[M T]

So, I found a very interesting dataset, the Oxford Government Response Tracker.  Basically, it tracks government action in response to the coronavirus.  The stringency index is based on Eight of the policy indicators (C1-C8) record information on containment and closure policies, such as school closures and restrictions in movement (the variables are: school closing, workplace closing, canceling public events, restrictions on gatherings, public transport closings, stay at home requirements, restrictions on internal movements, and international travel controls).  They show the relationship between the index and number of cases:

Uh, a country that is (say) 100x the population gets the same vertical scale on the left as the small country???  The scale on the right is some artificial scale, but I'd guess it is intended to be something proportional to the population.   So, if Lichtenstein did nothing and had all 38K population get sick, they would have looked ok just because the gray plot didn't go that high.  China would look worse, despite it having, what, 4 orders of magnitude more population.

Oranges versus barrels of apples.

And while someone is making this stuff up, how many countries in this set had 50 (or more) different regions that had different rules applied at different times? (Italy had some of that.)  Are these stringency steps when the last region applied them, or when the first did?

This needs a lot of work to pass the "smells like BS" test

Remind me, who on this board pointed out that any statistic based on quantity of positive tests was pretty much meaningless?  Or do we just pull that card out of our sleeve when we want to discount an outlier but ignore it for all the non-outliers.  (We have the same card for deaths, by the way.)

[M T]

So, I found a very interesting dataset, the Oxford Government Response Tracker.  Basically, it tracks government action in response to the coronavirus.  The stringency index is based on Eight of the policy indicators (C1-C8) record information on containment and closure policies, such as school closures and restrictions in movement (the variables are: school closing, workplace closing, canceling public events, restrictions on gatherings, public transport closings, stay at home requirements, restrictions on internal movements, and international travel controls).  They show the relationship between the index and number of cases:

Uh, a country that is (say) 100x the population gets the same vertical scale on the left as the small country???  The scale on the right is some artificial scale, but I'd guess it is intended to be something proportional to the population.   So, if Lichtenstein did nothing and had all 38K population get sick, they would have looked ok just because the gray plot didn't go that high.  China would look worse, despite it having, what, 4 orders of magnitude more population.

Oranges versus barrels of apples.

And while someone is making this stuff up, how many countries in this set had 50 (or more) different regions that had different rules applied at different times? (Italy had some of that.)  Are these stringency steps when the last region applied them, or when the first did?

This needs a lot of work to pass the "smells like BS" test

Remind me, who on this board pointed out that any statistic based on quantity of positive tests was pretty much meaningless?  Or do we just pull that card out of our sleeve when we want to discount an outlier but ignore it for all the non-outliers.  (We have the same card for deaths, by the way.)

[Genuine Realist]

So, I found a very interesting dataset, the Oxford Government Response Tracker.  Basically, it tracks government action in response to the coronavirus.  The stringency index is based on Eight of the policy indicators (C1-C8) record information on containment and closure policies, such as school closures and restrictions in movement (the variables are: school closing, workplace closing, canceling public events, restrictions on gatherings, public transport closings, stay at home requirements, restrictions on internal movements, and international travel controls).  They show the relationship between the index and number of cases:

Uh, a country that is (say) 100x the population gets the same vertical scale on the left as the small country???  The scale on the right is some artificial scale, but I'd guess it is intended to be something proportional to the population.   So, if Lichtenstein did nothing and had all 38K population get sick, they would have looked ok just because the gray plot didn't go that high.  China would look worse, despite it having, what, 4 orders of magnitude more population.

Oranges versus barrels of apples.

And while someone is making this stuff up, how many countries in this set had 50 (or more) different regions that had different rules applied at different times? (Italy had some of that.)  Are these stringency steps when the last region applied them, or when the first did?

This needs a lot of work to pass the "smells like BS" test

Remind me, who on this board pointed out that any statistic based on quantity of positive tests was pretty much meaningless?  Or do we just pull that card out of our sleeve when we want to discount an outlier but ignore it for all the non-outliers.  (We have the same card for deaths, by the way.)

I had been wondering the same thing about using positive tests as a measure. 

Since the availability of testing varies by region and over time, wouldn't a better measure be hospitalizations?

[Genuine Realist]

So, I found a very interesting dataset, the Oxford Government Response Tracker.  Basically, it tracks government action in response to the coronavirus.  The stringency index is based on Eight of the policy indicators (C1-C8) record information on containment and closure policies, such as school closures and restrictions in movement (the variables are: school closing, workplace closing, canceling public events, restrictions on gatherings, public transport closings, stay at home requirements, restrictions on internal movements, and international travel controls).  They show the relationship between the index and number of cases:

Uh, a country that is (say) 100x the population gets the same vertical scale on the left as the small country???  The scale on the right is some artificial scale, but I'd guess it is intended to be something proportional to the population.   So, if Lichtenstein did nothing and had all 38K population get sick, they would have looked ok just because the gray plot didn't go that high.  China would look worse, despite it having, what, 4 orders of magnitude more population.

Oranges versus barrels of apples.

And while someone is making this stuff up, how many countries in this set had 50 (or more) different regions that had different rules applied at different times? (Italy had some of that.)  Are these stringency steps when the last region applied them, or when the first did?

This needs a lot of work to pass the "smells like BS" test

Remind me, who on this board pointed out that any statistic based on quantity of positive tests was pretty much meaningless?  Or do we just pull that card out of our sleeve when we want to discount an outlier but ignore it for all the non-outliers.  (We have the same card for deaths, by the way.)

I had been wondering the same thing about using positive tests as a measure. 

Since the availability of testing varies by region and over time, wouldn't a better measure be hospitalizations?

[Farm93]

So, I found a very interesting dataset, the Oxford Government Response Tracker.  Basically, it tracks government action in response to the coronavirus.  The stringency index is based on Eight of the policy indicators (C1-C8) record information on containment and closure policies, such as school closures and restrictions in movement (the variables are: school closing, workplace closing, canceling public events, restrictions on gatherings, public transport closings, stay at home requirements, restrictions on internal movements, and international travel controls).  They show the relationship between the index and number of cases:

Uh, a country that is (say) 100x the population gets the same vertical scale on the left as the small country???  The scale on the right is some artificial scale, but I'd guess it is intended to be something proportional to the population.   So, if Lichtenstein did nothing and had all 38K population get sick, they would have looked ok just because the gray plot didn't go that high.  China would look worse, despite it having, what, 4 orders of magnitude more population.

Oranges versus barrels of apples.

And while someone is making this stuff up, how many countries in this set had 50 (or more) different regions that had different rules applied at different times? (Italy had some of that.)  Are these stringency steps when the last region applied them, or when the first did?

This needs a lot of work to pass the "smells like BS" test

Remind me, who on this board pointed out that any statistic based on quantity of positive tests was pretty much meaningless?  Or do we just pull that card out of our sleeve when we want to discount an outlier but ignore it for all the non-outliers.  (We have the same card for deaths, by the way.)

I had been wondering the same thing about using positive tests as a measure. 

Since the availability of testing varies by region and over time, wouldn't a better measure be hospitalizations?

I have a number of posts noting positive tests are problematic, but I am sure others noted it too since there are lots of flaws with test level comparisons.  Beyond when testing was available to citizens, you have to account for testing protocols, test result disclosure, skills of those conducting the tests, and reliability of the tests just to name a few.

Hospitalizations across a country are likely comparable, but even that could be flawed.
In CA or AZ, with a relatively large population of undocumented immigrants, a number of people in the population are likely to be very slow to visit a hospital if they or household members might face immigration consequences.
In other places, like states that resisted Obamacare, there might be more uninsured people.   Again the impact might be that a number of citizens will only appear if they are rather close to death since a few days in a hospital bed or a few hours in an ICU bed would bankrupt many.
Also some states limited tests to very specific populations while other states had drive-thru testing available to almost everyone in April.

Those are just some of the issues of looking at hospitalization rates across different states in the USA.

Global comparisons are even more problematic.
Korea citizens are known for going to doctors and utilizing their very low out of pocket healthcare system at the first hint of a problem.   A patient that tests positive from the MD visit would likely get sent to the hospital much sooner than in the USA.

In China, pharmacies are located in hospitals and many doctor visits are conducted there too.   Logical to conclude that those presenting with COVID19 like symptoms would be sent to an isolation unit in the hospital and not even allowed to go home to collect a few items.    

So in this example Korea and China hospitalization rate comparisons to the USA would be deceptive even if each country was tracking and publicizing results in comparable ways.  Korea is disclosing things in almost real time, while China is....well....less forthcoming.

Even deaths across different countries is an issue since politicians in many countries often have incentives to misrepresent the true tally.

It all gets messy, but I think deaths is best if hospitalization protocols are materially different between two population groups.   The protocols for hospitalization in the USA are similar enough that state v state comparisons of hospitalizations is likely viable.

[Farm93]

So, I found a very interesting dataset, the Oxford Government Response Tracker.  Basically, it tracks government action in response to the coronavirus.  The stringency index is based on Eight of the policy indicators (C1-C8) record information on containment and closure policies, such as school closures and restrictions in movement (the variables are: school closing, workplace closing, canceling public events, restrictions on gatherings, public transport closings, stay at home requirements, restrictions on internal movements, and international travel controls).  They show the relationship between the index and number of cases:

Uh, a country that is (say) 100x the population gets the same vertical scale on the left as the small country???  The scale on the right is some artificial scale, but I'd guess it is intended to be something proportional to the population.   So, if Lichtenstein did nothing and had all 38K population get sick, they would have looked ok just because the gray plot didn't go that high.  China would look worse, despite it having, what, 4 orders of magnitude more population.

Oranges versus barrels of apples.

And while someone is making this stuff up, how many countries in this set had 50 (or more) different regions that had different rules applied at different times? (Italy had some of that.)  Are these stringency steps when the last region applied them, or when the first did?

This needs a lot of work to pass the "smells like BS" test

Remind me, who on this board pointed out that any statistic based on quantity of positive tests was pretty much meaningless?  Or do we just pull that card out of our sleeve when we want to discount an outlier but ignore it for all the non-outliers.  (We have the same card for deaths, by the way.)

I had been wondering the same thing about using positive tests as a measure. 

Since the availability of testing varies by region and over time, wouldn't a better measure be hospitalizations?

I have a number of posts noting positive tests are problematic, but I am sure others noted it too since there are lots of flaws with test level comparisons.  Beyond when testing was available to citizens, you have to account for testing protocols, test result disclosure, skills of those conducting the tests, and reliability of the tests just to name a few.

Hospitalizations across a country are likely comparable, but even that could be flawed.
In CA or AZ, with a relatively large population of undocumented immigrants, a number of people in the population are likely to be very slow to visit a hospital if they or household members might face immigration consequences.
In other places, like states that resisted Obamacare, there might be more uninsured people.   Again the impact might be that a number of citizens will only appear if they are rather close to death since a few days in a hospital bed or a few hours in an ICU bed would bankrupt many.
Also some states limited tests to very specific populations while other states had drive-thru testing available to almost everyone in April.

Those are just some of the issues of looking at hospitalization rates across different states in the USA.

Global comparisons are even more problematic.
Korea citizens are known for going to doctors and utilizing their very low out of pocket healthcare system at the first hint of a problem.   A patient that tests positive from the MD visit would likely get sent to the hospital much sooner than in the USA.

In China, pharmacies are located in hospitals and many doctor visits are conducted there too.   Logical to conclude that those presenting with COVID19 like symptoms would be sent to an isolation unit in the hospital and not even allowed to go home to collect a few items.    

So in this example Korea and China hospitalization rate comparisons to the USA would be deceptive even if each country was tracking and publicizing results in comparable ways.  Korea is disclosing things in almost real time, while China is....well....less forthcoming.

Even deaths across different countries is an issue since politicians in many countries often have incentives to misrepresent the true tally.

It all gets messy, but I think deaths is best if hospitalization protocols are materially different between two population groups.   The protocols for hospitalization in the USA are similar enough that state v state comparisons of hospitalizations is likely viable.

[Genuine Realist]

So, I found a very interesting dataset, the Oxford Government Response Tracker.  Basically, it tracks government action in response to the coronavirus.  The stringency index is based on Eight of the policy indicators (C1-C8) record information on containment and closure policies, such as school closures and restrictions in movement (the variables are: school closing, workplace closing, canceling public events, restrictions on gatherings, public transport closings, stay at home requirements, restrictions on internal movements, and international travel controls).  They show the relationship between the index and number of cases:

Uh, a country that is (say) 100x the population gets the same vertical scale on the left as the small country???  The scale on the right is some artificial scale, but I'd guess it is intended to be something proportional to the population.   So, if Lichtenstein did nothing and had all 38K population get sick, they would have looked ok just because the gray plot didn't go that high.  China would look worse, despite it having, what, 4 orders of magnitude more population.

Oranges versus barrels of apples.

And while someone is making this stuff up, how many countries in this set had 50 (or more) different regions that had different rules applied at different times? (Italy had some of that.)  Are these stringency steps when the last region applied them, or when the first did?

This needs a lot of work to pass the "smells like BS" test

Remind me, who on this board pointed out that any statistic based on quantity of positive tests was pretty much meaningless?  Or do we just pull that card out of our sleeve when we want to discount an outlier but ignore it for all the non-outliers.  (We have the same card for deaths, by the way.)

I had been wondering the same thing about using positive tests as a measure. 

Since the availability of testing varies by region and over time, wouldn't a better measure be hospitalizations?

I have a number of posts noting positive tests are problematic, but I am sure others noted it too since there are lots of flaws with test level comparisons.  Beyond when testing was available to citizens, you have to account for testing protocols, test result disclosure, skills of those conducting the tests, and reliability of the tests just to name a few.

Hospitalizations across a country are likely comparable, but even that could be flawed.
In CA or AZ, with a relatively large population of undocumented immigrants, a number of people in the population are likely to be very slow to visit a hospital if they or household members might face immigration consequences.
In other places, like states that resisted Obamacare, there might be more uninsured people.   Again the impact might be that a number of citizens will only appear if they are rather close to death since a few days in a hospital bed or a few hours in an ICU bed would bankrupt many.
Also some states limited tests to very specific populations while other states had drive-thru testing available to almost everyone in April.

Those are just some of the issues of looking at hospitalization rates across different states in the USA.

Global comparisons are even more problematic.
Korea citizens are known for going to doctors and utilizing their very low out of pocket healthcare system at the first hint of a problem.   A patient that tests positive from the MD visit would likely get sent to the hospital much sooner than in the USA.

In China, pharmacies are located in hospitals and many doctor visits are conducted there too.   Logical to conclude that those presenting with COVID19 like symptoms would be sent to an isolation unit in the hospital and not even allowed to go home to collect a few items.    

So in this example Korea and China hospitalization rate comparisons to the USA would be deceptive even if each country was tracking and publicizing results in comparable ways.  Korea is disclosing things in almost real time, while China is....well....less forthcoming.

Even deaths across different countries is an issue since politicians in many countries often have incentives to misrepresent the true tally.

It all gets messy, but I think deaths is best if hospitalization protocols are materially different between two population groups.   The protocols for hospitalization in the USA are similar enough that state v state comparisons of hospitalizations is likely viable.

It's not my area of law, but I believe the requirement imposed on most health institutions to provide emergency care regardless of insurance would go a long way towards normalizing hospitalization stats.

[Genuine Realist]

So, I found a very interesting dataset, the Oxford Government Response Tracker.  Basically, it tracks government action in response to the coronavirus.  The stringency index is based on Eight of the policy indicators (C1-C8) record information on containment and closure policies, such as school closures and restrictions in movement (the variables are: school closing, workplace closing, canceling public events, restrictions on gatherings, public transport closings, stay at home requirements, restrictions on internal movements, and international travel controls).  They show the relationship between the index and number of cases:

Uh, a country that is (say) 100x the population gets the same vertical scale on the left as the small country???  The scale on the right is some artificial scale, but I'd guess it is intended to be something proportional to the population.   So, if Lichtenstein did nothing and had all 38K population get sick, they would have looked ok just because the gray plot didn't go that high.  China would look worse, despite it having, what, 4 orders of magnitude more population.

Oranges versus barrels of apples.

And while someone is making this stuff up, how many countries in this set had 50 (or more) different regions that had different rules applied at different times? (Italy had some of that.)  Are these stringency steps when the last region applied them, or when the first did?

This needs a lot of work to pass the "smells like BS" test

Remind me, who on this board pointed out that any statistic based on quantity of positive tests was pretty much meaningless?  Or do we just pull that card out of our sleeve when we want to discount an outlier but ignore it for all the non-outliers.  (We have the same card for deaths, by the way.)

I had been wondering the same thing about using positive tests as a measure. 

Since the availability of testing varies by region and over time, wouldn't a better measure be hospitalizations?

I have a number of posts noting positive tests are problematic, but I am sure others noted it too since there are lots of flaws with test level comparisons.  Beyond when testing was available to citizens, you have to account for testing protocols, test result disclosure, skills of those conducting the tests, and reliability of the tests just to name a few.

Hospitalizations across a country are likely comparable, but even that could be flawed.
In CA or AZ, with a relatively large population of undocumented immigrants, a number of people in the population are likely to be very slow to visit a hospital if they or household members might face immigration consequences.
In other places, like states that resisted Obamacare, there might be more uninsured people.   Again the impact might be that a number of citizens will only appear if they are rather close to death since a few days in a hospital bed or a few hours in an ICU bed would bankrupt many.
Also some states limited tests to very specific populations while other states had drive-thru testing available to almost everyone in April.

Those are just some of the issues of looking at hospitalization rates across different states in the USA.

Global comparisons are even more problematic.
Korea citizens are known for going to doctors and utilizing their very low out of pocket healthcare system at the first hint of a problem.   A patient that tests positive from the MD visit would likely get sent to the hospital much sooner than in the USA.

In China, pharmacies are located in hospitals and many doctor visits are conducted there too.   Logical to conclude that those presenting with COVID19 like symptoms would be sent to an isolation unit in the hospital and not even allowed to go home to collect a few items.    

So in this example Korea and China hospitalization rate comparisons to the USA would be deceptive even if each country was tracking and publicizing results in comparable ways.  Korea is disclosing things in almost real time, while China is....well....less forthcoming.

Even deaths across different countries is an issue since politicians in many countries often have incentives to misrepresent the true tally.

It all gets messy, but I think deaths is best if hospitalization protocols are materially different between two population groups.   The protocols for hospitalization in the USA are similar enough that state v state comparisons of hospitalizations is likely viable.

It's not my area of law, but I believe the requirement imposed on most health institutions to provide emergency care regardless of insurance would go a long way towards normalizing hospitalization stats.