Is the COVID denominator inaccurate?

[elburrito]

https://www.nejm.org/doi/pdf/10.1056/NEJMe2002387?articleTools=true

If you assume there are ten times more subclinical cases than tested and confirmed, are the case fatality rate, and the rate of 15-20% hospitalization for severe or critical illness, not significantly lower?

I was a little surprised by this statement in The NEJM editorial, “If one assumes that the number of asymptomatic or minimally symptomatic cases is several times as high as the number of reported cases, the case fatality rate may be considerably less than 1%. This suggests that the overall clinical consequences of Covid-19 may ultimately be more akin to those of a severe seasonal influenza (which has a case fatality rate of approximately 0.1%) or a pandemic influenza (similar to those in 1957 and 1968) rather than a disease similar to SARS or MERS, which have had case fatality rates of 9 to 10% and 36%, respectively.”

---

Members don't see this ad.

 

[deleted480308]

No one knows

 

[gbwillner]

I've been thinking the same... mortality rate across all carriers can't be defined unless you screen for said carriers. We are not doing this. Germany and Worst Korea are- they are the only ones who can really estimate accurate death rates, which are likely to be around .2-.5%. Likely stopping the virus is not possible at this point. I expect 30% of health care workers are already exposed.

 

[elburrito]

I've been thinking the same... mortality rate across all carriers can't be defined unless you screen for said carriers. We are not doing this. Germany and Worst Korea are- they are the only ones who can really estimate accurate death rates, which are likely to be around .2-.5%. Likely stopping the virus is not possible at this point. I expect 30% of health care workers are already exposed.

assume 0.5% mortality and conservatively assume only 50 million people (same as H1N1) in the country’s population (340 million) get infected. That still equals 250,000 dead, and assuming 15% hospitalization for 50 million, that equals 7.5 million hospitalizations.

 

[gbwillner]

S'more bad news... I'm reviewing the COVID-19 data and the following facts come out:

1. The death and hospitalization rates are growing exponentially
2. The doubling time of deaths is 3.35 days since 3/11 and seems to be getting faster, suggesting this thing is very contagious
3. At this time we have tested just over one million patients in the US. The positive test rate, however, has steadily increased over time. The small data sets meant the rate jumped a bit from 10-14% between 3/11 to 3/15; but since then there has been steady and incremental increases from 3/16 until now- from roughly 10% to 17% yesterday.

The meat trucks though are not for the dead from this flu hoax. They are for Soylent Green.

 

[elburrito]

https://www.thelancet.com/journals/laninf/article/PIIS1473-3099(20)30243-7/fulltext

overall infection fatality rate 0.66%.
Assume 50,000,000 are infected = 330,000 dead

 

[AZpath]

S'more bad news... I'm reviewing the COVID-19 data and the following facts come out:



The meat trucks though are not for the dead from this flu hoax. They are for Soylent Green.

You are showing you age. Great movie.

 

[D1b()4]

I am getting more and more skeptical regarding this theory of highly underestimated denominator. I am sure there are some asymptomatic/mildly symptomatic cases, and many of those do not get tested. However, we have already tested more than a million people. If the theory of highly underestimated denominator were true, I would expect us catching more and more mild cases as the testing availability increases resulting in decreasing overall fatality rate. Unfortunately, the fatality rate has been going up and will probably keep going up for some time since it usually takes more than a week to die from COVID-19, and we are still in the early stages of the epidemic. Also, strain on the healthcare system will also lead to increased fatality rate.

 

[gbwillner]

I am getting more and more skeptical regarding this theory of highly underestimated denominator. I am sure there are some asymptomatic/mildly symptomatic cases, and many of those do not get tested. However, we have already tested more than a million people. If the theory of highly underestimated denominator were true, I would expect us catching more and more mild cases as the testing availability increases resulting in decreasing overall fatality rate. Unfortunately, the fatality rate has been going up and will probably keep going up for some time since it usually takes more than a week to die from COVID-19, and we are still in the early stages of the epidemic. Also, strain on the healthcare system will also lead to increased fatality rate.

Counter- we are not really screening healthy individuals. Furthermore, the positive testing rate is steadily going up.

 

[D1b()4]

Counter- we are not really screening healthy individuals. Furthermore, the positive testing rate is steadily going up.

I made this argument to emphasize that the fatality rate of COVID-19 is several times higher than even severe seasonal influenza, and we also do not screen asymptomatic healthy individuals for influenza. If the positive testing rate is going up because we are detecting more and more mild/asymptomatic cases, the fatality rate expressed as percentage should be stabilizing or decreasing. Even if 50% of the affected individuals are asymptomatic, like some studies suggest, COVID-19 fatality rate is not anywhere near the fatality rate of severe influenza.

Here is some additional information to support my argument. Based on the Diamond Princess experience (Field Briefing: Diamond Princess COVID-19 Cases, 20 Feb Update), where they performed testing irrespective of the presence of symptoms, 51% of cases were asymptomatic at the time of diagnosis. However, it is not clear whether they had been symptomatic prior to testing or developed symptoms after they were isolated due to positive COVID-19 test. Per Chinese data, approximately 75% of asymptomatic cases eventually become symptomatic. Overall, I do not understand how COVID-19 fatality rate can approach that of severe seasonal influenza unless 90% of the COVID-19 infections are asymptomatic.

 

[cmz]

Counter- we are not really screening healthy individuals. Furthermore, the positive testing rate is steadily going up.

Should we be screening healthy individuals? There aren't enough tests in my area and we need to test the sick people first.

 

[bauber]

Iceland is testing asymptomatic people and up to 50% of those are positive, smallish sample size though, 6% of their population.

Iceland lab's testing suggests 50% of coronavirus cases have no symptoms | CNN

There's a lot about Iceland that other countries could envy: Its spectacular natural surroundings, its place among the world's happiest countries, and, now, its large-scale testing for the novel coronavirus, which could influence how the world understands the outbreak.

www.cnn.com

 

[Allegri]

OK but if you're screening everyone healthy you've got some # of false positives for a relatively low prevalence condition. What are the SN/SP of the test? I can never find this reported anywhere.

 

[Med Director New England]

Iceland is testing asymptomatic people and up to 50% of those are positive, smallish sample size though, 6% of their population.

Iceland lab's testing suggests 50% of coronavirus cases have no symptoms | CNN

There's a lot about Iceland that other countries could envy: Its spectacular natural surroundings, its place among the world's happiest countries, and, now, its large-scale testing for the novel coronavirus, which could influence how the world understands the outbreak.

www.cnn.com

The positive rate is NOT 50% in asymptomatic, the article states ~ 50% of all positive tests are asymptomatic. Big difference...

Says the overall positive rate is around 1%.

 

[bauber]

Thanks, that's what I meant. My bad on the phrasing.

 

[D1b()4]

Iceland is testing asymptomatic people and up to 50% of those are positive, smallish sample size though, 6% of their population.

Iceland lab's testing suggests 50% of coronavirus cases have no symptoms | CNN

There's a lot about Iceland that other countries could envy: Its spectacular natural surroundings, its place among the world's happiest countries, and, now, its large-scale testing for the novel coronavirus, which could influence how the world understands the outbreak.

www.cnn.com

I think 6% is an excellent representative sample of the population.

 

[gbwillner]

OK but if you're screening everyone healthy you've got some # of false positives for a relatively low prevalence condition. What are the SN/SP of the test? I can never find this reported anywhere.

These are highly variable. There are studies focusing on collection method. The best are sputum and nasopharyngeal swab- neither cracked 85% for sensitivity....

 

[D1b()4]

These are highly variable. There are studies focusing on collection method. The best are sputum and nasopharyngeal swab- neither cracked 85% for sensitivity....

Chinese kits are the best with 30% sensitivity.

 

[Med Director New England]

OK but if you're screening everyone healthy you've got some # of false positives for a relatively low prevalence condition. What are the SN/SP of the test? I can never find this reported anywhere.

the performance parameters of the various platforms testing for Covid (sensitively, specificity, NPV,PPV) are not known. If you ask any of the big labs doing the testing they can’t tell you and will tell you the platforms haven’t gone thru the usual analytical and clinical validations to make these claims. The EUA really only required dilution studies of known positive samples to demonstrate limits of detection. The specificity can be inferred b/c the primer targets are supposedly Covid specific and should in theory should not cross react with other Coronaviruses, but again these studies have not been done.

Which sample type has the highest sensitivity (OP, NP, lower respiratory specimen, sputum) is debatable. I am seeing examples of patients with multiple types sent at one time having only one of two or 2 of 3 come back positive. Sometimes the NP is negative while the OP is positive or the reverse or a specimen from the upper respiratory tree (OP or NP) is positive but the one from the lower respiratory tree is negative. Everyone seems to have settled on NP being preferred.

Lower respiratory specimens and sputums also initially were not being tested, but now are acceptable specimen types. Lower tree sample negativity also doesn’t exclude Covid disease only involving the upper tree. Sputum has the problem of proper specimen procurement it may just be spit and be FN for this reason.

My pulmonologist and ICU friends also tell me they are shying away from even doing bronchoscopy at all in covid patients due to risk of spread of virus during procedure. They also don’t see much value add in sending lower respiratory specimens at all if an OP or an NP sample is positive - clinical picture dictates treatment once Covid is diagnosed.

 

[deleted314957]

Wow, a pulmonologist shying away from a bronch. Never saw it in my day!


Sent from my iPad using Tapatalk

 

[elburrito]

Is there any basis for current mainstream media statements suggesting that we are looking at an "Apex" in April?

I'm concerned the govt is just feathering the nest for plausible deniability. That is, govt. was suboptimal in its initial phase pandemic response and fell behind the exponential growth curve. Now, compared to S. Korea, we are playing catchup and "struggling to control" this outbreak.

If the virus has doubling time in the population of 3 days, one month from now, 2 to 10th power is 1024 x current number of infections. My suspicion: an April "Apex" was never on deck. "Apex" is only a rhetorical device intended to shift the narrative away from government failure to respond early on, and when there is no curve flattening after said "Apex", the govt. will claim it is due to failure in social distancing.

 

[Guest8]

Is there any basis for current mainstream media statements suggesting that we are looking at an "Apex" in April?

I'm concerned the govt is just feathering the nest for plausible deniability. That is, govt. was suboptimal in its initial phase pandemic response and fell behind the exponential growth curve. Now, compared to S. Korea, we are playing catchup and "struggling to control" this outbreak.

If the virus has doubling time in the population of 3 days, one month from now, 2 to 10th power is 1024 x current number of infections. My suspicion: an April "Apex" was never on deck. "Apex" is only a rhetorical device intended to shift the narrative away from government failure to respond early on, and when there is no curve flattening after said "Apex", the govt. will claim it is due to failure in social distancing.

I would wager only current hotspots will hit their apex this month. Many states/counties haven't even begun to be hit by this yet, so no way are they near their apex or plateau - they'll just be starting their exponential growth phase soon. Especially those that have been resistant to lockdowns.

 

[chooks]

Sn/Sp data is indeed hard to come by. I have seen one IFU on a serological assay saying their IgG test was 100% Sn/100% specific. I do not believe that for a second. Most do not mention it at all. I have seen one tech briefing from a manufacturer that did comparisons with patient studies in Italy with very sensitive and specific results. But the EUA is a regulatory wild west - platforms are neither waived, nor moderate, nor high complexity, leaving things up in the air in terms of deciding what is the right course for required validation/verification activities from what I can tell.

We did a "mini-validation" on one platform and saw 80% Sn/100% Sp. The other side of the coin is knowing what the prevalence is so you can understand what the actual PPV and NPV. If you are trying to rule out patients so you can stop using PPE, the 80% Sn may not cut it based on your local prevalence (depending on how much risk you want to take with false negatives). The public (and clinicians) want whatever test they can get their hands on -- understandable in the situation we are in -- but little discussion around the inherent limitations of testing, other than the disclaimers at the bottom of reports that "these results must be interpreted in the overall clinical picture".

 

[yaah]

I think part of the problem is that media and many out there are using the "apex" as some sort of critical inflection point - apex in most of the data typically means that the increase in infections is going down. So if you have an increasing number of infections up to the apex, that simply means that you have now encountered about 50% of the infections you are going to see in that outbreak. Which also means that half of the new infections/deaths are yet to come.

 

[elburrito]

Why a study showing that covid-19 is everywhere is good news

If millions of people were infected weeks ago without dying, the virus must be less deadly than official data suggest

www.economist.com

 

[elburrito]

The argument in the economist article above relies on an inference that a significant proportion of ILI may actually represent COVID-19, and if so, the prevalence is much higher than previously thought, and the infection-mortality rate (not just the case-fatality rate) may be significantly lower than projected in the various models that have prompted our global shutdown.

Occam's razor: do not multiply entities beyond necessity. Is their inference reasonable?

Cuomo stated that the worst may be behind us. Has social distancing really flattened the curve or was this all a big hoax and much ado about nothing?

The infection rates, moratlity, and prevalence in well studied cohorts from Wuhan and Diamond Princess cruise were used to inform the models that gave us case fatality rates estimated as 2-3%+, which when coupled with exponential growth, may result in devastating mortality rates far beyond seasonal influenza.

 

[yaah]

I mean, I'm not sure most of the projections are that inaccurate. Hospitals in the hardest hit areas are beyond 100% of capacity. If they don't get to over 1000% of capacity does that prove it was "much do about nothing?" Not at all. The virus is surely less deadly than the 10% fatality rate in Italy or even the 4% or whatever it is in parts of the US. But that doesn't mean there aren't tens of thousands of desperately sick people, nor does it mean that social distancing did not significantly decrease that.

I think there's a lot of lack of perspective as to actual numbers. There are roughly 325,000,000 americans. A 0.1% fatality rate if everyone is infected is 325,000 deaths. I don't think there's a lot of reason to discount the models that showed without social distancing measures that up to 70% of the country could be infected within a few month period, which is massive strain on the healthcare industry even if the fatality rate is only 0.1%.

 

[elburrito]

GOP Lawmaker Goes Full Death Cult: 'Always Choose' Economy Over Lives

Rep. Trey Hollingsworth says lawmakers have to "put on our big boy and big girl pants" and end the shutdowns.

www.huffpost.com

 

[D1b()4]

The virus is surely less deadly than the 10% fatality rate in Italy or even the 4% or whatever it is in parts of the US.

I am not so sure about that. Once everything settles (there is a 10-14 day delay between diagnosis and death/recovery), I suspect our fatality rate will be somewhere between 8-10%. Even if 50% of the cases are subclinical, this still translates into 4-5% mortality rate.

 

[twospadz]

I would be careful using Diamond Back Princess as a model to extrapolate the percentage of people that would die if infected. Although we don't know the age of everyone on that cruise ship, we do know the average age of someone who takes international cruise vacations is 46. The average age of an American is 38. Therefore, you are likely dealing with a older sample and we do know that age plays a factor in mortality with COVID.

Also, I wouldn't make assumptions that 100% of the population would be infected with COVID. COVID has a basic reproduction rate of 3; which means that it will take 70% of the population getting immune to obtain herd immunity which would affect our calculations.

 

[Med Director New England]

13% of consecutive women presenting for Routine delivery at NY Presbyterian Columbia hospital were found to be Covid + when universally screened for Covid on admission to L&D.

None would have otherwise been tested for Covid and only a few had any any symptoms at all.

nyc area is a hotbed but 13% is a lot. The denominator could be huge.

 

[gbwillner]

13% of consecutive women presenting for Routine delivery at NY Presbyterian Columbia hospital were found to be Covid + when universally screened for Covid on admission to L&D.

None would have otherwise been tested for Covid and only a few had any any symptoms at all.

nyc area is a hotbed but 13% is a lot. The denominator could be huge.

Wow. I would think urban density as the best correlate for highest prevalence. NYC likely has the highest rate by far. I would guess Philly #2.

 

[D1b()4]

13% of consecutive women presenting for Routine delivery at NY Presbyterian Columbia hospital were found to be Covid + when universally screened for Covid on admission to L&D.

None would have otherwise been tested for Covid and only a few had any any symptoms at all.

nyc area is a hotbed but 13% is a lot. The denominator could be huge.

Of the 29 women asymptomatic at admission, fever developed in 3 prior to discharge (median length of stay 2 days). The only thing we know for sure is that 26 females testing positive for Covid-19 were asymptomatic during admission. We do not know what proportion had recently been sick or developed symptoms after discharge. Also, pregnancy is characterized by physiologic changes in immune system that might result in lower proportion of the women developing fevers/becoming symptomatic. To summarize, they haven't been followed long enough to determine whether they were truly asymptomatic, and the findings in pregnant females might not be generalizable to the entire population.

 

[Med Director New England]

True
Still a much higher number than I would have expected
This is the only study so far in the US (at Least that I am aware of) shedding any light on this question.

serology will definitely answer this question soon.

 

[D1b()4]

Approximately 60% of Roosevelt sailors with COVID-19 are asymptomatic. Again, it is not clear whether they had been recently ill or will develop symptoms eventually. I think an estimate that up to 50% of infected people are asymptomatic carriers is reasonable.

 

[Med Director New England]

Sero prevalence study by Stanford
Estimated to be at least 50x greater than known positives....

COVID-19 Antibody Seroprevalence in Santa Clara County, California

Background Addressing COVID-19 is a pressing health and social concern. To date, many epidemic projections and policies addressing COVID-19 have been designed without seroprevalence data to inform epidemic parameters. We measured the seroprevalence of antibodies to SARS-CoV-2 in Santa Clara...

www.medrxiv.org

 

[D1b()4]

Sero prevalence study by Stanford
Estimated to be at least 50x greater than known positives....

COVID-19 Antibody Seroprevalence in Santa Clara County, California

Background Addressing COVID-19 is a pressing health and social concern. To date, many epidemic projections and policies addressing COVID-19 have been designed without seroprevalence data to inform epidemic parameters. We measured the seroprevalence of antibodies to SARS-CoV-2 in Santa Clara...

www.medrxiv.org

The sampling is probably affected by self-selection bias and is not truly random: "We recruited participants by placing targeted advertisements on Facebook aimed at residents of Santa Clara County."
If I had cold-like symptoms two weeks ago, I would love to participate in that study to see whether I was actually exposed to Covid-19.

"Over 24 hours, we registered 3,285 adults, and each adult was allowed to bring one child from the same household with them (889 children registered)."
Letting the participants to bring their children residing in the same household also affects random sampling. In addition, children are known have mild/subclinical course of Covid-19.

 

[gbwillner]

The sampling is probably affected by self-selection bias and is not truly random: "We recruited participants by placing targeted advertisements on Facebook aimed at residents of Santa Clara County."
If I had cold-like symptoms two weeks ago, I would love to participate in that study to see whether I was actually exposed to Covid-19.

"Over 24 hours, we registered 3,285 adults, and each adult was allowed to bring one child from the same household with them (889 children registered)."
Letting the participants to bring their children residing in the same household also affects random sampling. In addition, children are known have mild/subclinical course of Covid-19.

The sensitivity of the test was also 67%

 

[Med Director New England]

Even if imperfect this is the best data there is currently, Sample was taken 4/3 and4/4 so pretty recent snapshot. Also an insensitive serologic test would underestimate the sero-positivity rate btw...

No doubt there will be additional similar studies with more robust numbers coming soon. Quest partnered with the CDC are retrieving archived sera from various locations throughout the US and at various time points in the last 12 weeks for the same type of study.

it is looking like the denominator will be huge - the bigger the better. Natural immunity will reduce Ro.

 

[D1b()4]

In my opinion, unless the self-selection bias is somehow accounted for, the findings of this study are not very useful.

Does anyone understand how they used the response to prior clinical symptoms question in their analysis? This piece of information would be helpful to get an idea whether the might be significant self-selection bias.
"The survey asked for six data elements: zip code of residence, age, sex, race/ethnicity, underlying comorbidities, and prior clinical symptoms"

 

[Med Director New England]

You got to start somewhere...
No study design will be perfect & all retrospective samples are inherently flawed.

that said testing as much serum as possible for antibodies with a reliable test will be the best way to estimate what % of the population (the denominator) have or had Covid infection. The more the numbers, the more The geographic diversity, the more points in time sampled during the pandemic the better this data will be. The Quest / CDC results will be useful. Again not perfect - anyone getting blood drawn had a reason for it with possible confounding factors. I am told they are including frozen sera at different time points thru the pandemic in the us, different parts of the country and some samples predate the pandemic as a source for reliable negatives when the us population was naive to the virus.

me and my spouse are gettingblood drawn next week for Ab testing. I am very interested in my results and am hoping I have serologic evidence of exposure.

 

[D1b()4]

me and my spouse are gettingblood drawn next week for Ab testing. I am very interested in my results and am hoping I have serologic evidence of exposure.

Let us know how it goes. I would love to get tested as well.

 

[Med Director New England]

Let us know how it goes. I would love to get tested as well.

I’ll post it

you shouldn’t have any trouble getting tested if you want it. Reasonable for any healthcare worker to get tested.

 

[yaah]

Two other illuminating studies:

1) Ohio prison has 73% of inmates testing positive (they're not all displaying symptoms), and 21% of all positive cases in Ohio are prisoners.

At one Ohio prison, 73 percent of inmates have tested positive for COVID-19

The latest Speed Read,/speed-reads,,speed-reads, breaking news, comment, reviews and features from the experts at The Week

theweek.com

2) Homeless shelter in Boston, 146 of 397 tested positive, none had symptoms. CDC reviewing ‘stunning’ universal testing results from Boston homeless shelter

These raise the question of course of false positives, but is that really likely to be the best explanation?

 

[D1b()4]

2) Homeless shelter in Boston, 146 of 397 tested positive, none had symptoms. CDC reviewing ‘stunning’ universal testing results from Boston homeless shelter
These raise the question of course of false positives, but is that really likely to be the best explanation?

Those are probably not false positives. Some of them developed symptoms eventually though the article does not state how many:
"The 146 people who tested positive were immediately moved to two different temporary isolation facilities in Boston. According to O’Connell, only one of those patients needed hospital care, and many continue to show no symptoms."

 

[Med Director New England]

Two other illuminating studies:

1) Ohio prison has 73% of inmates testing positive (they're not all displaying symptoms), and 21% of all positive cases in Ohio are prisoners.

At one Ohio prison, 73 percent of inmates have tested positive for COVID-19

The latest Speed Read,/speed-reads,,speed-reads, breaking news, comment, reviews and features from the experts at The Week

theweek.com

2) Homeless shelter in Boston, 146 of 397 tested positive, none had symptoms. CDC reviewing ‘stunning’ universal testing results from Boston homeless shelter

These raise the question of course of false positives, but is that really likely to be the best explanation?

Saw the mass study....
I think the prison setting and a homeless shelter are probably a bit cruise ship like. Rampant spread once virus introduced

Hopefully herd immunity is taking hold

 

[ccous]

Saw the mass study....
I think the prison setting and a homeless shelter are probably a bit cruise ship like. Rampant spread once virus introduced

Hopefully herd immunity is taking hold

Already widespread infection and a huge amount of asymptomatic or mild cases, and a pretty decent amount of ‘immunity’ in these individuals (as in many months of persistent protective antibodies), is really our only hope of not having this virus become a horrific headache for the next couple of years.

Hopefully upon reinfection (which I’m sure happens given the nature of similar viruses and viruses in general), the subsequent infections do not become severe due to the prior antibodies. It would suck to have a mild infection and then get hit with a worse one, or have a severe case and have it wipe you out the next time.

And let’s hope Remdesivir can actually treat the severely ill. Maybe even eventually give to those who are moderately ill to prevent progression.

 

[D1b()4]

Here is probably a better estimate of true prevalence (13.9% in NY state):

New York releases antibody testing data: 14% of population may be infected with coronavirus

New York may have 13.9% of its population infected with coronavirus, preliminary state results showed Thursday.

www.usatoday.com

This is probably somewhat higher than the true prevalence because it tested people out in the society shopping who are more likely to be positive.

 

[elburrito]

Coronavirus (COVID-19) Overview

COVID-19 is a new type of coronavirus that causes mild to severe cases. Here’s a quick guide on how to spot symptoms, risk factors, prevent spread of the disease, and find out what to do if you think you have it.

www.webmd.com

 

[Med Director New England]

Let us know how it goes. I would love to get tested as well.

Just got my IGG test back : POSITIVE

this was run on the abbot Immunoasaay (not POC). I took a look at the package insert to review the validation data with interest in test specificity.

Abbot ran 997 archived sera drawn before the Covid outbreak. Only 4 of them were positive, presumably cross reacting with other corona viruses. So the test looks pretty specific.

In retrospect I likely had this back in mid Feb, I do recall a mild thing that I dismissed as a cold lasting about a day. I was a little fatigued for maybe one additional day. Never felt like I had a fever. Never failed the temperature checks that started about 3-4 weeks ago for me at work.