Miracle, or disaster?

[M T]

Cases in Santa Clara County (SCC) grew exponentially for 6 weeks beginning the 2nd week of June.  July 12-18 had 3 days higher than any day from a previous week.  The previous week had 5 days higher than any day from a previous week. Etc.  July 12-18 was the sixth week of rising case numbers that had all the signs of exponential growth.

If I told you that in the next week (July 19-25), SCC had 56% fewer cases than in the preceding week (dropped from 1636 during July 12-18 to 719 during July 19-25), you might think it was a miracle.   You might say "Open the bars and let's drink to that."  or "Open the churches and let's praise God."   I'm sure some will say that.

But, before you toast the rapid change in numbers, or sing a hymn of thanksgiving, take a look at the testing figures: SCC had 56% fewer tests in the week of July 19-25 than the previous week (dropped from 42,826 to 18,736).

56% fewer cases and 56% fewer tests.   Coincidence?  I think not.

On July 23, California changed the rules for whose tests can get processed.  This has affected the number of tests of SCC residents that can be processed.

I think pretty obviously, this change, while it may be necessary for fast & equitable testing in the state or country, has caused probably 900 or more cases of COVID-19 to be missed in this past week alone.

In some other county, one might try to say that maybe this decrease was a result of the closing of certain businesses in the middle of the prior week (July 15, as applied to SCC).   Other than some businesses that were open for two days (July 13 & 14), SCC already had those businesses closed.  So there was no real effect.  (Indeed if there were any effect, it would be that those two days would have INCREASED cases during the week when cases plummeted, due to incubation times.)

The exponential growth SCC had for 6 weeks probably didn't go away.  I suspect those missing cases are in the community and possibly causing R to rise.

I think this has been a disaster for SCC, at least.   I think it also highlights the problem of inadequate testing virtually everywhere in the US.

You will see that the % positive rate dropped simultaneously with this change in the rules.  That suggests to me that the disease rate in those who can't get the test now is higher than those that can.  I hope the county can look at the preceding week's testing and identify if the new rules had been in place, would the % positive rate been lower or not?  (By the way, SCC has been reporting incorrect 7-day % positive numbers.  I sent a note to them Thursday night showing how their numbers are not calculated as the state specifies. (They were averaging the daily % positive values.))


In mid-June, President Trump reportedly joked, "If we stop testing right now, we'd have very few cases, actually."   I find it ironic that California is proving him right.

[edited 7/31 8PM. Previously had 54% for both reductions. I meant 56% for both, and have corrected it.]

[OutsiderFan]

I find it ironic that California is proving him [POTUS] right.

Um, surely you had your tongue firmly planted in cheek, right?

I mean there isn't really any need to point out the number of infections that exist have no relationship to those identified by testing, especially given how horrible our testing is, right? Not only are we not doing enough testing to catch as many infections as there are, we're not letting people know if they are positive fast enough to make sure the infected people isolate and can be contact traced, which furthers spread. 

I looked at the Santa Clara County testing dashboard this morning and was equally alarmed as you, MT, by the dark shading over recent test results, and the accompanying "we're still waiting for results" message.

It's also hard for me to understand the drop-off in national cases reported. Last week we hit 78k on 7/24. This week we haven't gone past 69k on any day. I find it really hard to believe things started dropping off so fast.  It seems last week there might have been a bit more recognition of how serious the pandemic is in places that weren't taking it seriously before (as evidenced by the 80% of Americans now favor mandatory masks), but it would take two weeks for any case reductions to show in the data, yet here we are with fewer cases being reported this week.  We should have gone over 80k reported infections this week, based on the data and now pretty reliable prediction models, and did not. 

The thing to watch for is deaths reported.  If we see deaths continue climbing, and we see a death rate higher than the number of cases reported over the last two weeks, in a few weeks, we'll know for certain that the reason case numbers dropped this week was because testing capacity was overwhelmed, not because there were fewer cases.

[dabigv13]

SoCal hospitalizations and ICU use peaked a week ago and have been steadily decreasing since. Hope that continues.

[BostonCard]

In mid-June, President Trump reportedly joked, "If we stop testing right now, we'd have very few cases, actually."   I find it ironic that California is proving him right.

If a tree falls in the woods and no one hears it, did it still fall?  I mean, excepting the asymptomatic patients, most of the untested patients with COVID-19 will still feel like crap and, more importantly, will still infect others.  Thankfully, they will also develop immunity.

Number of cases has never been a great metric, because we've always known that there are many people who go undiagnosed.  The things to look at have always been percent positivity, hospitalizations, ICU utilization, and deaths.

BC

[Goose]

I mean there isn't really any need to point out the number of infections that exist have no relationship to those identified by testing, especially given how horrible our testing is, right?

NO relationship? Really? OF, IMHO that is just silly. Even if the number of tests is restricted, one can pretty well be assured that the % positive is proportional to the prevalence of the disease. Given what the restrictions are, it could be argued % positive should increase. What one can't do is compare raw number of positive tests from before the restrictions to after the restrictions. It is perfectly reasonable to be alarmed by these changes in testing priorities, but IMHO it is important to not make the problem bigger than it is. The problem is big enough to not require that.

Not only are we not doing enough testing to catch as many infections as there are, we're not letting people know if they are positive fast enough to make sure the infected people isolate and can be contact traced, which furthers spread.

I basically agree with this statement. However, I think it covers three separate problems. The problems are 1)Finding all the infections, 2) Getting the results fast enough to be effective, and 3) Making sure the infected people isolate.

The "propagation delay" (#2) problem is one reason the change in testing priority was made. It should speed up getting results to the people that are "most likely" to be positive. However, at this point in the epidemic it is difficult to understand why SCC is in this bind. At one point Stanford was doing all the testing for SCC. Is this still true, or have they developed their own capability? Are they relying on commercial labs and the commercial labs are overloaded? Did SCC decide that we had "won" and done nothing to ensure they were ready for a second (or third, or Nth wave). It wouldn't be a tragedy if the bought a "extra" 96 well PCR machine that they turned out not to need. It would be a tragedy if they need it and don't have it. Why don't we know, and why isn't the Press asking pointed questions?

The problem of locating everybody with this disease (#1) is turning out to be a disaster. As Fauci said, this virus is a bitch. Testing capacity is an issue, but it may not be the main issue. Initially, the assumption was people get the virus, become ill, get tested, test positive, isolate and all is well. If those people can't get tested, it is a problem. In SCC I think they can. However, we now know that there are lots of people who get the virus and do not become ill. Can those people pass on the disease? For sure, at least some of them can. How many is "lots"? 2X? 10X? We don't know. We also know that people who do become ill can transmit the disease for some period of time (several days) before they become ill. So testing sick people isn't going to do the job. OK, test everybody. Logistically impossible. Even tiny Iceland couldn't do that. Test a sample? Fine, but it has to be a large enough sample (which runs into limited testing speed) to detect the virus, and what do you do with the result? So some neighborhoods seem to have lots of cases. What do you do? What they do in Italy is lock it down, nobody goes in or out. They could (have the legal authority) do that in SCC, but they won't. And by the way, the non-hot neighborhoods aren't "clean". It is just below your test sensitivity.

Contact tracing and isolation could help find many asymptomatic spreaders. However, nobody in the Bay Area, and probably nobody in any USA urban area has a contact tracing and tracking capability that is "working". San Mateo County at least is (was) telling the truth. On July 19th, they reached 34% of the cases who tested positive. On July 19th, 18% of those cases were able to isolate. On July 19th they reached 41% of the contacts. However, since they only reached 34% of the "index" cases, the number of contacts was undoubtedly "small" relative to the number of "actual" contacts, even if 100% of the cases reached cooperated (unlikely). How small the number of contacts was I don't know, because that is not a number they provide. I am working on that. I strongly suspect SCC is worse.



So we don't lack for problems. Testing is one, but maybe not the main one. Places like Taiwan have really scrambled (using quarantine, testing, contact tracking etc.) to ensure there weren't many cases of "community spread". The USA never really did that. We never drove the number of detected cases low enough to ensure  we could contain them. We sat on our hands while we needed to be developing our contact tracking. SCC made all kinds of statements about how may people they would have on the job by July 15th. Unfortunately, that would have been too late, as we were already opening up. Did they reach that level of staffing? I don't know. Anybody have a link?

[BostonCard]

I believe that the bottleneck is testing reagents.  Like literally not enough pipettes or extraction solution, etc.

https://www.mckinsey.com/industries/phar...ic-testing

Executing a test requires some 20 different reagents, consumables, and other pieces of equipment. Of those materials, major shortages have been reported in RNA-extraction kits and certain reagents, including enzymes and primers.7 The global manufacturing capacity for molecular-assay tests is estimated to be between 37 million and 38 million tests a week, given current availability of the various test components, with RNA-extraction kits being the bottleneck to higher capacity (Exhibit 3).8 That compares with fewer than 10 million tests a week being conducted around the world, according to our research.

BC

[Goose]

I believe that the bottleneck is testing reagents.  Like literally not enough pipettes or extraction solution, etc.

https://www.mckinsey.com/industries/phar...ic-testing

Executing a test requires some 20 different reagents, consumables, and other pieces of equipment. Of those materials, major shortages have been reported in RNA-extraction kits and certain reagents, including enzymes and primers.7 The global manufacturing capacity for molecular-assay tests is estimated to be between 37 million and 38 million tests a week, given current availability of the various test components, with RNA-extraction kits being the bottleneck to higher capacity (Exhibit 3).8 That compares with fewer than 10 million tests a week being conducted around the world, according to our research.

BC

Just to be clear, I think you mean that pipettes or extraction solution, etc. are not the limiting factor in testing. After reading the McKinsey document, I conclude it is a consultant's document. The relevant explanation of the reagent issue is

There are two potential explanations for the gap. First, a significant quantity of the reagents being manufactured are those that run on open systems—that is, less integrated systems that can run a wider range of test methods. Such reagents cannot be used with the high-throughput machines that tend to be used in developed countries and are closed systems requiring cartridges loaded with proprietary reagents manufactured by the OEM. Second, as Exhibit 3 shows, most of the available manufacturing capacity is based in China, potentially making access to it more difficult, given validation and export considerations.

Keyword is "potential". Unfortunately they offer no evidence that Roche, ABI etc. aren't delivering on time all that has been ordered. It should be easy to find out if that is the case, but the authors didn't. I have no idea myself, but if it were a bottleneck, I would have expected to hear about it from the press, or from places like UW and Stanford who are running such tests. We have heard about shortages at Stanford early on, but not subsequently. If I ran a testing lab with samples backing up and my vendor wasn't delivering, I assure you I wouldn't be quiet about it. It would seem to me the authors can't justify the statement that the 10M tests per day limit is caused by reagent limitations, given the capacity limit there of 38M tests and no real evidence that the proprietary reagents are actually the bottleneck. 
There has been a significant difference between reported testing capacity and actual tests being performed since the beginning. My first post noting this was on April 19th. I didn't know why then, and even after reading the McKinsey document, I don't know why now. Apparently nobody does, or it would be fixed by now. What I can't understand is the reason that e.g. Quest would not increase capacity if they are running flat out. They have had time to do so. Even if the lead time on a high throughput PCR machine is 90 days (and as I understand it it is more like 30 days), that can't be the issue. I can only conclude they aren't at capacity yet. WTFO?

[BostonCard]

I wonder if some of this is supply chain.  Just because the manufacturing capacity exceeds demand, doesn't mean that there isn't a lag time between increased demand and when the supplies reach where they need to be in time to match the spike in supply.

Especially with travel impacted, and a lot of these supplies being manufactured on the other side of the world.

BC

[Goose]

I wonder if some of this is supply chain.  Just because the manufacturing capacity exceeds demand, doesn't mean that there isn't a lag time between increased demand and when the supplies reach where they need to be in time to match the spike in supply.

Especially with travel impacted, and a lot of these supplies being manufactured on the other side of the world.

BC

I think this is potentially an issue, but six months in to the outbreak, you would expect the testing supply chain pipeline to be full by now. The purchasing people also should know what the lead times are, and I would expect the lead times should be going down, not up. Independent of the cause, if I am a lab who can't test because I don't have a reagent (for whatever reason), I would be screaming to the press and the public about it. We aren't seeing that.

I think it more probable that there are administrative delays that make the process take more time than it should. There are lots of "steps" in getting a sample from the test subject to the remote lab and back, and each one of those steps probably isn't optimized for speed. Nobody has analyzed the system from the point of view of minimizing the delay. Each step may not take real long, but is death by 1000 cuts. Several different organizations are involved in each step. Tests are probably marshaled several different times in the process. Each time one waits for "enough" tests to accumulate, there is a delay. Obviously, I don't know this for a fact. However, given it is a system that has been created basically from nothing in the last 5 or 6 months, it is unlikely it is well optimized.

[chrisk]

They should be able to repurpose military logistics to make this happen. If we could handle the logistics for D-Day 75 years ago, we could handle this.

But many of the tools that were used in WWII have either been ignored or atrophied.

[Goose]

They should be able to repurpose military logistics to make this happen.  If we could handle the logistics for D-Day 75 years ago, we could handle this.

Remember that it took three years to get ready for D-day. Also, most military logistics isn't very applicable to our current problems.

But many of the tools that were used in WWII have either been ignored or atrophied.

The entire Federal, State, and County public health system has atrophied. Much of that atrophy is recent, i.e. the last 30 years or so. In the 1950s it was still in very good shape. As Fauci said, we had so many  successes in the last 60 years that we just let it slip away. We didn't need it any more.

[BostonCard]

It shouldn't take more than 48 hours to turn around the tests.  Assuming the testing site is away from the lab, you hold the tests at the testing site until the end of the day, and then batch deliver them to the laboratory site.  Even if the tests are held overnight and processed the next day and not run until overnight day 2, you can have the test resulted by morning, 48 hours after it is drawn.  Delivery of the results can be electronic (and therefore immediate).

BC

[dabigv13]

I wonder if some of this is supply chain.  Just because the manufacturing capacity exceeds demand, doesn't mean that there isn't a lag time between increased demand and when the supplies reach where they need to be in time to match the spike in supply.

Especially with travel impacted, and a lot of these supplies being manufactured on the other side of the world.

BC

I think this is potentially an issue, but six months in to the outbreak, you would expect the testing supply chain pipeline to be full by now. The purchasing people also should know what the lead times are, and I would expect the lead times should be going down, not up. Independent of the cause, if I am a lab who can't test because I don't have a reagent (for whatever reason), I would be screaming to the press and the public about it. We aren't seeing that.

I think it more probable that there are administrative delays that make the process take more time than it should. There are lots of "steps" in getting a sample from the test subject to the remote lab and back, and each one of those steps probably isn't optimized for speed. Nobody has analyzed the system from the point of view of minimizing the delay. Each step may not take real long, but is death by 1000 cuts. Several different organizations are involved in each step. Tests are probably marshaled several different times in the process. Each time one waits for "enough" tests to accumulate, there is a delay. Obviously, I don't know this for a fact. However, given it is a system that has been created basically from nothing in the last 5 or 6 months, it is unlikely it is well optimized.

This is incorrect. There are real delays in reagent supply chain. Our hospital recently had to add a new type of testing machine, not cheap. We could have bought more of the other testing machines we already have that functions very well, but by diversifying we can ensure more backup capability, and use different types of reagents because we couldn't get enough from what we had to return labs quick enough. We needed 48 hour turnaround but were starting to come up to 72 hrs. This is pretty unusual- normally there is one type of lab you can order for any particular lab. Now there's three, two in house and one send out (Quest, which takes the longest, and only used for low priority tests now).

This is not including rapid tests, only PCR.

Now, different orgs have different throughput and needs. My wife's practice uses a more smaller, perhaps more boutique, lab without big numbers and bought a lot of reagents when able. They can turn around a test in 6 hours.

The main issue again is our prevalence. If this thing wasn't raging everywhere we'd surely have results uniformly in less than 24 hrs.

Here's NYT on this-
https://www.nytimes.com/2020/07/23/healt...rtage.html

[M T]

The main issue again is our prevalence. If this thing wasn't raging everywhere we'd surely have results uniformly in less than 24 hrs.

The CDC reports it was told of 1.9M tests last week.  That's roughly 270,000 tests per day.  Just a drop in the bucket when you compare it to every day items that are mass produced around the country.  I found this "In 2017 in the United States alone, an estimated 65 billion 12 oz. beers were consumed, along with over 4 billion bottles of wine."  That's 1.25B bottles of beer all in all per week (about 650x the number of tests).  You can't spin up such production or the distribution chain in 6 months.

Is COVID hitting the US harder than we thought even in mid-May?  Certainly.  SCC scaled its tracing for a maximum of 75 cases per day.  L.A. County scaled for a peak of 2500 (they peaked about 3300, averaging 2000 per day over the week).   And, California isn't the hardest hit area.  I suspect the supply & production of reagents were scaled for less disease than we have.

(For comparison, Wuhan tested its entire city over a 2 week period:  9.9M tests in 17 days (582K tests per day) beginning May 14.  Can you imagine the squawking if NYC or LA tried to require every one to take a test?)

[allrightynow]

The CDC reports it was told of 1.9M tests last week.  That's roughly 270,000 tests per day.  

It may not change your point, but actually the US is currently averaging about 800,000 tests per day according to covidtracking.com

[magnus]

I read an article recently (can't seem to find it now) that said the issue was that we were too reliant on the more automated testing machines.  One where you could just put it in and it'd do everything for you.  And unfortunately, while we have a ton of these machines.  We don't have whatever it was that was needed for these machines to process these tests (I'll call them test kits).  These companies all had their own proprietary kits.  The lack of a common standard means that if you are using company ABC's most prolific machine that can process 6k tests a day.  You're in a bind because that company has 100k such machines out there, but can only produce 1 million test kits  a day, so you're still short 5 million from maxing out those machines.

I'm making up the numbers, but the testing crunch, at least for the US, was attributed to the preference for these more automated machines, each with their own proprietary test kits.

And so like dabigv13 said hospitals are resorting to have to buy multiple brands so that they won't be hamstrung by one company running out of test supplies.  Unfortunately, that likely means that the hospitals are running way under capacity considering all the machines they have.

Edit:
Found the article:
https://www.reuters.com/article/health-c...SL2N2EV01Z

[Hurlburt88]

given that there are still snags in testing and potentially also with PPE in some locations, probably very valid concerns in this article about getting things organized when vaccines start to become available

https://www.reuters.com/article/us-healt...SKCN24W1ON

[M T]

The CDC reports it was told of 1.9M tests last week.  That's roughly 270,000 tests per day.  

It may not change your point, but actually the US is currently averaging about 800,000 tests per day according to covidtracking.com

Thanks for the correction.  I wasn't careful enough to recognize that the CDC apparently doesn't get all the tests (but that surprises me). In the text above the table I used was "The number of specimens tested for SARS-CoV-2 using a molecular assay and reported to CDC by public health laboratories and a subset of clinical and commercial laboratories in the United States are summarized below."

[BrevinsBest]

I've taken 2 COVID tests. The first one, done through the SMC Verily/Google project at the convention center took 2 weeks for me to get my results. The second test, done through Stanford, took 2 days for results. Waiting 2 weeks for results is worthless.

[Goose]

I think this is potentially an issue, but six months in to the outbreak, you would expect the testing supply chain pipeline to be full by now. The purchasing people also should know what the lead times are, and I would expect the lead times should be going down, not up. Independent of the cause, if I am a lab who can't test because I don't have a reagent (for whatever reason), I would be screaming to the press and the public about it. We aren't seeing that.



I think it more probable that there are administrative delays that make the process take more time than it should. There are lots of "steps" in getting a sample from the test subject to the remote lab and back, and each one of those steps probably isn't optimized for speed. Nobody has analyzed the system from the point of view of minimizing the delay. Each step may not take real long, but is death by 1000 cuts. Several different organizations are involved in each step. Tests are probably marshaled several different times in the process. Each time one waits for "enough" tests to accumulate, there is a delay. Obviously, I don't know this for a fact. However, given it is a system that has been created basically from nothing in the last 5 or 6 months, it is unlikely it is well optimized.

This is incorrect. There are real delays in reagent supply chain. Our hospital recently had to add a new type of testing machine, not cheap. We could have bought more of the other testing machines we already have that functions very well, but by diversifying we can ensure more backup capability, and use different types of reagents because we couldn't get enough from what we had to return labs quick enough. We needed 48 hour turnaround but were starting to come up to 72 hrs.

It sounds to me like your lab had a capacity issue due to the increase in cases. When you are running at 100% of capacity and have any delay in your supply chain, you can't ever recover. Thus testing becomes permanently delayed. The increase in cases also probably increased consumption of everything, making it necessary to order more supplies earlier than expected. The new machine will increase capacity, and it probably does make sense to order from another manufacturer to reduce dependence on single-source materials. It also may be you could get that machine faster.

This is pretty unusual- normally there is one type of lab you can order for any particular lab. Now there's three, two in house and one send out (Quest, which takes the longest, and only used for low priority tests now).
[/quest]
Yes, these are the marshaling delays I referred to.

[quote]

This is not including rapid tests, only PCR.



Now, different orgs have different throughput and needs. My wife's practice uses a more smaller, perhaps more boutique, lab without big numbers and bought a lot of reagents when able. They can turn around a test in 6 hours.

They may have some capacity that lies unused most of the time. They probably doing the test the "manual" PCR way. Morel labor intensive, but  more responsive. You don't need to fill up the machine, etc.

The main issue again is our prevalence. If this thing wasn't raging everywhere we'd surely have results uniformly in less than 24 hrs.



Here's NYT on this-

https://www.nytimes.com/2020/07/23/healt...rtage.html

For sure the prevalence has really upped the need for testing, and it would appear the suppliers haven't expanded as fast as it requires. Thanks for the link. The NYT article seems to say it is random shortages of everything that is limiting testing throughput.

I read an article recently (can't seem to find it now) that said the issue was that we were too reliant on the more automated testing machines.  One where you could just put it in and it'd do everything for you.  And unfortunately, while we have a ton of these machines.  We don't have whatever it was that was needed for these machines to process these tests (I'll call them test kits).  These companies all had their own proprietary kits.  The lack of a common standard means that if you are using company ABC's most prolific machine that can process 6k tests a day.  You're in a bind because that company has 100k such machines out there, but can only produce 1 million test kits  a day, so you're still short 5 million from maxing out those machines.

I'm making up the numbers, but the testing crunch, at least for the US, was attributed to the preference for these more automated machines, each with their own proprietary test kits.

And so like dabigv13 said hospitals are resorting to have to buy multiple brands so that they won't be hamstrung by one company running out of test supplies.  Unfortunately, that likely means that the hospitals are running way under capacity considering all the machines they have.

I doubt that. They are running all the tests that they can. Their "capacity" may be reduced by the lack of custom plastic or the lack of primer kits. Quoting how many tests per day your machines can run as your capacity is an error that has been prevalent during this whole epidemic. Your capacity can be limited by lack of technical staff, lack of reagents, or lack of people to do the paperwork. At one time or another all of these have been cited as a reason for delays in testing.

Edit:
Found the article:
https://www.reuters.com/article/health-c...SL2N2EV01Z

Good article. Thanks for the link. It does contain some real examples of limitations due to insufficient reagent supplies. This has been absent in most of what I had seen previously.
However, I think the Reuters article leaves the impression that using the "old style" method of testing was an option to deliver adequate testing in the USA. It isn't. The article does point out that it is labor-intensive, but it doesn't point out adequately that it is SKILLED labor intensive. Training people takes just as long or longer than increasing supply from sole-source vendors does, assuming you can even find the people qualified to do the work. The impression is also given that you can just order a "generic" reagent and all will be well. It is not that simple.The reagent has to be validated for that test. It can be done by the vendor, or by the user, but it has to be done. There are approved protocols that must be followed. For example, the CDC test cited in the article was only approved with certain lots of reagents from two vendors. Others may have been added later, but it isn't a free-for-all. The USA is doing it mostly using high-throughput machines because with our population automation is the only way we can come close to doing the number of tests required. Nothing is stopping hospitals (and State labs) from doing it the "old" way. If it were viable, it would be done that way.