The Messiest Phase of the Pandemic Yet

The numbers are remarkable. More than 100 million people in the United States have likely been infected by SARS-CoV-2 and 167 million people are fully vaccinated. Yet despite this huge population of people with at least some level of immunity, the Delta variant has sent case and hospitalization numbers soaring. Florida is on its way to having twice as many people hospitalized now than during any previous wave, when essentially no one was vaccinated.

One way to think about it, as the epidemiologist Ellie Murray has laid out, is that if Delta is as transmissible as the CDC thinks, we need a much higher percentage of our population vaccinated for immunizations and natural infection alone to cause the virus to peter out. Even when the huge majority of people in a given place have gotten the coronavirus or a shot, there might still be outbreaks, as the Brown University public-health expert Ashish Jha fears will happen in South Dakota after the Sturgis Motorcycle Rally.

These realities have already smashed the more optimistic projections of late spring, including my own. Having stared at these numbers for months and months with the COVID Tracking Project at The Atlantic, I never thought that we’d see hospitalization numbers higher than they were during the winter peak in any state. But here we are.

It’s time for a data-driven reset on the basic knowns and unknowns of this pandemic, a task that must be undertaken with great humility. The virus keeps changing, and so does our understanding of the social and biological components of the pandemic. But in exploring both the knowns and the unknowns, we can see how complex the pandemic has become—and that we’re still lacking crucial data because of the failings of state and federal government.

The Knowns

Let’s begin with the best bit of good news. Based on the available data, all the vaccines given in the United States appear to confer a solid level of immunity against severe outcomes such as hospitalization and death. Over a three-month period this summer, the CDC recorded 35,937 deaths from COVID-19—but just 1,191 of those who died were fully vaccinated. In other words, 96.7 percent of deaths this summer have been in the unvaccinated. Hospitalization data look similar, with few fully vaccinated people requiring hospitalization.

The CDC’s data mirror what other institutions have found. The New York Times was able to compile data from most states on the percentage of people with breakthrough cases who were hospitalized. Although the proportion of breakthrough patients varied by an order of magnitude from as low as 0.2 percent of total hospitalizations in Texas to 4.7 percent in Arkansas, in every state more than 95 percent of hospitalized people were unvaccinated. This is also consonant with data from the United Kingdom, which, because of its National Health Service, has better data than exist in the U.S. Many other, smaller studies in various states show very similar results.

In Israel, data showed that fully vaccinated patients who were hospitalized were much more likely to have comorbidities such as hypertension, diabetes, and heart failure.

So this is known: At this moment in the pandemic, fewer than 5 percent of the people being hospitalized and dying are fully vaccinated.

Although the randomized controlled trials on vaccine efficacy indicated that the vaccines conferred substantial protection from symptomatic infection—with efficacies touted at about 95 percent for the mRNA vaccines—their real-world performance is almost certainly lower, though to what extent is not exactly clear. (More on that in a minute.)

At the same time, more and more evidence suggests that some people with breakthrough infections can transmit the virus. Combine those two facts with Delta’s extremely high transmissibility, and we’ve found ourselves in a world where even well-vaccinated communities can see quick growth in cases. Back in the pre-variant days of the pandemic, 70 percent vaccination was seen as a rough goal to achieving herd immunity, the point at which viral growth could no longer be sustained in a community. Yet San Francisco, which has 70 percent of its population vaccinated, has nonetheless seen a similar case surge to the one in Maricopa County, home to Phoenix, Arizona, where only 43 percent of residents are vaccinated.

Although, statistically, counties and states with higher vaccination rates have lower case counts and hospitalization rates, they have still become areas with high levels of community spread.

Read next: Why is it taking so long to get vaccines for kids?

There are probably different transmission dynamics within these cities. Young, unvaccinated people are likely responsible for a good deal of transmission. There are, after all, still 50 million kids under 12 who are not eligible for the vaccines. But it’s also likely that older, vaccinated people are responsible for some spread as the amount of virus increases in the community.

In a number of places, this has not caused major increases in hospitalizations, but that’s not universally true. Perhaps the most startling example is The Villages, in Florida. Centered on a retirement community, this metropolitan area has close to 90 percent of its over-65 population immunized, yet it has seen a surge of cases and hospitalizations.

Although, again, statistically, places where more people are vaccinated are faring better than places where fewer people are vaccinated, there is enormous variability lurking in the numbers. Some of it may be explainable by policy decisions and political allegiances. But some of it is also just luck.

Back in the spring, when the variant we were most worried about was called Alpha, Michigan and almost Michigan alone got absolutely torched, matching its peak for hospitalizations from the winter. This didn’t happen anywhere else, though some epidemiologists expected it to, based on the experience of European countries. Alpha just kind of went away, and it seemed like the U.S. might be in the clear.

Enter Delta. In this surge, a piece of Missouri began to take off before the rest of the country. Would it be like Michigan? As we all now know, the answer was no. The southeastern United States is now experiencing huge outbreaks as many states come close to matching or surpassing their pandemic peaks in cases and hospitalizations.

The health-care system in north Florida is under pressure that few places have seen at any time during the entire pandemic. Why there? Why not somewhere else with similar vaccination rates and political opposition to viral countermeasures? No one knows with total certainty, and we’re unlikely to ever find out.

One of the rare blessings of the pandemic has been that children have proved uniquely able to deal with the attacks of SARS-CoV-2. Their risk of serious illness has always been very, very low. And the available evidence suggests that this has not changed. COVID-NET, which is a CDC sample of hospitals, shows that the rate of hospitalizations for kids has varied in the pandemic from about 0.3 hospitalizations per 100,000 to 1.5 hospitalizations per 100,000. That rate is rising quickly now, but it remains within the historical envelope of the pandemic in the United States—at about one hospitalization per 100,000 children. Similar trends hold true for age subgroups such as 0–4, 5–11, and all under 18. And hospitalization rates for children younger than 18 remain considerably lower than the peak hospitalization rates of adults 18–49 (10 in 100,000), 50–64 (28 in 100,000), and 65 and older (72 in 100,000).

That said, the CDC’s COVID-NET data do not cover the whole country—they pull data from 99 counties across 14 states, representing about 10 percent of the U.S. population. And they have no data from hospitals in the worst-hit state, Florida.

If we look at the pediatric hospital admissions gathered by the Department of Health and Human Services, we can see that they are at record highs now, surpassing the worst day of the winter surge and still headed straight up. Although the overall numbers are still low—the seven-day average of hospital admissions is fewer than 300 per day—children in the Southeast are being hospitalized at almost double the rate as at any other time in the pandemic. Some states are seeing two or even three times as many admissions of kids as at any time in the pandemic.

It’s important to note that there could be multiple reasons why we’re seeing this increase in pediatric hospital admissions. Taken alone, the increases do not mean that the average Delta-variant infection is more severe in children than previous SARS-CoV-2 strains. In an early Scottish study, the data were worrisome, but the bigger problem is that almost all the data available for school decision makers are from before the Delta wave. The nation’s children will be back in the classroom before we’ll definitively know whether Delta hits kids harder than previous variants.

Still, the kids’ numbers are tracking quite closely with their adult numbers—and it’s not totally clear whether this is a meaningful shift compared with earlier stages of the pandemic.

And what all the data taken together seem to suggest is that there is an absolutely monster wave of coronavirus circulating in the South, and that our current case numbers do not come close to capturing the true number of infections in that region of the country.

Breakthrough infections for vaccinated people were always going to happen. No vaccine provides perfect immunity, and the immune system is strange and somewhat unpredictable.

But there was some logic to the hope that maybe these infections wouldn’t transmit the virus forward. Because the large majority of vaccinated people have mild symptoms, the thinking went, perhaps they would have lower viral loads, and therefore be less likely to spread the virus.

How well the vaccines protect against any infection (not just symptomatic infection, hospitalizations, or death) is a hotly disputed topic. A variety of data suggest that vaccination does help prevent exposures to the virus from becoming infections, and that, obviously, helps slow the spread of an outbreak.

But it’s also become clear that vaccinated people who do get infected can spread the virus. The most recent piece of evidence came when American scientists were able to culture virus from samples taken from vaccinated people who’d gotten infected. Those same people showed similar viral loads to unvaccinated people. And yes, even those with asymptomatic infections.

Although that’s bad news, there is some good news too: Breakthrough infections appear to be significantly shorter than infections in the unvaccinated. That would reduce the amount of time that people with breakthrough infections could spread the virus.

There will undoubtedly be many more studies along these lines, and the papers cited above are preprints, meaning that they have not yet been peer-reviewed. But the data, including unpublished studies cited by public-health officials, are pointing in the same direction: Breakthrough infections are happening. And when they do, those people can spread the virus.

The Unknowns

We already know that we’ve been undercounting the true number of infections over the course of the pandemic. Sure, we have a tally of cases, but that count is almost entirely of cases confirmed by a positive test result. And as the conditions of the pandemic have changed, the relationship between that case count and the actual number of infections has varied. It’s such a basic question that it seems absurd to ask, and yet we simply don’t know how many Americans have had COVID-19.

This is not a purely academic question. Natural infection should confer some level of immunity, though whether natural immunity is as protective as the vaccines is unclear. Regardless, it’s important to know how many naive immune systems are out there for the virus to get at. We know the number of fully vaccinated people with reasonable precision—call it a bit shy of 170 million people. But how best to estimate how many people have been infected? The CDC has done some testing of the levels of antibodies in the U.S. population, but the data are incomplete and imprecise; you could not simply multiply the percentages of people with antibodies by the number of people in the country and get an accurate number.

Through hospitalization and death data, we know the rough shape of the infection waves. There was a large New York–centric wave in spring 2020; a smaller, southern-focused summer wave; then last winter’s massive nationwide wave. In 2021, there was a small spring wave centered on the upper Midwest, and now there’s the current massive summer wave in the Southeast.

One can add up all the cases from those waves and find roughly 36 million confirmed cases. But for each era of the pandemic, there has been varying testing availability and usage—not to mention a large pool of asymptomatic infections. That’s led to very different and still-unknown case-detection rates through time. Public-health officials know they severely undercounted in the beginning of the pandemic, a problem that improved through the year. But in 2021, the availability of at-home tests and COVID-19 denialism—particularly in the less-vaccinated right-wing areas where the virus is flourishing—among other factors, may have driven down the number of cases that we’re confirming. Vaccinated people with mild cases may also have less incentive to get tested, because they know they are unlikely to have major complications. The CDC itself initially did not recommend that fully vaccinated people get tested after exposure, before changing that guidance in late July. Some institutions also dropped testing regimes for vaccinated people, and some testing sites scaled back their services. All of this is to say: Both interest and access may be lower than earlier in the pandemic.

Read: The vaccine scientist spreading vaccine misinformation

The CDC does make an estimate of the total number of infections. That number was 120 million with a range from 103 million to 140 million before the Delta wave. How many people have been infected since June 1? The CDC has counted about 3 million cases, but who knows what the relationship of that number is to the true number of infections.

Then there is one final unknown regarding immunity: What is the overlap between the people who have been infected and the vaccinated? The U.S. does not have these data, but they’re a pretty important component of our current situation. If there were no overlap between the 170 million vaccinated, and there have been 150 million infections, we’d be looking at 320 million people with some immunity, nearly the whole country.

But it’s likely that there is a good deal of overlap. And the more overlap, the more dry tinder there is to keep this pandemic going. When everyone in the United States has been vaccinated or infected, it won’t mean that the pandemic is over, but our collective immune systems will have become a more formidable opponent for the many strains of SARS-CoV-2.

2. How well do the vaccines work to prevent infection?

As noted, all available data show that the vaccines remain remarkably effective at reducing the risk of hospitalization and death from COVID-19. But past that very important outcome, the data are much murkier.

So the effectiveness of the vaccines is a matter of perspective. What people might refer to as vaccine effectiveness can have different meanings, and therefore the nature of their data and calculations can vary. If we want to talk about vaccine effectiveness precisely, we need to specify effectiveness against an outcome (infection, symptomatic disease, hospitalization, death). We also need to define the temporal parameters: across how long of a time period? When were the vaccines administered? We need to break out the different vaccines. We need to have a rough understanding of the variants in circulation when a given study was done. And finally, we need to specify which population is under discussion—young, old, immunocompromised, health-care workers, etc.

Sure, all these factors can be rolled up, and had to be rolled up during the vaccine approval process, into a single number to determine vaccine efficacy. That number came out to 95 percent in the original trials for the mRNA vaccines.

Effectiveness is what comes from empirical observations. As these results have been released, what we’ve usually heard is something like this from the CDC vaccine-effectiveness page: “mRNA COVID-19 vaccines offer similar protection in real-world conditions as they have in clinical trial settings, reducing the risk of COVID-19, including severe illness, among people who are fully vaccinated by 90 percent or more.”

But here’s the thing. Change one of the crucial variables, and the picture changes. That’s led to the publication of multiple conflicting studies. A New England Journal of Medicine study found Pfizer’s effectiveness against symptomatic disease from the Delta variant to be 88 percent. That’s great!

But a preprint paper working with Mayo Clinic data found much lower effectiveness against infection, especially for the Pfizer vaccine, which the authors contend had an effectiveness of just 42 percent against infection after Delta became prevalent in the populations that they studied. These findings are both surprising and disconcerting.

Read next: Masks are back, maybe for the long term

There is wide variability in international studies as well. In slides prepared by the CDC for the expert panel that provides recommendations on vaccines, we can see the same kind of difficult-to-explain results. Pfizer looks great in the English/Scottish and Canadian data, even against infection and symptomatic disease, but Israeli and Qatari data do not show the same performance.

It may be that these data can be reconciled in some way. For example, the NEJM study looked at symptomatic disease, while the Mayo Clinic paper may have picked up more asymptomatic disease. But even that would not be too reassuring at a population level because, as noted above, it now seems likely that vaccinated people with asymptomatic infections can spread the virus, at least sometimes. And the Israelis, at least, didn’t seem to show a major difference between vaccine effectiveness in preventing infection and symptomatic infection.

There are many other possible explanations. Could the effectiveness of the vaccines fade more quickly than hoped, so those with less recent vaccinations are more likely to get infected? Could there be a problem with the distribution of some of the Pfizer doses, which require the most intense cold storage of any of the immunizations? Maybe the way that the studies picked their subject populations or did the data work pushed the results one way or another. And none of this touches on the effectiveness of the Johnson & Johnson vaccine, which had a lower efficacy in trials.

For the time being, it seems prudent to assume that it’s possible that one or more of the vaccines will be found to have substantially lower real-world performance in preventing Delta infection and/or symptomatic disease.

3. Why have so many more people been hospitalized in the United States than in the United Kingdom?

After a glorious June, when cases in the U.S. fell to their lowest levels since the beginning of the pandemic, more virus began to circulate around the country. The United Kingdom had just seen a surge, but it did not result in an accompanying surge of hospitalizations or deaths. That seemed to portend good things for the United States.

On June 1, when the Delta wave began to take off in the U.K., approximately 40 percent of its population had been fully vaccinated. The wave ran high—reaching 80 percent of the case peak from the winter—but hospitalizations reached only 15 percent of the winter peak before the wave began to recede. This was fantastic news from a British perspective.

Fast-forward a month and cross the Atlantic Ocean. When the Delta wave began to take off in early July in the U.S., roughly 47 percent of the U.S. population was fully vaccinated. But in the American context, hospitalizations have not only risen to 50 percent of their pre-pandemic peak but continue to rise. Several southern states are seeing their all-time peaks in hospitalizations, despite three previous waves of infection and millions of vaccinated residents.

Florida had a larger share of its population vaccinated at the start of the American Delta wave than the U.K. did when it saw the variant’s exponential rise. Yet, in Florida, the state now has nearly double the number of COVID-19 patients in hospitals than it has ever had during the pandemic.

It will take a long time to tease out the different factors between the U.S. and the U.K. Obviously, for example, the United States is a much larger country with distinct types of urban structures.

But there are several other immediate pathways for thinking about why things are playing out so unlike in the U.S. The U.K.’s vaccination strategy was substantially different from the American one, despite the overall similarity of vaccination rates. It could also be that American unvaccinated people were spread more unevenly through the country than the unvaccinated in the British context, with different epidemiological effects.

Looking at Florida, though, one thing stands out. For reasons few epidemiologists could understand, the state had not been hit as hard as neighboring places with similar populations and politics. Look at almost any metric before the Delta wave, and Florida fared pretty well relative to New York, California, or Illinois. Not until the current Delta wave has Florida experienced a surge comparable to those seen in other big states.

The U.K., by contrast, was hit with two massive COVID-19 waves in which the death rate was nearly twice what it was in the U.S. That suggests that a much greater percentage of the U.K. contracted the virus, giving them some natural immunity. The virus may have run out of bodies to attack.

Perhaps, in Florida, the state’s good fortune in previous waves—along with the political opposition to societal countermeasures—could be one of the factors driving this gigantic increase in COVID-19.

4. What percentage of infections are we confirming as “cases”?

Positivity rates—as my colleagues at the COVID Tracking Project argued over and over—are a fraught metric, especially when used as a threshold for crucial decisions. However, as a coarse measure of whether testing is adequate, they do help tell the story of the case numbers that are coming out of the American South.

Note that the goal for positivity rate in most states was under 3 percent. In the Florida panhandle and adjacent counties in Alabama and Georgia, the positivity rate in many counties is greater than 25 percent. That’s comparable to many jurisdictions in the days of highly constrained testing supplies during the first wave of the pandemic. Not coincidentally, these areas are also seeing massive increases in hospitalizations, and because severe illness is largely occurring in unvaccinated people, we will also see a rise in deaths.

Read next: How the pandemic ends is different now

On its face, elevated positivity rates have historically meant that public-health surveillance was missing a greater share of the infections in a community. But there’s more evidence that this is what’s happening. Relative to previous waves, the ratio of cases to hospitalizations is lower. Last winter, we confirmed 12 million cases in December and January. This wave, we’ve confirmed fewer than 3 million cases since July 1. Last winter, we peaked at more than 120,000 COVID-19 patients in the hospital at one time. Right now, we’re already over 64,000. So we are showing 25 percent of the cases and 50 percent of the hospitalizations.

In addition, at-home tests such as the Abbott BinaxNOW and other types of testing in institutional settings like schools may not be getting reported to authorities. It’s really all a mess.

There are more precise ways to look at these data in particular hospital systems and areas, but the upshot is that either Delta is making people sicker—which, as noted above, is a real possibility—or our case-detection rate has fallen. Or, just to muddy things, maybe both.

5. How many people will die?

For people in countries with access to vaccines, the good news is that it seems almost certain that fewer people will die in this wave of COVID-19 than in the winter surge. Fewer both in absolute terms and in the percentage of COVID-19 infections as the vaccines make many more people less vulnerable to severe illness. This is an unmitigated good (and one that is a moral imperative to extend to the rest of the world).

But millions of unvaccinated people are still getting infected. And for them, the old mathematics of COVID-19 will hold. Older people who get sick are more likely to die. The more comorbidities an infected person has, the more likely they are to die.

Here, again, on a national level, there is good news. The most vulnerable group—people older than 75—are being infected at about 10 percent of the rate at the winter peak. That’s a major decline.

But in specific areas, which is to say, yet again, Florida, that trend is not holding. There, the rate of hospital admissions is up for every single age bracket, from young people to those older than 80. And people who are between 50 and 79 years old are being admitted to the hospital at higher rates than ever in the pandemic. Some of those people are going to die, and the numbers will not be small.

There is also hope that better therapeutics and improved care practices will push the death rate down. But there is pressure in the other direction too. With the hospitals in hard-hit areas under tremendous burden, they are less likely to be able to provide the highest standard of care.

Florida is already reporting a seven-day average of more than 150 deaths a day, a number that seems likely to rise as the statistics work their way through the system. Florida’s peak over the winter was about 180 deaths a day.

The big question in all this, then, is: Does Florida portend what’s likely to happen in the rest of the country? That’s not yet clear, and let’s hope not.

6. What are the risks of long COVID?

This section is more a list of questions than it is of answers. There’s so much that we don’t know about the risks of long COVID. For example, how susceptible are adults with mild infections to long COVID? How about kids? How about breakthrough infections? And asymptomatic infections?

Post-viral syndromes have long been understudied. Long COVID is a bundle of the unexplained aftereffects of a virus that we’ve never encountered before. That’s an especially hard thing to study. Yet, pushed by patient advocates, scientists are trying to get a handle on the depth of the problem.

According to one meta-analysis of research, at least some symptoms persist longer than two weeks for 80 percent of COVID-19 patients. An English survey found that more than 10 percent of people who had COVID-19 said the disease’s effects were still having a “significant effect on my daily life” 12 weeks after infection. Another found only 2 percent of people experiencing symptoms 12 weeks after infection. And another found 38 percent of post-COVID people with at least one symptom 12 weeks out. Many research studies and anecdotal stories speak to the prevalence of these problems. But the specifics are really hard to pin down, as are the risk factors.

And what of post-vaccination infection? Will the immunizations prove effective at reducing long COVID too? One small study of health-care workers found extended symptoms in some people with breakthrough infections. If the Delta variant really begins to move across the whole country, there will be a lot of breakthrough infections.

If you are relatively young and healthy, you could always bet that you’d probably come out of a COVID-19 infection just fine, neither hospitalized nor dead. That’s doubly true for the vaccinated. But long COVID is the big, spinning question mark in all the risk calculations that human beings must continue to make.

Dave Luo and Lindsey Schultz contributed reporting to this story.