It’s the nightmare scenario.
After a spring that saw America’s coronavirus toll spiking to more than 36,000 cases and 2,700 deaths per day, followed by a summer in which daily cases climbed as high as 75,000 and daily deaths again cleared 1,400, it appears for the moment that the overall trajectory of the U.S. pandemic is headed in the right direction. On average, new daily cases are back down to 36,000. Daily deaths have fallen to about 700.
Yet experts have long feared that colder weather, increased indoor activity, school reopenings and holiday gatherings could combine to create a fall wave with the potential to dwarf previous peaks — and now America’s most prominent COVID-19 modelers are projecting just that.
So is it time to freak out about the fall?
Maybe not just yet.
On Thursday, the team at the University of Washington’s Institute for Health Metrics and Evaluation (IHME) released its latest pandemic forecast. It was not encouraging. In the likeliest scenario, the IHME researchers now predict that an additional 220,000 people will die of COVID-19 by the end of 2020, bringing America’s total death toll to 410,000 and more than doubling the current tally of 190,000, which took the U.S. nine tragic months to reach, in less than half that time.
Estimated daily infections — which include all reported cases plus those undetected by testing, and which the IHME currently pegs around 140,000 — would nearly triple to 373,000 by mid-November. In early December, daily deaths would surpass April highs.
And things could get even worse than that, the institute warns. If, for instance, America continues “the gradual easing of social distancing mandates” and refuses to “re-impose” them even if the pandemic worsens — essentially pursuing the sort of natural “herd immunity” strategy that President Trump’s new coronavirus adviser favors — then the IHME predicts that total U.S. COVID-19 deaths would soar to 620,000 by the end of the year, with daily deaths surpassing 12,000 and daily infections surging to 1.8 million.
The only way to keep the death count below 300,000 this year, according to the IHME, is for everyone — that is, 95 percent of Americans — to immediately start wearing a mask whenever they leave the house. Given that just 45 percent of Americans currently mask up in public, that’s probably not going to happen.
So yes, the IHME’s forecast is frightening. The question is whether it’s credible.
The truth is, nobody — not even the top coronavirus experts — has any clue how the pandemic is going to evolve over the next four months. That’s frightening in and of itself, of course. But it’s also why projecting the exact number of deaths and infections four months in advance is almost certain to be misleading.
Just consider the number of variables at play, particularly as summer gives way to fall.
Scientists are still uncertain what role children play in transmission, and how that role changes with age. So anticipating the effect of school reopenings is basically impossible.
We know the virus tends to spread more readily through the air than was previously believed, but what will that mean as temperatures drop and people retreat indoors? Will they be less likely to gather as a result? Or will they be more likely to gather instead — and in riskier, unventilated settings?
The same goes for the effect of temperature and humidity itself; studies suggest that weather “probably influences COVID-19 transmission,” much like other coronaviruses. But researchers don’t know how much.
And what about distancing fatigue? What happens if the irresistible allure of a hearty Thanksgiving dinner or cozy Christmas morning persuades more Americans to give up, once and for all, on staying 6 feet away from friends and family?
All that and we haven’t even mentioned the potential impact of a vaccine, which is expected to arrive, at least in limited quantities, by the end of 2020 — or the ripple effects of the current college outbreaks, which may seed hot spots elsewhere as students return home for the holidays.
The whole point of a forecast like the IHME’s is to account for as many of these variables as possible in mathematical form. That’s what a model is. And to be fair, the IHME’s main projection makes some reasonable assumptions: that testing, which is down slightly, will go up again once cases start to climb; that mask use will remain constant; that distancing will increase as the pandemic worsens.
But the IHME also makes another assumption that might not be so reasonable. According to an analysis by Joshua Salomon, an infectious disease and modeling expert at Stanford University, the variable “driving projections of 220K more deaths by New Years” is “not behavior” and “not mask use” — “nor is it testing.”
Instead, according to Salomon, it’s “seasonality” — the IHME’s assumption that COVID deaths will rise in the fall and winter the same way pneumonia deaths tend to rise in the fall and winter.
“Seasonality is captured using weekly, state-specific vital statistics data on pneumonia mortality from 2013 to 2019,” Salomon explains. “As far as I can surmise, the estimated rise from 900 to 2900 daily deaths derives entirely from this seasonal effect.”
It’s certainly possible that whatever seasonal factors trigger our annual spikes in pneumonia mortality have the same impact on COVID-19, and that, as a result, we’re doomed to double our national death toll by charging into the final months of the year at the dangerously high rate of 36,000 new cases per day.
But again, nobody has enough information yet to mathematically estimate how pneumonia seasonality and COVID-19 seasonality are linked — let alone to say whether they’re linked at all. Layering that assumption on other assumptions to forecast a very precise number of deaths (410,451) by a very precise date (Jan. 1) suggests a lot of certainty where there’s actually very little.
To get a sense of how wrong the IHME forecast could be, it’s worth comparing it with a different model. At covid19-projections.com, Youyang Gu, an independent data scientist trained at the Massachusetts Institute of Technology, has been using a data-driven approach with a layer of artificial intelligence (read: fewer subjective assumptions) to forecast the trajectory of the pandemic.
According to Gu’s model, daily deaths will fall to 400 by Nov. 1 — less than a quarter of the IHME’s Nov. 1 projection of 1,825. By then, according to Gu, total U.S. COVID-19 deaths will hover right around 219,000 — about 35,000 less than the IHME forecast.
Beyond Nov. 1, Gu’s model stops making projections. Most other models don’t look more than a week or two into the future. Gu used to project out three and a half months, but he has narrowed his time frame to two months because “dynamics can change in the fall and it’s uncertain how [they] will change.”
“Back in June,” he explained on Twitter, “the main driver [was] the reopening, and that's something that we can reasonably model.” But “no one knows what’s going to happen beyond November, so … any model with forecasts for beyond November needs to be taken with a grain of salt.”
“Forecasting certainty when there is very little,” Gu wrote on Twitter, “can undermine public trust in the scientific community.”
It can also terrify people. To be sure, a little fear can go a long way during a deadly global pandemic. The more Americans mask up, keep their distance, avoid indoor gatherings, isolate students at the first sign of an outbreak and test and quarantine holiday travelers, the fewer of them will likely die this fall from COVID-19.
But it’s too early to quantify the forecasted impact of those measures, or of the weather, or how those measures and the weather will interact.
For now it should be enough to say, as Gu did recently on Twitter, that knowing what we already know, tens of thousands of additional Americans, and perhaps even more than 100,000, are likely to lose their lives this year because of COVID-19. Here’s hoping the unknown isn’t deadlier than that — and that fear of the unknown keeps America from letting it get even deadlier.
Read more from Yahoo News: