are common. [0]

I never read anything about SARS being mild or asymptomatic in the majority of patients...in fact it was the opposite. It was so severely symptomatic that the virus killed itself with natural selection. If you had SARS-COV-11 and exhibited high viral load, you were likely too sick to go spread it. The only serological testing they did on SARS-COV-1 deemed asymptomatic cases to be uncommon [1]. Quite the opposite of COVID-19.

0: https://www.cebm.net/covid-19/global-covid-19-case-fatality-...

1: https://www.ncbi.nlm.nih.gov/pmc/articles/PMC3035549/

This is good news. Thank you.

I am not convinced that this data is conclusive yet.

In two closed environments where everyone was tested, the Diamond Princess and the Washington Choir, the CFR is currently 1.55% and 4.44% respectively, with still cases in critical condition.

Of courses both populations are older, but it seems safe to say that the IFR is much higher than the numbers you quoted for older populations.

What your safe to say claim misses however is that the CFR drops off dramatically below a certain age, and very dramatically, especially if you compare between people in their mid 40s and blow to people in their mid-60s and above. Thus, your own mention of the populations of both cruise ships heavily tending towards elderly passengers (and infected) would almost certainly skew the CFR very strongly upward, and even with this upward skew, it's still so far resting at a fairly moderate 1.44% in the case of the Diamond Princess. As for the Washington Choir, not sure of why its CFR is almost three times higher, perhaps more of the passengers were older than in the DP? Either way a look at all current aggregate data on fatality counts by age from nearly any region of the world you'd like to look at clearly shows that same near drop off a cliff as soon as you reach age groups below 60.

Agree for the most part, especially with the claims with respect to SARS having a higher IFR

However, given the high R0 and still somewhat high IFR for older populations and even a relatively high IFR compared with the flu IFR for younger populations this virus still has the potential to wreak havoc.

What we are seeing is that a disease with a high R0 can be much more deadly than a disease with a lower R0 and higher IFR.

Did we ever do randomized tests due SARS? You linked one, but I would presume more would be needed for a decision?

The main study which backs up his guess is the report from the WHO's trip to Wuhan, where they actually did what was suggested above and mass-test random samples from the population, finding a very low incidence of positives. Personally I would find the WHO on a trip to China as not the most reliable of sources, so take it with an enormous grain of salt. But it's one data point.

Meanwhile, your link in [0] flat-out says we don't know the percent of asymptomatic cases. It is well-documented they exist. I have also seen some papers estimating the percent of cases that are asymptomatic as well under 20%, and as high as the 80-90% range, the difference between them making an enormous difference to the implied mortality rate.

A month ago, people pointed to South Korea's 0.5% fatality rate, a country which tests very aggressively, as an indication that the true fatality rate is much lower. Well, they still test just as aggressively, and now their mortality rate has risen to 1.96%. The disease started spreading in a population that skewed young there, and but now the disease has started to hit older demographics who have a much higher mortality rate. And of course, "asymptomatic" cases can also mean pre-symptomatic cases, as happened with the Diamond Princess, which had an observed 46% of positive cases be asymptomatic...only to have this rate fall to ~17% later.

Yeah basically the extreme correlation of mortality with age makes it hard to state a mortality rate. Small changes in the population age distribution create big changes in mortality.

I've posted this in an earlier thread but I don't think that low estimates make sense. The virus killed 0.44% of total population of Bergamo

4500/1112187

https://translate.google.com/translate?hl=en&sl=it&u=https:/....

https://en.wikipedia.org/wiki/Province_of_Bergamo

EDIT: and there are harder hit communities: 400/40000 https://www.ecodibergamo.it/stories/valle-brembana/il-grido-...

take the age bias into account. 45 for Bergamo and 29.6 for the world.

This first link is utter nonsense. It bases on things like 6 people having died on Diamond Princess. That number is now 12, and growing.

I’ve seen these reports but I don’t understand the math. The true simple CFR meeting actual deaths as reported over actual cases as reported should be adjusted for the course of disease which would make it in places like Italy well over 10% and in other places perhaps 4-6. % or more and we know China’s data, the only place with a major outbreak where it’s really run its course, is bullshit. How do you get to 0.1%? Maybe I’m missing something here but you’d have to believe there were probably 50-100 times the Number of reported cases that were actual cases which would mean the US would have to have had 50-100 million people already have contracted the disease but 98-99 pct of them had no idea. I’m not seeing that as a passes the smell test possibility. But again maybe I missing something here.

> Did I miss something, or is this pretty much the same disease as SARS but with a higher R0?

Yes, SARS-1 and SARS-2 (Covid-19) are more similar than not. But debating IFR, CFR or R0 with the testing data we have is pointless.

You can learn a lot about the nature of SARS-1 by reading about the 2002 warzone in Toronto hospitals, which wiped out whole ICU teams. Sounds just like corona today in Italy or NY.

I think there are also several issues that are causing the reported CFR to be understated, that counteract to some degree the “denominator problem” from undiagnosed mild cases. The one that I think is the most important is the simple understanding of cohort analysis which I just don’t ever see explained or included any calculations. The number of cases is rising exponentially, and there’s a six week course roughly to the disease, so the number of deaths at any point in time is six weeks behind the number of diagnosed cases at any point in time and the number of cases is increasing exponentially, so this seems like a big error. I have to say I’m truly surprised by how we are this far into this thing and it doesn’t seem like anybody has even done some of the basic mathematical maneuvering you would need to do to understand this or at least they haven’t done that and explained it very publicly.

Did we miss the chance of going deeper on the Diamond Princess? Maybe a full, in depth check on everyone in the ship would be the ideal approach.

Everyone on the Diamond Princess was checked. That's why it keeps being cited as a closed population with statistically valid sampling (100%). It's the re-projections from that ship onto the age range of the general population that yields low fatality rates comparable to flu.

If I was on the diamond princess during this outbreak and in a balcony room, that balcony door would have been wide open the entire time.

> . It's the re-projections from that ship onto the age range of the general population that yields low fatality rates comparable to flu.

Using old data that misses fully half of the deaths so far... (and there are still people in serious condition).

For SARS resolved CFR was a more accurate estimator at all points in time: https://academic.oup.com/aje/article/162/5/479/82647

We clearly saw this coming. In fact, in 2006 President Bush signed “National Strategy For Pandemic Influenza, Implementation Plan”, Homeland Security Council, May 2006. As an epidemiologist, what do you think went wrong? How come we are caught off guard? Here is a comparison of our policies and fight against the Spanish flu and Covid-19. Remarkable how stay-at-home and social distancing policies worked then too. https://thepeel.news/stay-at-home-orders-worked-a-century-ag...

My grandparents were dating at the time of the Spanish flu. They lived in a little town in the middle of Iowa and practiced social distancing by having their dates at my Grandmother's home, doing quilting, instead of socializing with others.

It strikes me that R0 could be a very rubbery figure because surely it's dependent on the social setting. An R0 of kids crawling over each other in a kindergarten must be different to an retired estate where most people spend their time in their own house.

The final R0 is I guess is how all those values average out in a society. Is it normal for all the different societies around the planet to have the same values?

R0 is the expected number of cases generated by one infected case in a susceptible population. Usually derived through mathematical models, it is not a measure / rate of transmittablity or infectivity of the pathogen, but an indicator used to predict severity: R0 <= 1 implies the disease is endemic, otherwise; an epidemic.

R0 is dependent on parameters taken into consideration and the model itself. R0 makes sense only in the context laid out by the model.

> An R0 of kids crawling over each other in a kindergarten must be different to an retired estate where most people spend their time in their own house.

Ref: https://news.ycombinator.com/item?id=22818413

> The final R0 is I guess is how all those values average out in a society.

See: https://en.m.wikipedia.org/wiki/Basic_reproduction_number#No...

Study from the first community transmission center in Germany. Brand new. https://www.land.nrw/sites/default/files/asset/document/zwis...

Preliminary results and conclusions of the COVID-19 Case Cluster Study (Gangelt municipality)

Prof. Dr. Hendrik Streeck (Institute for Virology)

Prof. Dr. Gunther Hartmann (Institute for Clinical Chemistry and Clinical Pharmacology, Speaker of the Cluster of Excellence ImmunoSensation2)

Prof. Dr. Martin Exner (Institute for Hygiene and Public Health)

Prof. Dr. Matthias Schmid (Institute for Medical Biometry, Informatics and Epidemiology)

University Hospital Bonn, Bonn, 9 April 2020

Background: The municipality of Gangelt is one of the places in Germany most affected by COVID19 . It is assumed that the infection is due to a carnival session on 15 February 2020, as several people tested positive for SARSCoV2 in the aftermath of this session. The carnival session and the outbreak of the session are currently being investigated in more detail. A representative sample was taken from the community

Gangelt (12,529 inhabitants) in the Heinsberg district. The World Health Organization (WHO) recommends a protocol in which, depending on the expected prevalence, 100 to 300 households are randomly examined. This random sample was coordinated with Prof. Manfred Güllner (Forsa) to ensure its representativeness.

Aim: The aim of the study is to determine the status of SARS-CoV2 infections (percentage of all infected persons) in the community of Gangelt, which have been and are still occurring. In addition, the status of the current SARS-CoV2 immunity shall be determined.

Procedure: A serial letter was sent to about 600 households. In total, about 1000 inhabitants from about 400 households took part in the study. Questionnaires were collected, throat swabs taken and blood tested for the presence of antibodies (IgG, IgA). The interim results and conclusions of approx. 500 persons are included in this first evaluation.

Preliminary result: An existing immunity of approx. 14% (antiSARS-CoV2 IgG positive, specificity of the method >.99 %) was determined. About 2% of the persons had a current SARS-CoV-2 infection detected by PCR method. The infection rate (current infection or already been through) was about 15 % in total. The case fatality rate in relation to the total number of infected persons in the community of Gangelt is approx. 0.37 % with the preliminary data from this study. The lethality rate currently calculated in Germany by Johns-Hopkins University is 1.98 %, which is 5 times higher. The mortality in relation to the total population in Gangelt is currently 0.15 %.

Preliminary conclusion: The lethality calculated by Johns-Hopkins University is 5 times higher than in this study in Gangelt, which is explained by the different reference size of the infected persons. In Gangelt, this study covers all infected persons in the sample, including those with asymptomatic and mild courses. In Gangelt, the proportion of the population that has already developed immunity to SARS-CoV-2 is about 15%. This means that 15% of the population in Gangelt can no longer become infected with SARS-CoV-2, and the process has already begun until herd immunity is achieved. This 15% of the population reduces the speed (net reproduction rate R in epidemiological models) of a further spread of SARS-CoV-2 accordingly.

By adhering to strict hygiene measures, it can be expected that the virus concentration in a person infected can be reduced to such an extent that the severity of the disease is reduced, while at the same time immunity is developed. These favourable conditions are not given in the case of an exceptional outbreak event (superspreading event, e.g. carnival session, après-ski bar Ischgl). With hygiene measures, favourable effects with regard to total mortality can be expected.

We therefore expressly recommend implementing the proposed four-phase strategy of the German Society for Hospital Hygiene (DGKH). This strategy provides for the following model:

Phase 1: Social quarantine with the aim of containing and slowing down the pandemic and avoiding overloading critical supply structures, especially the Health care system

Phase 2: Beginning of the withdrawal of quarantine while ensuring hygienic conditions and behaviour.

Phase 3: Lifting of the quarantine while maintaining the hygienic conditions

Phase 4: State of public life as before the COVID-19 pandemic (status quo ante).

(Statement of the DGKH can be found here:

https://www.krankenhaushygiene.de/ccUpload/upload/files/2020... adug_Lageeinschaetzung.pdf)

Note: These results are preliminary. The final results of the study will be published and presented to the public as soon as they are available.

"By adhering to strict hygiene measures, it can be expected that the virus concentration in a person infected can be reduced to such an extent that the severity of the disease is reduced, while at the same time immunity is developed. These favourable conditions are not given in the case of an exceptional outbreak event (superspreading event, e.g. carnival session, après-ski bar Ischgl). With hygiene measures, favourable effects with regard to total mortality can be expected."

This is the most interesting part of this and should be discussed more. There seems to be a lot of evidence that viral load is a factor. If social distancing and proper hygiene doesn't just lead to less infections but to more mild infections that's a big dea.

My understanding is that viral load is always a factor for every virus. Laboratories don't use soap and water, because they want to kill 100% of the viral particles, not 99.9%. They work with samples that have extremely high concentration of viral load.

(Source: lecture by prof. Krzysztof Pyrć)

That last link is broken. This should be the correct one: https://www.krankenhaushygiene.de/ccUpload/upload/files/2020...

Santa Clara county did surveillance testing which likely saved California https://www.cdc.gov/mmwr/volumes/69/wr/mm6914e3.htm despite the shortage of test kits. Just imagine if the nation had sufficient tests available.

"During March 5–14, among patients with respiratory symptoms evaluated at one of four Santa Clara County urgent care centers serving as sentinel surveillance sites, 23% had positive test results for influenza. Among a subset of patients with negative test results for influenza, 11% had positive test results for COVID-19."

Seems to me like Santa Clara was just testing sink people to me

It continues:

At the time the investigation began, testing guidance recommended focusing on persons with clinical findings of lower respiratory illness and travel to an affected area or an epidemiologic link to a laboratory-confirmed COVID-19 case, or on persons hospitalized for severe respiratory disease and no alternative diagnosis (1).

At the time, the standard was to test people with travel history to affected areas or when they were hospitalized for unexplained respiratory illness.

Compare that to testing people that went to urgent care because they felt sick. There's a difference.

Still, there's no reason to say that any of these selection criteria result in sample sizes big enough to be statistically relevant against larger populations. Also, testing "people with travel history to affected areas" does not amount at all to being a representative segment.

Santa Clara's results are basically only useful for Santa Clara. _Maybe_ the Bay Area in addition, but not really any farther.

So, we DO need surveillance testing, on a much wider scale, across the US and globally to be frank.

The poster up thread is claiming that the surveillance testing in Santa Clara lead California to act as if there was community transmission.

That's super useful information when most testing is using restrictive criteria like travel to affected areas.

They aren't arguing that it was enough, they are pointing at it as an example of what broader testing can help do, if only it existed.

So I don't really understand your refutational tone ("Still", "relevant", "only useful", "DO", etc). Is one of us misunderstanding the flow of the thread?

You are getting downvoted because you're arguing for global surveillance, which is not a popular topic on this board. And that doesn't include some real fears of actual dictator-ish moves, with COVID as a justification.

i'm not scared that hungary might use covid-19 as a justification to abandon democracy in exchange for dictatorship. i see they did. an observation is a terrible thing.

Test a group known to be 10x more likely get infected and you can get meaningful results with a much smaller sample size. 'sentinel surveillance sites' to try and catch the wave before it sweeps through the general population.

Testing people without symptoms would likely result in getting nothing given the limited sample size. The point is that the sampling was done with less bias, without the preconceived notion of who may actually have it (at least among fever patients).

This, unless we do that there is no way to establish true IFR so we wont know until its in hindsight how deadly the covid is? I dont know why this is not the priority 1 for the CDC.

on the similar lines, Germany is doing something clever, they are mixing samples to do a kind of binary search with covid tests so they can clear a large number of people with few tests. we should do that without waiting for antibody tests etc. it should be noted that this approach would only work if the infected sample set is very small so the window of doing this is going to close soon.

> I dont know why this is not the priority 1 for the CDC.

What actions would we do differently based on the result? Not knowing is frustrating for everyone, but we should save our resources for stuff that's actually going to improve outcomes.

IFR is a really important peice of information as it effects the overall management of outbreak. consider these scenarios: A) IFR is really low and most people dont seen anything past slight discomfort, so given the number of 'cases' being treated now it would mean a much larger portion of populus is already infected. translation we can reach herd immunity faster & more safely. B) IFR is high (closer to CFR), which means the current case count is more or less in line with actual infected people. which means we really need to focus on lockdown and slow the growth so we can manage the hospital capacity.

as you can see both of these scenarios have different optimal responses but we cannot decide unless we get this info. so far we only have a few biased sample sets (from us pov) diamond cruise & skorea which dont shed much light on our situation.

Testing should be saved for two purposes:

1) to guide care in a hospital setting. 2) to determine the current status of the population.

Randomized testing would give us a view into where the virus is spreading and allow us to allocate resources effectively. It would also give us a view into what percentage of a given population is a carrier, or has already recovered. All of these data would be greatly helpful. What isn’t helpful is using tests on people who aren’t either part of a study or under active care.

Literally every epidemic model being used to justify lockdowns depends on IFR as a critical parameter, and the best we can do right now is guess at the value.

If the IFR is 1/10th (quite possible!) or 1/50th (less likely, but still possible) of the current estimate, it completely changes the calculus for locking things down. A fifty fold under-estimate of infections would mean that approximately half of NYC has already been infected, for example, and that we're not far from herd immunity.

We just don't know.

> Germany is doing something clever, they are mixing samples to do a kind of binary search...

Do you have a link for that? I can find some papers suggesting you can detect one positive sample mixed with 31 negative, but nothing about that being applied.

Just googled this and I found a discussion here: https://twitter.com/koeppelmann/status/1245037948099330048

"Testing capacity in Germany will be increased by up to factor 10 to up to 400,000 a day (!) by doing pooled testing. E.g mix 16 samples and if negative - all are negative, otherwise binary search for the positive(s). Could of course be used worldwide."

He quotes a tweet with a link to this study: https://idw-online.de/de/news743899

There are a bunch of serological surveys going on right now in the US: https://www.sciencemag.org/news/2020/04/unprecedented-nation...

Both Munich and Prague have started such studies with several thousand random people in each. That should give us a much better estimate for the number of "silent" cases.

Random testing here in Iceland over the last week showed 0.6% infected.

https://grapevine.is/news/2020/04/06/random-sampling-reveals...

This is a useful albeit very small data point, because Iceland received its cases from both Europe (alpine ski cluster) and the US. So it puts a very loose upper bound on the incidence in the US up to the point travel to Iceland was effectively halted.

So far only PCR tests are available at scale. Those PCR tests are very sensitive (e.g when detecting traces of long broken virus on surfaces) but even they are only reliably positive for a few days into the infecting. Apparently, the immune system wins the battle for the upper throat quite easily, even in patients who later die from pneumonia. If you did a random sample study with a test that only identifies a positive for a short time you'd still have a huge uncertainty from not knowing reliable averages about how long someone infected "lights up" in the test. There just isn't much information upside from background sampling PCR, particularly when serological tests (those than detect who already had the virus) will soon be available. Those are not useful at all to combat spread anyways, so they can be used for background testing without competing for more immediate use cases.

Dr. Fauci addressed that the other day when asked about when schools might reopen. He hedged a lot, but basically said he thought it could happen for the Fall. At the same time he was talking about that as the time frame for wide-spread anti-body testing that would show just how far the virus had penetrated. (And that such testing and continued contact tracing etc. would be what facilitated cautious steps towards reopening society)

There was a town in Italy, roughly 10k population, where they tested every single person regardless of symptoms. The result highlighted how many asymptomatic people tested positive. AFAIK the test was not repeated so there was no data on evolution / spread. But it was still an eye opener.

(My apologies for not citing the article, I had read it in the paper edition of the NY Times, and failed to find it online.)

I remember reading this story in NYT also. Are you thinking about the town of Vò?

If so, here is a related article about how they tested everyone there: https://www.theguardian.com/world/2020/mar/18/scientists-say...

Recently in Argentina there were news of a case were 400 people in a boat were exposed, quarantined and then tested. From official government sources, of all that tested positive, 70% were asymptomatic.

Edit: I rechecked, the numbers are right but apparently this has been the first test. It hasn't passed enought time to see if those that tested positive, but were asymptomatic, later on develop symptoms.

This doctor from Stanford agrees with you and is trying to run this type of test: https://www.youtube.com/watch?v=-UO3Wd5urg0

I think Iceland is aiming for full population test

At the moment, that data would not make any significant difference on the actual decisions that are made today. Or can you think of anything you would change with the knowledge that 20% of the population have been infected, versus 2%? As long as thousands die each day, there really is no alternative to the current strategy.

It would possibly help planning for the next phase. But mostly it would feed our curiosity.

Contrast to using limited tests to find the greatest number of infected people: every additional person knowing without a doubt that they are infected is one more person doing their best not to spread it further.

It’s wasteful to test randomly until we are able to test high risk groups every day or so. Cashiers, for example are at high risk of infection, and ha e possibly hundreds of contacts per day.

For population-randomized testing, antibody tests should be far superior anyway. They allow detection of. It just active cases but also past infections.

Sure, let's just keep the economy closed, who cares if the death rate is 10%, 1% or 0.1% (seasonal flu), what difference does it make right?

Great job not getting it. Hope your unique skills are transferable to viruses.

As said by someone else, tests are in short supply. So other ways are being tested, like testing wastewater[0].

[0]: https://www.sciencedaily.com/releases/2020/03/200331092713.h...

Not an expert, but this is something I've been wondering about too. But then I had another thought, What happens if people refuse the random test. Even if those who refuse are only a small amount of people who get drawn in the covid testing lotto, that likely introduces a serious compounding variable.

So then the question is, do you go out into the general population with law enforcement, and force people who were drawn in the testing lotto to be tested at gunpoint if they don't comply? I think the answer is that we don't have the will to do that. And perhaps even if this would be incredibly valuable data, it would be morally wrong to gather it in the way we'd have to gather it to have a truly random sample without such a confounding variable.

I am not a doctor, not an epidemiologist. Just some random thoughts.

Just pay them, e.g. 500USD each, or however much is enough to reduce the number of refusals to a negligible proportion. It's cheap compared to the cost of making incorrect decisions because you don't know how many people have the virus.

That's a perverse incentive. You'd have to pay everybody $500 each, not just the refusals, since everyone is going to refuse in order to gain $500.

You can even refuse 500USD in hope to get 1000USD...

60-80% of people routinely hang up when being called for political surveys, yet surveys still manage to get within 2 or 3% of election results. This is a non-issue.

Nope. They’re often off by 7-11%. As evidenced in many elections.

Refusals should only matter if they correlated with infection (or lack thereof). I don't see why that would be the case here though.

Really? You don’t see any way that someone who wouldn’t be willing to let the government conduct a test on them wouldn’t also be less likely to follow CDC guidelines on social distancing, etc?

90% of people will comply. Let's say it's only 50% though. Those people are no more or less likely to be infected. So it's 100% perfectly fine not to test everyone if the goal is to determine the number infected overall.

If looking to go beyond, hey let's get testing capacity to the point where those who are sick and desperately want to be tested can be tested, then we can attack the 1984 fantasy.

"But but what if I can't force those to be tested who don't want it" is an absurd and insane concern given that only a tiny minority of people can be tested who desperately want to.

Once we have testing capacity then those who want to force testing should go out and try to force people to be tested and happily accept whatever happens to them.

You are reading a lot of stuff into my comment that is not there.

1. I’m all for more testing, and would gladly be tested. 2. You really don’t think the cohort of people who would refuse testing wouldn’t also be less likely to follow CDC guidelines? This increasing the chance they are infected. 3. All I was trying to say, was that random testing will be confounded by things like people refusing. Perhaps it gets us close enough. Perhaps people who refuse are more likely to be infected. Or maybe less likely, because they are loners who don’t go out thanks to their doomsday prepping or something. I think it is worth being aware of.

You seem to be incredibly confident that the cohort who would refuse testing will have the exact same infection rate as everyone else. Care to share why you are so certain of that?

You might check out https://seattleflu.org/

I don’t know if they’re the only ones doing random testing, but my understanding is that they were able to put pressure on the CDC when no one in the US was legally allowed to test other than the CDC. they were somehow able to circumvent testing restrictions by claiming research status since they were in the middle of a two year flu study.

In any event, I suspect the cost and difficulty of good randomized is why they needed such good funding for only a 2 year study.

It's in South Korea where there is an abundance of tests. Where tests or people to administrate them are scarce, we have to prioritize those with symptoms.

There is an argument to me made for the exact opposite. Those who are already sick will be treated according to their symptoms, a diagnosis has little bearing on the treatment they will receive and therefore low value. Meanwhile, proper random sampling of the population will give more accurate prevalence numbers and trends, allowing for much better epidemic response.

South Korea (and Singapore and HK etc.) had a specific policy of contact tracing instead of random sampling, which works if you start it early enough and can actually keep ahead of suspected contact spread.

Sick people should be tested when they require medical intervention as it guides their care. It would be a shame to assume COVID when it might be something treatable.

I'm sure the decision tree is complex, and there are branches where what you say is absolutely true. However, given the testing kit bottleneck, if identification of "something treatable" is crucial, it might be better to test for those treatable diseases that are likely given presenting symptoms.

In the extreme case, if you require a ventilator right now or you'll die, nobody needs to wait on verification of the specific viral cause to begin treatment.

"Those who are already sick will be treated according to their symptoms, a diagnosis has little bearing on the treatment they will receive and therefore low value."

It is not low value for the healthcare providers, though, who can spare measures against getting infected to those who actually have COVID and thus both save time and scarce resources.

The cost of false negatives would be too high. Hospitals have to treat all respiratory cases as potential COVID right now, even if initial tests are negative.

In Alberta we are only testing pretty sick people until this week, but are high in the global rankings for tests per capita. About 2% of cases have been positive. Widespread, random testing would not make as good a use of resources, since the chemicals to make the tests are becoming a little scarce until Chemical manufacturers catch up

> the chemicals to make the tests are becoming a little scarce until Chemical manufacturers catch up

To the best of my knowledge this isn't true. Only specific product lines by specific manufacturers are approved for clinical use by most governments. As far as I know, the reagents used for RNA extraction and PCR are cheap and readily available.

This is probably what you're looking for: https://www.eurosurveillance.org/content/10.2807/1560-7917.E...

I believe Wuhan government eventually went door-to-door to test the population exhaustively.

I also don’t get it - shouldn’t work better testing a random population sample and interpolate the numbers?

Reply on news.ycombinator.com