Tucker: Are coronavirus lockdowns working?

[silentwitness]

More discussion.

 

[josie_L]

Oh my gosh, yes, it is, I just read several articles one of which is from the Lancet using that specific term.

It is refered to as “crude” because it is not reflective of the real morbidity rate.

And in the case of Covid-19, the deaths are measured by cases per 100 000.

From a Lancet article:

5

The authors argue that crude case fatality ratios obtained by simply dividing the number of deaths by the number of cases can be misleading because there can be a period of 2–3 weeks between a person developing symptoms and that case being detected and reported, and because surveillance of a novel virus is biased towards detecting severe cases, especially at the beginning of an outbreak when test capacity is low.

 

[dvdjs]

Oh my gosh, yes, it is, I just read several articles one of which is from the Lancet using that specific term.

Here is a goo place to look at the proper use of the terminology.
https://www.cdc.gov/csels/dsepd/ss1978/lesson3/section3.html

The word “crude” is used formally when looking at incidence within a total population, typically undifferentiated by cause. One should differentiate between the: a) case fatality rate; b) death-to-case ratio; and c) the crude mortality rate. Lieu did not refer to c), but actually gave b), although calling it a). The difference between a) and b) is can be minor of large, depending on the duration of the the cases. and the time window sampled. There are a numerous refinements that will be made to the current death-to-case ratios to determine the case fatality rate. As I have mentioned, these include the expansion of the time window to cover the full duration of a case from onset to resolution, more comprehensive measurements of the incidence of illness, and numbers of deaths associated with them. Time will tell.

 

[josie_L]

The researchers are making this distinction and so am I, the point is Lieu ISN’T. I think I’ve made that clear by stating that his present mortality or whatever you want to refer to it as is not VALID NOR REFLECTIVE OF THE TRUE RATE.

Because his rate/ratio is simply the dividing of numbers of deaths by known cases, which is exactly what the researchers refered to as a CRUDE MORTALITY RATE because it is not factoring in all unknown cases.

Whether we will ever get a true case mortality rate will depend upon how much testing is done in the aftermath of the first wave.

 

[dvdjs]

is not VALID NOR REFLECTIVE OF THE TRUE RATE.

This is the nub.
A well defined parameter whose value properly reflects the data available is VALID.
Thus, the ratio of the reported deaths to confirmed cases is, validly the the ratio of the reported deaths to confirmed cases.

The fact that such a parameter, while giving an indication of the present state of pandemic, should not be taken as COVID-19 mortality rate is not arcane. I have been pointing out from the outset that there are several known unknown to be evaluated before a disease specific death rate can be evaluated. Some of these will tend to lower the disease specific death rate as compared to the trending ratio of deaths to confirmed cases. Some will tend to raise them.

At this point, it is far from clear which of these will dominate in the final analysis. As such, it is arguably better when characterizing the present status of the pandemic with the actual numbers that we know - the raw data - then it is to use estimates of the known unknowns.

Whether we will ever get a true case mortality rate will depend upon how much testing is done in the aftermath of the first wave.

For large pandemics, final analyses are informed by anomalies in crude death rates.

 

[Maximus1]

I think he said it didn’t count because other examples exist

 

[josie_L]

I will reiterate the current CRUDE mortality rate as termed by scientists is not the actual morbidity rate. And ALL research being done to decipher a rate closer to what it would actually be puts it at a percentage far lower than 5.6 %.

The figure quoted by Lieu is simply of propagandic use in this contentious struggle between left and right.

 

[Annie]

As such, it is arguably better when characterizing the present status of the pandemic with the actual numbers that we know - the raw data - then it is to use estimates of the known unknowns.

The problem with this is that it can give a completely misleading idea about what is happening.

If a town of 10,000 has 100 confirmed cases and 10 die, that is a death rate, according to your system, of 10%.

In a similar town where more testing is happening, maybe 1000 are confirmed cases and 10 die, then you have a death rate of 1%, according to your system.

Everything else remains the same, bit in the town that hears 10% death rate, the people are terrified and the town with 1% not so much.

 

[dvdjs]

I will reiterate

It would be better to avoid equivocal uses of "crude’.
The strict technical use reefers to an aggregates that is undifferentiated by, for example, age.
It is also used adopted to a more generic use for which, “raw” (vs treated) is more commonly used.
I suppose that one extend it further to use in the discussion of a data set in which elements remain to be added. But the further one strays from the strict technical use the less clear.

The problem with this is that it can give a completely misleading idea about what is happening.

Actually, it is more often the lack of clarity that allows for misleading.
If one reports of the crude death rate in a town’s population it is what it is. Since the data represents past action, I am not sure who should be terrorized. One might be more misled by a statement that goes beyond a simple recounting of the facts and makes anticipates the anticipates that final rate - particularly one that is accompanied, as it should be, by a, say, 95% confidence interval. I suppose that say might be terroized by such estimates.

The figure quoted by Lieu is simply of propagandic use

No it is a simple, direct, and accurate account of actual facts; the virus doesn’t care about left or right.

 

[Annie]

One might be more misled by a statement that goes beyond a simple recounting of the facts and makes anticipates the anticipates that final rate

I agree that more information is needed for clarity. You seem well-versed in statistics, so you might not be aware of how people who are not so knowledgeable react to them.

Saying that there is a death rate in town X (the first town I mentioned above) of 10% without all any qualification is scary to the average person, and sounds like the situation is worse than in town Y (the second town I mentioned), even though the situation is probably very similar.

 

[dvdjs]

Saying that there is a death rate in town X (the first town I mentioned above) of 10% without all any qualification is scary to the average person

Perhaps. But in the scenarios you gave, that was that fact and the death rate was indeed 10-fold worse than in town Y. It is of course of great interest to understand the basis of the difference, in order to adopt behaviors so that town Z may have a more like Y than X, and that Y does not evolve toward X.

 

[LeafByNiggle]

dvdjs:

As such, it is arguably better when characterizing the present status of the pandemic with the actual numbers that we know - the raw data - then it is to use estimates of the known unknowns.

The problem with this is that it can give a completely misleading idea about what is happening.

If a town of 10,000 has 100 confirmed cases and 10 die, that is a death rate, according to your system, of 10%.

In a similar town where more testing is happening, maybe 1000 are confirmed cases and 10 die, then you have a death rate of 1%, according to your system.

Everything else remains the same, bit in the town that hears 10% death rate, the people are terrified and the town with 1% not so much.

This analysis highlights a problem with focusing only on the mortality rate, as if that is the only statistic that matters. The risk to the public is due to two factors: the mortality rate and the infection rate. If we hear about a disease that is 80% fatal, but only infects 2000 people per year in the US, we would not be too worried, would we? Only 1600 people per year die of that disease. What ultimately matters is my chance of dying. In the hypothetical case I just described, my chances of dying in the next year from that disease would be 0.00049%. Now let’s suppose that some new test was developed that revealed that not 2000, but 20,000 people per year were infected, with most of them being asymptomatic. Then, of course we would have to revise our mortality rate from 80% to 8%. We would still have 1600 people per year dying of that disease, regardless of whether it is calculated as 80% of 2,000 or 8% of 20,000. So the risk to me as a member of the public would be exactly the same. I would still have the same 0.00049% chance of dying in the next year from that disease.

Applying this to covid-19, regardless of the mortality estimates, the risk (and therefore the seriousness) of this disease is dependent on the death rate as a fraction of the population. In that sense, covid-19 looks very serious, no matter how you calculate the mortality rate separately from the infection rate, for the overall death rate is dependent on both.

In the course of two months, we have already seen more of the population dead from this virus than from any seasonal flu has killed over the course of an entire flu season (except the 1918 pandemic, where we didn’t have antibiotics for secondary infections or ventilators, etc. to keep people alive). Those are the facts that are more relevant to the risk posed by this virus - not the isolated mortality rate.

 

[dvdjs]

The risk to the public is due to two factors: the mortality rate and the infection rate.

Yes. And I would add that the infection rate itself - apart form mortality - is of particular importance as we are learning of the potential of long term health problems in the aftermath of recovery.

In the course of two months, we have already seen more of the population dead from this virus than from any seasonal flu has killed over the course of an entire flu season

Yes.

 

[Annie]

Perhaps. But in the scenarios you gave, that was that fact and the death rate was indeed 10-fold worse than in town Y. It is of course of great interest to understand the basis of the difference, in order to adopt behaviors so that town Z may have a more like Y than X, and that Y does not evolve toward X.

In my hypothetical, the only difference between X and Y is testing. Say in X, only people sick enough to go to the hospital are tested, but in Y, people sick enough to go either to the doctor or to the hospital are tested.

Other than the testing, the two towns are completely the same.

Thus, the death rate as measured against confirmed cases is meaningless as it can easily manipulated by increased testing. Town Z can be identical to X and Y, but have testing of many more people, and in this way “decrease their death rate.”

 

[LeafByNiggle]

Thus, the death rate as measured against confirmed cases is meaningless as it can easily manipulated by increased testing. Town Z can be identical to X and Y, but have testing of many more people, and in this way “decrease their death rate.”

That is exactly why the death rate as measured against the total population is more relevant than the death rate as measured against confirmed cases.

 

[Annie]

Which is exactly what I have told dvdjs on 2 or 3 occasions in the recent past

 

[dvdjs]

In my hypothetical, the only difference between X and Y is testing. Say in X, only people sick enough to go to the hospital are tested, but in Y, people sick enough to go either to the doctor or to the hospital are tested.

Sorry, I thought you were talking of the death rate, rather than the case fatality rate, but I see that you meant the latter. It is quite fair to note that if testing is limited, the CFR is likely to be inflated, just as it is fair to indicate that unaccounted deaths and unresolved cases tend to deflate it. That is a point that I have made repeatedly.

Our situation on testing is not great, but we have reached the stage that about 1 in 7 tests are positive. I have plotted the state-by-state CFRs vs the positive fraction of tests.
Leaving out the four outliers where the caseload has overwhelmed testing, there does seem to much correlation with CFR and testing. And the average CFR is about 4%.
(Please Note: This uploaded content is no longer available.)

 

[josie_L]

And that’s what I’ve been trying to reiterate over and over again with dvdjs. Thank you!

 

[dvdjs]

That is exactly why the death rate as measured against the total population is more relevant than the death rate as measured against confirmed cases.

IOW the crude death rate vs the case fatality rate.

It could be more relevant. But that is not necessarily the case. (And the data that I pposted don’t support the idea.) In the growth stages of an epidemic, when transmission is starting and the degree of infection low, the death rate has the disadvantage of understating the severity of the outbreak, especially if it were confused with the CFR - which evidently happens. The does CFR have its own limitations, but apart from the very beginning of the outbreak before testing is taken up on earnest, or situations in which testing is limiting, the CFR will be more stable than the dearth rates.

Both of these have the advantage of being simple and direct measures of the state of the pandemic. Whatever limitations that imposes, or terror that it causes, it faithfully reflects the actual data, rather that a modellers conception of what the numbers will be when the picture is more complete.

 

[josie_L]

I never said the virus cares, but are you seriously in doubt of the misusage by both left and right commentators regarding the morbidity rate from covid-19?

It is being used as a propagandist tool for both sides, one which is undermining the severity of covid and the other over-exaggerating it.