Post #673: Reopening Georgia and elsewhere, Post #2

Above:  Our clear downward trajectory, justifying removal of restrictions.  (Yes, that’s sarcasm.  But that’s the actual data.Source:  Calculated from Johns Hopkins coronavirus database., accessed 4/29/2020.)

I’ll start by repeating what I said 20 days ago, in Post #616, 4/9/2020.  At that time, I predicted that eventually, some people would start saying this:

“Our containment strategies worked, and we avoided mass deaths. And so the lack of mass deaths is evidence that our containment strategies were not needed.”

Substitute “are” for “were”, above, and you now know everything there is to know about the logic behind re-opening the economy.  If you believe that the thinking behind this movement to re-open state economies goes one inch deeper than that, or draws from some sound empirical basis, you would be mistaken.

Maybe fortune will favor those who open early.  Maybe not.  Maybe the cost of additional deaths is outweighed by the economic gains.  Maybe not.   Maybe, ultimately, economic gains will be negative, if this slows down the ultimate recovery from this pandemic. Or, maybe there will be no uptick in infections and deaths.  Nobody knows.  It’s all a roll of the dice.

No matter how you slice it, this is exactly the sort of situation where preparing for failure matters.  Failure is always an option.  Everybody will celebrate if this is a success.  But is anybody preparing in the states where this fails?


Background

Source:  Clipart-library.com

Between the time I wrote part 1 of this post (4/25/2020) and today (4/29/2020), it looks like many states, largely those with Republican governors, have decided to follow Georgia’s lead, and begin relaxing restrictions on commerce and restrictions on individuals’ movement within their states.

Given this re-opening, I think there’s value in following through on the prior post, and answering these four questions.

1:  What, exactly, do the “Federal guidelines” for re-opening states say?

2: Did Georgia meet the basic criterion for re-opening businesses under those guidelines? 

3: How will Georgia tell if this is working out or not, and what criteria will be used to “pull the plug” on this re-opening, if necessary?  How tough will it be to do that?

4:  But, at the very least, once Georgia does this, we’ll at least be able to tell whether or not it works, right? 


Briefly …

I’ll just go ahead and get my prejudices out in the open.  Near as I can tell, the Federal “guidelines” are almost purely a political document.  They provide cover for governors who wish to re-open businesses now.  They don’t provide much, if any, any actual guidance, beyond common sense.

One dead giveaway is the total and complete absence of numbers.  Other than “14”, a nice round two weeks.  This means that deciding whether or not you are adhering to the guidelines is purely a matter of judgment.  It’s certainly not a matter of actually measuring your state’s performance and outcomes, and comparing that to agreed-upon thresholds. (And I’m far from the only one to have noted this.  This is a big-business think tank’s take on those guidelines, calling out their lack of quantification.)

The Federal guidelines for re-opening the economy mix a few ill-defined ad-hoc rules with a few nuggets of common sense.  Based on the generally poor grammar, the numerous undefined terms, and the use of many dramatic words (“surge”, “crisis”), I’d have to guess that this didn’t take long to write.

There really is no more depth than that.  You probably think I’m kidding, because you think that nobody would take this kind of a roll of the dice, based on so little thought.  Let me just repeat the Federal government’s guidance to you, the citizen, from that document.  Ready?  Here’s what the Federal government wants you to do:

CONTINUE TO PRACTICE GOOD HYGIENE.  PEOPLE WHO FEEL SICK SHOULD STAY HOME.

Good advice in any era, for sure.  But do you feel emboldened to resume pre-COVID lifestyle now?

Bottom line:  If you think there might be somebody in charge who knows what they are doing, and that’s the reason for re-opening these economies now, you should probably re-think that. It’s possible that nobody could develop a good plan for economic re-opening, given the novelty of the situation.  It’s completely clear that we have not done so.

So this is, near as I can tell, a complete and total roll of the dice.  Either way.  Either re-opening businesses, or keeping them shut.  It appears based on no empirical evidence whatsoever.  There are no quantifiable guidelines.  It’s just a bunch of vague suggestions that seem plausible.

Given that reality, the stupidest possible way to go about this is to pretend that it’s not a gamble, roll the dice, and then refuse to admit it if you lose.  But my guess is that this is the direction we’re heading.  As I noted in an earlier post, it looks like there is no plan in case of failure.  And, possibly, no way for a governor to admit failure.  But we won’t know for about a month, because it will take that long for the impact of this to begin trickling into the data.


1  What do the Federal re-opening guidelines say?

Source:  Clipart-library.com

In so far as I can tell, the total content of the Federal guidelines for “re-opening America” is in this small .pdf.  If there’s more to it than this, I have yet to find it.  Let me place a copy here.

Guidelines-PDF

 

Phase I:  Something about your count of cases has to be falling.  The US does not track people who are cured (other than those discharged from the hospital), we have to infer that they are requiring a “downward trajectory” in daily new cases.  This is a requirement that the number of new cases discovered each day is falling.

A state should meet the following criteria before re-opening business:

  1. Downward trajectory in new cases with symptoms of influenza AND separately, those with with symptoms of COVID-19;  AND
  2. Downward trajectory in cases formally diagnosed with COVID-19 OR a downward trajectory in the fraction of all tested cases that test positive; AND
  3. Hospitals are not providing crisis care, AND you have a robust system for testing at-risk health care workers for COVID-19.

The first point I would make is that nothing is quantified, or even defined.  None of the italicized terms is defined and there are no numerical thresholds  specified.  Basically, as long as you can look at some trend line, and it looks less-than-horizontal to you, then you’re good to proceed.

This appears to rule out states like Virginia, where daily new cases continue to rise.  But not much else.

Why fourteen days?  That doesn’t appear to be based on any statistical analysis of length of random variations in trends, and it certainly isn’t based on the biology (median 5 day incubation period).  I think that’s based on being a good round number, full stop.

Why no criterion for new cases per 100,000 population?  If you’re worried about additional face-to-face contact spreading the virus, the criterion ought to be based on the likelihood that any one person is infectious-but-asymptomatic.  Instead, no matter how high the new case count, under these guidelines, as long as it appears to be dropping, you are good to go.

The crazy thing is, that’s the well-defined section.  As you go further into the criteria for re-opening businesses, it’s just word salad.  It’s a bunch of good-sounding but totally meaningless phrases.  Just a sampling:

Ability to quickly and independently supply sufficient Personal Protective Equipment and critical medical equipment to handle dramatic surge in need.

Do I really need to say this?  Nothing in that statement is defined or quantified.  Quickly?  Sufficient?  Critical?  Dramatic?  Surge?  This is just a bunch of urgent–sounding words strung into sentences.  There is no way to tell whether or not a state meets this criterion.

And the rest of the criteria are just like that.  One more:

Ability to quickly set up safe and efficient screening and testing sites

Quickly?  Safe?  Efficient?  (Why efficient?)  They don’t even bother to suggest how many you might reasonably need.  Some, I guess.  So if you have plans for two or more, you have this one covered.

Ability to surge ICU capacity

Is “surge” really a transitive verb now?  In any case, what does that mean?  How could a state tell if it qualifies?  It can’t, plain and simple.

By contrast:  If somebody said, you need to have a written plan for expanding staffed ICU beds by 20% with one week’s notice, signed by the relevant parties, that would be a criterion.  But “ability to surge” ICU capacity?  That’s meaningless.

You get the drift.  The whole “criteria” section reads like that.

Worse, you can go on to additional phases as long as there is no evidence of “rebound”.  And there is no definition of what “rebound” means.  Not merely no quantitative threshold.  Literally no hint as to (e.g.) how long the numbers have to be rising before you call that a “rebound”.

The advice to persons and businesses is nothing more than common sense, plus what the CDC recommends.  So, the direction to individuals is “CONTINUE TO PRACTICE GOOD HYGIENE” and “PEOPLE WHO FEEL SICK SHOULD STAY HOME”.

Tough to argue with that under any circumstances.

Near as I can tell, the only section that has any specifics is the guideline as to which businesses can open at which phases.

  • Under Phase I, bars and schools have to remain closed.  Nursing homes have no visitors.  All others open, with some restrictions.
  • Under phase II, bars and schools can re-open.  Nursing homes still have no visitors. All others open, with fewer restrictions.
  • Under Phase III, everything is open.

Makes me wonder what they have against bars, other than the fact that the President doesn’t drink.  And, given the incredibly low rates of serious infection among children, and the Chinese experience that children were, by and large, not disease vectors, it makes me wonder about the basis for opening schools being the last thing you do.

One thing in their favor is that they tell low-risk individuals to take their chances first.  This matches my seed-potato analogy (Post #G00) for who should re-start the economy.

Or, rather, they tell “vulnerable” individuals to stay home, and, wonder-of-wonders, that’s the one term in the entire document that they bother to define.  A vulnerable individual is one who is elderly (not defined beyond that), or anyone with high blood pressure, chronic lung disease, diabetes, obesity, asthma, or compromised immune system.  This appears well-grounded, as that is the list of conditions that appear to be risk factors for hospitalization and death from COVID-19, derived from statistical analysis of both China and US hospitalizations.

I doubt that the writers actually understood the medical definition of obesity, because that rules out 42% of US adults.  Clearly, if people paid attention, the combination of those restrictions would keep the majority of Americans at home.  I doubt that’s the intent here.

In conclusion, this isn’t a serious document.  Not in the sense of providing needed technical guidance based on facts and science.  I can only guess it was tacked together by people who wanted to sound serious, hence the many dramatic words.

If you’ve ever read or tried to implement government regulations, you realize this is just a bunch of vague phrases.  They may sound good to people who don’t know much.  But for those of us who have dealt with (e.g.) Medicare regulations, this is useless.  All show, no substance.


2: Did Georgia meet the basic criterion for re-opening businesses under those guidelines? 

 

No.  Judge for yourself. Here’s their daily new case count as of the time they decided to re-open.  That’s Georgia on the top, versus Virginia on the bottom.  The big “?” is meant to symbolize “wonder what’s going to happen next”. Doesn’t look like a “downward trajectory” to me.

But, in all fairness, that’s every bit as much of a “downward trajectory” as the national data at the top of the page are.


3: How will Georgia tell if this is working out or not, and act on that?

Source:  Clipart-library.com

Per the guidelines, they will look for evidence of “a rebound”.  How is that quantified?  How big a rebound is needed to trigger reversal of this?  What happens in the event of a rebound.  None of that is spelled out.

But if you were serious about this, you’d have to deal with two factors:  Time lags, and random variation.

Time lag.

The time lag is what I discussed in Post #G01, the hog-slaughter cycle Some weeks are going to elapse between the “re-opening”, and the point at which results from that will begin to appear in the data.  That’s from the biology (median five day incubation period), the health care system (median of maybe six days from onset of symptoms to delivery of test results), and from the tentative acceptance of this by the population. Business didn’t return to normal in a day.  It’s going to ramp up over time.

This long lag between the time you act, and the time you see the results, generates a classic problem of control theory.  You can’t do a course correction until you see where you are headed.  And that can be a recipe for a wreck.

Rather than get into the details, let me just give a simple analogy of driving a car.

Suppose you are driving a car where there’s a one-minute lag between the time you hit the gas, hit the brake, or turn the wheel, and time that the car responds.  You could still drive that car, as long as things proceed very slowly, and there are no unexpected obstacles in your way.

You could even drive that in traffic, as long as everybody played by the same rules.  If all the other cars were just as slow and cautious as you are, you could probably eventually figure out an approach to controlling the car that would mostly work.

But if you’re in fast traffic, this would be a disaster.  That is, if some other actors in the system are moving quickly, not subject to time lags, you wouldn’t be able to react in time.  If other actors in the system move much faster than you, or obey a different set of rules, you’ll eventually end up on a collision course and be unable to correct your path in time.

COVID-19 is a fast actor, when unconstrained.  And that’s the potential problem here.  Originally, it was thought that each infected individual managed to infect an average of about 2.3 others.  But that was based on the Chinese experience, where, among other things, everybody wore masks when in public.  Later, that estimate was tentatively (and perhaps incorrectly) raised to infecting about 5 others, before social distancing and lockdown.

The whole point of social distancing and lockdown was to reduce the speed of spread.  That is, reduce that new-infections number.  Ideally, to get it below 1.0, so that the pandemic would eventually die off.  Needless to say, we haven’t gotten it below 1.0 yet in Virginia, but that may have occurred in New York.

As we remove these restrictions, we’re going to increase the rate of infection spread, some.  We’re going to speed up COVID-19.  We may or may not do that to the point where “a rebound” occurs.  But with the time lags involved, you’re going to have three or four weeks of “rebound” in the pipeline before anybody has an inkling that this may have gone wrong.

The states are, in effect, deciding to drive that crippled car in fast traffic.  Whether or not one or more states is on a collision course, it’ll take us maybe a month to find out.

Random variation.

Random variation is the other enemy to understanding what’s happening.    The thumbnails above illustrate that, in two ways.

First, each state is going to have its own pre-existing trend.  So it’s not as if the default trend, absent any action, is “flat”.  It is what it is, and it varies across states.

Second, if you follow the ups and downs of the bars on these graphs, you can see that there often is no clear trend.  Or worse, what appears to be a trend varies.  One week, it looks like “downward trajectory”.  A few days later, it looks like “rebound”.

With that as the background, it takes a considerable length of time before you can confidently spot a change in the trend.  Here’s the data for New York State, for which there appears to be agreement that it has, for now, passed the peak of new cases.  At what point along that line would you have been confident in saying that?  As of 4/21/2020, maybe.  As of 4/25/2020 (last peak), maybe not.

For both of these reasons, determining exactly what is happening, when, is far from clear-cut.  That’s just the way health care data tend to be, and there’s no cure for it.  You only know what the trend is when you are looking at it in your rear-view mirror.  It will take some time before any change in trend, from re-opening these states, becomes clear and apparent.


4:  But, at the very least, once Georgia does this, we’ll at least be able to tell whether or not it works, right? 

In a word, no.  But this is an extras-for-experts on statistical analysis.  To be honest, nobody in my readership will care.  So let me just state a few things without references.

What we have here is a “natural experiment”.  Some states choose to re-open.  Others don’t.  They each re-open in their own unique way.  In some states, people will enthusiastically do some things.  In others, a few may reluctantly do others.  Each state will have its own pre-existing trend.

And the goofiest thing for people to understand is that we don’t really even have 50 states’ worth of independent data.  If state populations were equal, we’d have “50 degrees of freedom”.  But in fact, CA, TX, FL and NY together contain more than a third of the US population.  At some level, what happens in Alaska is more-or-less rounding error, and doesn’t really much matter in any overall statistical analysis.

The only practical upshot is that even if this re-opening policy flops, those who favor it will be able to fuzzy up the numbers for weeks after any change in trend is apparent. It’s not a situation that lends itself to clarity.  Given who the actors are here, it’s a given that there will be people eager to take advantage of that.