We Should be Preparing for COVID Hospitalizations to Double Within a Matter of Weeks
Over the past few months, I have said and written that we should look to Europe's experience with COVID to anticipate what may subsequently happen in the United States. As you may recall, in the spring it wasn't long before the tragedy that befell Lombardy, Italy was reflected in parts of the Mid-Atlantic region of the United States.
Over the summer I became particularly interested in studying COVID in France. Newly confirmed-cases there started rising in mid-July while the total number of people hospitalized continued to fall. If that were to have continued, then it may have boded well for those of us in the United States.
But it did not continue. Hospitalizations in France started rising slowly in mid-September, spiked in mid-October, and only now show some signs of decelerating. This should be of great concern to the US for a very specific reason. For the past fourteen weeks, the rise and fall of the number of COVID hospitalizations in the United States have tracked those of France four weeks prior. The relationship has been extraordinarily consistent over all that time.
If that relationship holds true in the coming weeks, which I believe to be a possibility, then hospitalizations are poised to accelerate in a way the American people, and perhaps their government, are not expecting. So, I have summarized this research for a number reasons.
Provide an objective framework for understanding how COVID has progressed through the US and other countries.
Motivate people to continue to actively consider the risks their activities pose for themselves as other people, particularly during the coming holidays.
Ensure policy makers are or remain aware of the need to prepare for a sudden acceleration of COVID hospitalizations and deaths.
France and the US
The chart below shows COVID hospitalizations per 100k residents over time in France (white line) and the US (blue line). This chart relates that in terms of hospitalizations, France's experience with COVID in the spring was over twice as severe as that of the US. No one in New York City would believe that, nor should they.
First, COVID was felt very unevenly during the early months of the crisis. Second, France started consolidating hospital data from its departments in March. While 44 states had started reporting hospital data by the end of April, it wasn't until July until all had. This isn't as large a problem with understanding the situation in the US as it might appear because, generally the states with the most severe outbreaks were the ones that started reporting the data the earliest.
It doesn't impact this analysis at all because we're concerned with the right hand portion of the chart. That is, what occurred as hospitalizations in both countries fell during the mid-to-late summer and then subsequently rose later in the year.
Notice that the US curve from its summer peak in late July to the present looks remarkably similar to the shape of France's curve. As the question in chart poses, what would happen if we shifted France's curve forward a few weeks?
Well, you'd get the chart below, which is one of the most uncanny relationships I've experienced in my professional career. You can analyze any subset of these data, and the relationship remains stable.
If you regress the daily observations against one another, you get a coefficient of determination (R^2) of over 97%. To the extent that can be put in English that means over 97% of the change in direction of one of the lines can be predicted solely by reference to the other line. That's amazing.
You've already guessed that I'm going to conclude that because hospitalizations have escalated in France that's the reason I think they are going to escalate here. And you're right, but that alone wouldn't have prompted me to believe so strongly in that conclusion.
Productive discussion of COVID in the United States is almost impossible because no matter where one starts, another will immediately question one's motives. Sometimes this becomes comical. I gave a COVID talk in early September to an audience of varied professional backgrounds. Afterward, I learned that one asked whether I was a "Trumper," and another asked permission to send the information to an individual in the Biden campaign because they thought it would be helpful to them.
I don't think we can help but see COVID through a political lens. Ultimately, I decided to look at the two comments as a sort of collective compliment. I have strenuously tried to keep my analyses as neutral as possible and to present them in ways that allow readers to draw their own conclusions. I hope that's a fair way to look at it.
Which is why starting the discussion in France will be helpful in understanding the course of the contagion without the information being immediately contorted.
COVID in France
Metropolitan France (the part of France that is in Europe) is the home of 65 million people who reside in one of 96 departments. 94 of these departments are located on mainland Europe. The island of Corsica holds the other two. As in the United States, the departments were not hit equally by COVID in the spring.
I have grouped these departments into three segments based on how severely the were impacted in the spring. I creatively named these segments High (12 departments), Medium (22 departments), and Low (68 departments). Severity is measured by the number of hospitalizations per 100k residents.
For reference the outbreak was 5X as severe in the departments in High segment as they were in the Low segment and nearly 2X that of the Medium segment. As I mentioned above, the severity of the pandemic is measured by the percentage of residents of an area who are hospitalized. Severity tells you where resources are likely to be the most strained. Magnitude is a simple count of the number of hospitalizations. It tells you where the most people are impacted. These two measures are often correlated, but they measure two different things.
I've defined the above segments by severity for a specific reason. If you define them by magnitude, you'll just end up with segments defined more by population than they are by COVID--which really isn't a very meaningful thing to do when you're trying to understand COVID.
The chart below shows the severity of the pandemic in these three segments from March through yesterday. It demonstrates how dramatically unevenly COVID hit France in the spring. In addition to that difference there are four other important things to note in this chart:
The High segment departments had more than 3X the population density than the least-impacted departments.
France's departments experienced a spring wave, no summer wave, and all the Segments are now experiencing a strong fall wave.
The fall wave is already stronger for departments in the Low segment than the spring wave was. This will be so for the departments in the Medium segment within a matter of days. This means the fall wave will be worse than the spring wave for at least 80% of the French people.
The severity of the pandemic is now much more-equally distributed throughout France. Currently, the severity of outbreak in the most-impacted segment (Medium) is less than 50% greater than that of the least-severely impacted (Low)
This fourth factor is absolutely critical to understanding why France's fall outbreak is different than the one in the spring. As the pandemic subsided over the summer, COVID quietly embedded itself in the less-densely populated portions of the country.
Among other reasons, as the weather turned cooler and people headed back indoors COVID returned with near-equal ferocity in most parts of the country. That's what viruses do. People who live in cities aren't significantly more likely to die of the flu than people residing in outlying areas. After awhile it mostly evens out. COVID is getting there fairly quickly.
COVID's dispersion is particularly difficult for France because 58% of its population resides in the Low segment. This is where magnitude matters. You can clearly see the implications of this in the chart below. The magnitude of the outbreak has reversed. It is is now greatest in the Low segment and least in the High segment
France is not an Exception in Europe
If France's experience with COVID were unique in Europe, then linking its experience to the United States would have much less power. But that isn't the case. Every country of Europe doesn't publish comprehensive hospitalization statistics, but they all publish the number of COVID deaths, which is depicted below for the 10 Western European nations with populations greater than 10 million people (white line) as well as for The US (blue line).
France's experience is right in line with the rest of Europe's. The number of COVID deaths in these select European countries has increased by 520% in the past four weeks and by 65% in just the past week.
The comparable figures in the United States are 32% and 14%, which is remarkable when you think about it. Unless there is some natural barrier or set of actions the US and its people have taken to do so well compared with the Europeans, the burden of proof may very well be on others to explain why our experience on this side of The Atlantic could remain so different.
The United States
The US' experience with COVID is similar to that of France, but not the same. Like, France, the spring outbreak was strongest in densely-populated states and weakest in low-density states, but unlike France, not all parts of the country had their first waves in the spring.
Rather than three segments defined by the severity of the outbreak in the spring, the US has five segments defined by the severity and timing of their initial outbreaks. The five segments are defined as:
Spring Wave: States that had a severe outbreak in the spring, followed by a long, slow decline through the summer.
Summer Wave: States that had their first severe outbreak in the summer.
Dual Wave: States that had outbreaks in both the spring and the summer, but of lesser degrees than the spring and summer wave states.
California: Due to its size, and its (until recently) falling hospitalization rates, it merits a segment all to its own.
Fall Wave: States that had relatively-less severe outbreaks in the spring and summer and appear to be currently on their way to their first peak.
The statistics for each follow and the chart of hospitalizations over time follows those:
Spring Wave: 10 States; Pop: 70 MM/389 per sq mi
Total Population Fatality Rate: 0.11%
Summer Wave: Total Population Fatality Rate: 0.07%
6 States; Pop: 71 MM/127 per sq mi
Dual Wave: 5 States; Pop: 38 MM/125 per sq mi
Total Population Fatality Rate: 0.08%
California: 1 State; Pop: 40 MM/254 per sq mi
Total Population Fatality Rate: 0.05%
Fall Wave: 29 States; Pop: 110 MM/ 47 per sq mi
Total Population Fatality Rate: 0.04%
As you can see the Wave segments don't resemble those of France at first glance, but there are critical similarities
The most densely populated states were the ones hit most severely in the spring after which they began a long-slow recovery
The segment least impacted early on (spring and summer) were the least densely-populated states
In the early parts of the epidemic the severity of the crisis varied tremendously.
COVID's severity is now much more evenly-distributed throughout the country.
Hospitalizations are now increasing in all segments
The chart below shows the magnitude of the outbreaks:
Like France, it's now the least densely populated areas of The United States where the plurality of the hospitalizations are.
Putting it all Together
Summarizing. . . Overall US COVID hospitalization trends are not only similar to those of France. So, are the underlying dynamics of the outbreaks. Namely, the expansion of the virus from high density areas to the rest of the two countries, the narrowing of regional differences in the pandemic's severity, and the recent upturn in the number of hospitalized COVID patients.
Despite the recent upturn in US hospitalizations, it has so far avoided the steep increases Europe has seen in COVID hospitalizations and deaths. The question is can it continue? I don't believe so for a couple of reasons:
Recent trends point to continuing, sharp increases in US COVID hospitalizations.
These trends are consistent with relationship we observed earlier between hospitalizations in the US and France and by extension the rest of Europe.
COVID's Momentum in the US
The increases in deaths and hospitalizations in Europe adhered to a similar pattern. A series of small, weekly increases, followed by a quick acceleration like that we saw in the spring and summer waves. The US is now in its eighth consecutive week of rising hospitalizations across the majority of states. In contrast, the spring and summer waves were driven by 10 and 6 states respectively.
Just a few facts should place recent increases in US hospitalizations in perspective.
The total increase in hospitalizations over the past eight weeks is greater than the total during the summer wave.
The week ended this past Saturday, October 31 saw the greatest increase in weekly hospitalizations since the week ended July 18.
The week ended this past Saturday November 7 saw the greatest increase in weekly hospitalizations since the week ended April 11.
Yesterday, Sunday November 8 saw the greatest increase in hospitalizations on a Sunday since April 5.
The momentum is tremendous, broad-based, and won't be easily reversed. The clearest way I can think of to demonstrate the strength and breadth of the current wave is to simply show you the data for hospitalization increases for the past two weeks, segmented by save and state.
The top chart below depicts net increases in net COVID hospitalizations for the past week. The bottom one for the week prior. Notice just how many states are showing increases and recall that the spring wave was driven primarily by 10 states and the summer wave by 6.
Also note that the increases for 37 states in the most recent week were greater than the increases for the prior week. As it is in Europe, the pandemic is gaining significant momentum in the United States
The question is not whether COVID is gaining momentum. The question is how fast and for how long. For that answer we really have no better guide than what has happened in France, a country which like the US also saw considerable hardship in the spring, a broadening of the virus' footprint over the summer, and a relaxed form of social distancing in many parts of the country.
I can understand that some might object to fashioning a model for the US based primarily on the relationship its data have to a country thousands of miles away. My answer to that is that a lot of the models you see are all over the place and using another country as a basis for comparison was one of the few things that worked in the spring.
I was an early critic of the model's touted by the White House in the late March and April. Within two weeks of their creation I concluded they were underestimating deaths on the way up their curves and underestimating them on the way down. I concluded in April they were systematically underestimating future deaths in the US.
One of the things the models did well was to estimate the timing of the peaks in the hard-hit states of the Northeast. And why was that? They concluded COVID's progress in states that quickly implemented lock downs would be similar to the curve experienced in Wuhan, China, the location of the original outbreak. It would be unfair of me not to note that predicting the course of this contagion in April was an impossible task, and that The Institute for Health Metrics and Evaluation, the creator of those forecasts, has very clearly revised its methodology.
Now we're getting down to it. Above is a repeat of the chart comparing the intensity of the pandemic (hospitalizations per 100k residents) in the US and France. Recall that to put the countries on the same stage of the the pandemic, we shifted the data for France forward by 28 days.
This means the data for France displayed on November 8 for the above chart is actually the data from October 11, 28 days prior. This also means that there are 28 days of hospitalization data from France that are literally off the chart to the right. These days represent the path the United States may take over the next four weeks.
So, let's put those days back on. The data for France depicted on December 6 is the data as of yesterday, Sunday, November 8. There are many ways to estimate how closely the US' future path with follow France's I'm just going to choose one that seems reasonable. The US projection in the the chart below (yellow line) preserves the vertical distance between the two curves as of November 8.
For the purpose of making the point in the headline of this article (that we should prepare for a near-term doubling of hospitalizations), all I need to have accomplished is for the reader to believe the shape of the yellow curve in the chart above is within reason. Let's assume I've achieved that. As the chart says, if the US follows this path, then it would have 21 hospitalizations per 100k people by November 15 and 43 by November 30. Here's the problem with that. Those numbers equate to 69,000 and 141,000 hospitalizations. Given the rate of acceleration we're now experiencing 69,000 the first figure seems well within reason. How about 141,000? That seems impossible.
But that's the problem with relying on conventional thinking when potting COVID's future. When France had 5,459 hospitalizations (8.4 per 100k residents) in mid-September while its government was strategizing how to to create plans to "live with the virus" they didn't envision that in a little over a month hospitalizations would have increased by over 200% to 19,891 (30.6 per 100k residents) by March 28 when they ordered another lock down.
A country and its people can't only prepare for what they think or hope will happen. They must prepare for the range of things that may happen, and while I certainly hope the calculations I made above are incorrect, I am planning my activities for the next few weeks with them in mind, and I am making the same recommendation to my friends, family, and anyone who has read this far. If you find my line of reasoning plausible, then please forward this article to anyone to whom you think will find it persuasive. I started analyzing COVID in January because I thought I might be able to do some good. Perhaps, this is it.
I'll conclude with my hope the US is preparing for this eventuality, but I have reason to doubt that. The CDC's consensus forecast (updated Nov 5, as of this writing) for the number of COVID deaths for the last week of November is approximately 7,000. You can see that in the chart to the right below. To the left you can see just how much uncertainty there is in their ensemble of forecasts, which isn't comforting.
Over the past few months, the daily number of COVID deaths equates to somewhere between 2.0 - 2.5% of the number of people hospitalized two weeks prior. If hospitalizations approach 70,000 by November 15, then the rate of deaths per week will be on the order of 10,000 - 12,000 by the end of the month. The upper part of that range is beyond any of the forecasts.
I wish we didn't have to think about a scenario like this, but if it comes to pass it would be nice to be prepared. As for what those preparations should be, I have believed since the early summer that much of the cure states have adopted is worse than the illness itself, but that's the subject for another article. But if you are open to that possibility, then please consider the point-of-view expressed in The Great Barrington Declaration.