From early on in the pandemic, there have been two sites that have competed to offer a view of the future, offering predictions of where COVID-19 cases and deaths go over coming months. One of these, the IHME projection from the Institute for Health Metrics and Evaluation (IHME) at the University of Washington. That projection, which is not really a model in the traditional sense, has been extremely sensitive to current conditions and prone being off by a significant degree—going under when the cases are on the increase, and overshooting as cases decline. That model currently projects that the United States will still be seeing very little decline in cases, hospitalizations, or deaths right through the spring of 2022. It also projects that another 100,000 Americans will die this year of COVID-19.
The IHME numbers have gained a lot of attention over the last two years, in part because it was that site that was first waved in front of Trump to make it clear that COVID-19 posed a serious threat, and that hundreds of thousands might die. However, there is another site, the COVID Scenario Modeling Hub (CSMH), which runs a series of competing models from different organizations and universities. Taking a consensus model from that site has, in most instances, produced a much more reasonable prediction of where things are going next.
And let’s hope that’s the case now. Because the model being thrown up at the CSMH is drastically different at this point than the future projected by IHME. Taking into account both the possibility of childhood vaccinations being available sometime in the next two months, and anticipating that no “super variant” that is more contagious than delta and able to resist current vaccines appears, the modeling hub projects that cases in the United States are at the start of a steep decline. Their models suggest that by the end of the year the United States could be back to a point roughly where it was when the delta wave began, and by the beginning of spring 2022, cases could be well below 10,000 a day.
That’s not “back to normal,” but it’s much better than where we are today.
At the beginning of July, the average number of daily cases of COVID-19 was down to around 12,000 a day. That was the lowest number since cases first began to ramp up in March 2020, and the way the numbers had come down from a high of over 250,000 a day at the start of the year made it seem as if COVID-19 wasn’t just declining, but on its way out the door.
Because IHME is simply a projection of current trends, what it showed for the future at the end of May appeared extremely rosy.
At the end of May, IHME projected the number of new cases of COVID-19 dropping sharply through the summer.
But even as the United States was experiencing that low, officials in the U.K. were warning that the dominant variant in that country was moving from the version that had recently been renamed alpha to the now all too familiar delta. Within an astoundingly short period, cases there swelled back to a peak that nearly matched the worst of the mid-winter wave. And then the U.S. started following the same track, just a couple of weeks behind, as delta became dominant across the nation and areas like rural Missouri exploded with new cases. This new “pandemic of the unvaccinated” became a lingering event that just kept generating more and more cases throughout July and August. By the beginning of September, the number of daily cases in the United States was back to 160,000.
Looking at the same chart on IHME today demonstrates the extent to which the model has changed.
At the same point in September, IHME projects a dire future in which, without universal masking, cases remain very high for months.
When looking at either of these charts, the numbers can be confusing because IHME is estimating what they believe to be the “real” number of daily cases rather than the number of positive tests in a given day based on the rate of testing and the rate of positives within each state. But in general, what the numbers show is a rate of cases that’s actually worse in January 2022 than it is today. The “worse” variant in these charts represents what happens if all social distancing and masks rules are dropped and people begin traveling, dining out, etc. at prepandemic rates. The predicted results there are … terrible doesn’t begin to cover it.
By contrast, here’s the projection of the Ensemble model—made up of results from over 100 individual models and projections—at the CSMH.
Ensemble model, CSMH on Sep. 14.
The numbers here are also deceptive, because while most people are used to seeing the number of COVID-19 cases reported on a daily basis, the CSMH totals numbers per each week. So the fearsome number of 1.14 million cases tagged for the first week of September is actually 162,000 cases per day, matching the values seen at WorldOMeters.
The difference here is profound. At year’s end, Ensemble calls for 26,000 cases a day—which is painfully high, but 83% lower than today. On the other hand, IHME numbers for the end of the year are actually 6% higher than the current rate of infection.
Which one is right? It’s neither, of course. They’re just projections. However, back at the end of May, as IHME was making that rosy projection, MIT’s Technology Review did a rating of the various COVID-19 models and declared the Ensemble model from the CSMH to have produced the best results. In some ways, this shouldn’t be surprising because Ensemble represents something like the wisdom of a very highly educated crowd. That wisdom suggested that should the United States stop effectively increasing the number of people vaccinated, cases of COVID-19 would begin to rebound in August and September. Granted, the overall model didn’t capture either the scale or the sharpness of the increase, or peg its start in July, but it was certainly sounding a warning at a time when everything, not just the IHME numbers, tended to suggest COVID-19 was on the run.
There is no certainty (look at that vast swath of “maybe” represented by the gray space of the 95% uncertainty values in the chart above) but—assuming that kids do start to get vaccinated, and there is no “omega” variant in our near future—cases of COVID-19 should trend down for the next several months, delivering a spring 2022 that is, frustratingly, better than where we were this year, but still not good enough to declare “masks off, everyone party.”
After all, look what happened in the U.K.
Cases in the U.K. fell rapidly off an early July peak … then they stopped falling.
When cases there peaked on July 20 and began to go down almost as steeply as they had gone up, everyone seemed to make an immediate assumption that delta was going to go as quickly as it came. However, that peak in cases was also when Boris Johnson declared “freedom day” and dropped mask rules and social distancing guidelines over most of the nation. Predictably, within a few weeks of that change, cases actually began to rise again. Despite a relatively high level of vaccination, the lack of restrictions was enough to boost the delta wave back into growth mode.
So, in this case, let’s hope that the U.K. is not a model for where the U.S. goes next, because there is one point on which the IHME projection and the Ensemble model absolutely agree: If the U.S. drops masks and ends other restrictions, all bets are off.
The IHME numbers have gained a lot of attention over the last two years, in part because it was that site that was first waved in front of Trump to make it clear that COVID-19 posed a serious threat, and that hundreds of thousands might die. However, there is another site, the COVID Scenario Modeling Hub (CSMH), which runs a series of competing models from different organizations and universities. Taking a consensus model from that site has, in most instances, produced a much more reasonable prediction of where things are going next.
And let’s hope that’s the case now. Because the model being thrown up at the CSMH is drastically different at this point than the future projected by IHME. Taking into account both the possibility of childhood vaccinations being available sometime in the next two months, and anticipating that no “super variant” that is more contagious than delta and able to resist current vaccines appears, the modeling hub projects that cases in the United States are at the start of a steep decline. Their models suggest that by the end of the year the United States could be back to a point roughly where it was when the delta wave began, and by the beginning of spring 2022, cases could be well below 10,000 a day.
That’s not “back to normal,” but it’s much better than where we are today.
At the beginning of July, the average number of daily cases of COVID-19 was down to around 12,000 a day. That was the lowest number since cases first began to ramp up in March 2020, and the way the numbers had come down from a high of over 250,000 a day at the start of the year made it seem as if COVID-19 wasn’t just declining, but on its way out the door.
Because IHME is simply a projection of current trends, what it showed for the future at the end of May appeared extremely rosy.
At the end of May, IHME projected the number of new cases of COVID-19 dropping sharply through the summer.
But even as the United States was experiencing that low, officials in the U.K. were warning that the dominant variant in that country was moving from the version that had recently been renamed alpha to the now all too familiar delta. Within an astoundingly short period, cases there swelled back to a peak that nearly matched the worst of the mid-winter wave. And then the U.S. started following the same track, just a couple of weeks behind, as delta became dominant across the nation and areas like rural Missouri exploded with new cases. This new “pandemic of the unvaccinated” became a lingering event that just kept generating more and more cases throughout July and August. By the beginning of September, the number of daily cases in the United States was back to 160,000.
Looking at the same chart on IHME today demonstrates the extent to which the model has changed.
At the same point in September, IHME projects a dire future in which, without universal masking, cases remain very high for months.
When looking at either of these charts, the numbers can be confusing because IHME is estimating what they believe to be the “real” number of daily cases rather than the number of positive tests in a given day based on the rate of testing and the rate of positives within each state. But in general, what the numbers show is a rate of cases that’s actually worse in January 2022 than it is today. The “worse” variant in these charts represents what happens if all social distancing and masks rules are dropped and people begin traveling, dining out, etc. at prepandemic rates. The predicted results there are … terrible doesn’t begin to cover it.
By contrast, here’s the projection of the Ensemble model—made up of results from over 100 individual models and projections—at the CSMH.
Ensemble model, CSMH on Sep. 14.
The numbers here are also deceptive, because while most people are used to seeing the number of COVID-19 cases reported on a daily basis, the CSMH totals numbers per each week. So the fearsome number of 1.14 million cases tagged for the first week of September is actually 162,000 cases per day, matching the values seen at WorldOMeters.
The difference here is profound. At year’s end, Ensemble calls for 26,000 cases a day—which is painfully high, but 83% lower than today. On the other hand, IHME numbers for the end of the year are actually 6% higher than the current rate of infection.
Which one is right? It’s neither, of course. They’re just projections. However, back at the end of May, as IHME was making that rosy projection, MIT’s Technology Review did a rating of the various COVID-19 models and declared the Ensemble model from the CSMH to have produced the best results. In some ways, this shouldn’t be surprising because Ensemble represents something like the wisdom of a very highly educated crowd. That wisdom suggested that should the United States stop effectively increasing the number of people vaccinated, cases of COVID-19 would begin to rebound in August and September. Granted, the overall model didn’t capture either the scale or the sharpness of the increase, or peg its start in July, but it was certainly sounding a warning at a time when everything, not just the IHME numbers, tended to suggest COVID-19 was on the run.
There is no certainty (look at that vast swath of “maybe” represented by the gray space of the 95% uncertainty values in the chart above) but—assuming that kids do start to get vaccinated, and there is no “omega” variant in our near future—cases of COVID-19 should trend down for the next several months, delivering a spring 2022 that is, frustratingly, better than where we were this year, but still not good enough to declare “masks off, everyone party.”
After all, look what happened in the U.K.
Cases in the U.K. fell rapidly off an early July peak … then they stopped falling.
When cases there peaked on July 20 and began to go down almost as steeply as they had gone up, everyone seemed to make an immediate assumption that delta was going to go as quickly as it came. However, that peak in cases was also when Boris Johnson declared “freedom day” and dropped mask rules and social distancing guidelines over most of the nation. Predictably, within a few weeks of that change, cases actually began to rise again. Despite a relatively high level of vaccination, the lack of restrictions was enough to boost the delta wave back into growth mode.
So, in this case, let’s hope that the U.K. is not a model for where the U.S. goes next, because there is one point on which the IHME projection and the Ensemble model absolutely agree: If the U.S. drops masks and ends other restrictions, all bets are off.