Check the COVID-19 Data Tracker from the U.S. Centers for Disease Control and Prevention (CDC), and you’ll get a rundown of the latest case numbers, hospitalizations, and deaths. Those categories might seem straightforward, but the data, say many experts, are telling us a lot less than we think they are.
That’s because it’s getting increasingly difficult to parse who is hospitalized or dies from COVID-19, and who is hospitalized or dies from another reason but with COVID-19. Across the U.S., “COVID-19 hospitalizations” represent all kinds of patients: those who need hospital-level care for severe cases of COVID-19; those with risk factors like heart disease or kidney issues who got infected, then had a heart attack, stroke, or kidney failure and needed to be hospitalized; and those who were admitted for one health condition but tested positive for COVID-19 at some point during their stay or several weeks afterward. COVID-19 plays a role of varying importance in all of these hospitalizations. “The situation is murky because we don’t know if COVID-19 is to blame for their worsening chronic health, or whether they developed a COVID-19 opportunistic infection that is [having] more of a bystander effect,” says Dr. Susan Cheng, professor of cardiology and director of public health research at Cedars-Sinai. “It’s hard to parse these things out except in the most extremely obvious cases.”
Among public health experts, there’s a simmering debate over what U.S. COVID-19 numbers really reflect. In a widely discussed and controversial column, George Washington University professor Dr. Leana Wen recently argued in the Washington Post that deaths reported due to COVID-19 are likely overcounted, as some of them might have been more attributable to other causes but were listed as COVID-19 deaths because the individual also tested positive. In Los Angeles County, academic and public health researchers reported last year that in the county’s public hospital, 67% of people testing positive for COVID-19 were not hospitalized because of their infections. Others disagree: since COVID-19 often exacerbates health events and conditions, the numbers, they say, may be undercounting the impact of COVID-19 on deaths.
“I don’t think we are overcounting COVID-19 deaths,” says Dr. Carlos del Rio, professor of medicine at Emory University and president of the Infectious Diseases Society of America. He notes that most of the deaths are occurring among older people who are more vulnerable to the worst effects of COVID-19. “I think the data suggest that we are still seeing a fair number of deaths [from COVID-19], and they are occurring in people with high risk for complications,” he says.
Even in the third year of the pandemic, getting the numbers right matters. Being able to accurately identify who is still getting gravely sick from COVID-19 could help public-health officials better target those who would benefit most from booster doses and antiviral treatments. As the country’s health officials move toward simplifying COVID-19 immunizations, knowing who is experiencing severe COVID-19 could also tailor immunization recommendations, such as increasing the number of doses, for the most vulnerable so they can avoid the more serious symptoms of disease. Such detailed hospitalization and death data would also help health officials to learn a lot more about how COVID-19 is interacting with other common health issues.
Why the numbers are such a mess
The CDC’s data come from hospitals or state health departments, which are required to report daily admissions of patients who have COVID-19 and deaths of patients with COVID-19. In some states, hospitals report COVID-19 hospitalizations directly to the CDC, while in others, state health departments collect the data and provide it to the federal government. (The CDC did not respond to requests for comment on how it presents COVID-19 hospitalization and death data.)
But what hospitals consider a COVID-19 admission often differs. “Right now, the health care system is still struggling to keep up,” says Cheng. “We’re doing the best we can with the knowledge we have to code [cases and deaths] as appropriately as possible. But we’re not even close to the ideal state of being able to talk about what that means in practice about [getting consistency in] how we are coding these things.”
Some groups recognize this problem and have standardized how they classify COVID-19 hospitalizations and deaths. For example, in King County, Wash., which includes Seattle, the health department reviews every COVID-19 hospitalization record to “understand whether people are coming in primarily because of a COVID-19-related condition or if COVID-19 is incidental to something else,” says Dr. Jeff Duchin, health officer for public health in Seattle and King County. By their standards, COVID-19 hospitalizations include people who are admitted and have positive COVID-19 tests either within 14 days prior to their hospitalization, or up to 21 days following their discharge, and those who have a COVID-19-likely illness with no other reasonable medical explanation for their admission. Duchin says there’s an 80% concordance between the reviewers’ determination of whether COVID-19 contributed to the hospitalizations and what the medical records suggest. “We are trying to reflect the true burden of disease from COVID-19 on the health care system as best we can,” he says.
But even though hospitals and health departments in one Washington county are all on the same page, comparing hospitalizations in Seattle to those in another city using the CDC’s COVID-19 Data Tracker won’t necessarily mean you’re comparing the same thing.
Hospitals also use different criteria for determining when a patient who tests positive for COVID-19 is no longer a COVID-19 patient. Some states consider people who test positive at any time during their hospital stay a COVID-19 case, even if they test negative eventually, while others, including New York, no longer log patients as COVID-19 cases if they test negative. Others stop counting people as COVID-19 patients once their symptoms go away, or after two weeks pass following a positive test if regular testing isn’t performed.
The same discrepancies muddle the information on deaths. Hospitals rely on death certificates, which doctors fill out when patients pass away, to determine causes of death. But doctors don’t have a national set of criteria for determining whether COVID-19 caused a particular patient’s death. At Emory, Del Rio says doctors there use the intensity of treatment for a patient’s COVID-19 infection as a guide for determining what role the virus played in the person’s deteriorating health and ultimate death. “If a patient who is positive for COVID-19 is treated with steroids and then passes away, we say COVID-19 contributed to their death,” he says. “If a person with COVID-19 is not treated with a steroid, we do not say COVID-19 contributed to their death.”
Even the way states report COVID-19 deaths to the CDC is subject to interpretation. Doctors have the option of listing primary and secondary causes of death; in Florida and New York, for example, if a doctor records COVID-19 as either the primary or secondary cause of death, the state reports that as a COVID-19 death.
The need for better data
The Council of State and Territorial Epidemiologists is currently devising a new definition for what should be coded as a COVID-19 death, versus what should be considered a death with COVID-19, which could help doctors in hospitals to make more consistent determinations of COVID-19 mortality. That would potentially help to nationally standardize how deaths from the coronavirus should be recorded.
But even if every state health department and hospital counted COVID-19 deaths and hospitalizations the same way, the data would still be woefully incomplete. Very little testing for the virus is now being done—even at hospitals, since studies show that routine testing, including of people without any symptoms, does not necessarily reduce viral spread among health care workers and patients. Based on the growing evidence, at the end of 2022, the Society for Healthcare Epidemiology of America, a professional organization of public health and infection control providers recommended against routine screening of newly admitted hospital patients, leaning instead toward testing only people who had COVID-19 symptoms. Many states, including Maryland and Florida, follow these guidelines.
That policy means that cases are going unrecorded. If all patients were tested, “then we could definitely know, for example, if we saw X% increase in admissions due to heart issues…and a comparable increase in positive COVID-19 cases,” says Beth Blauer, data lead for the Johns Hopkins Coronavirus Resource Center. For many patients dying of things like heart disease, “their condition may have likely been accelerated by COVID-19, but we don’t know because they aren’t being tested.”
The issue highlights a deeper problem—one that preceded the pandemic—about how health information in the U.S. is collected. U.S. health data have notoriously lacked detailed demographic information on race, ethnicity, age, and other health conditions for patients who are hospitalized and die in the health care system. It’s a failing that CDC director Dr. Rochelle Walensky acknowledged in numerous press briefings early in the pandemic, when it wasn’t clear how COVID-19 was affecting the health of different racial and ethnic groups. “The data is terrible, and it deeply lags,” says Blauer. “There is no real-time understanding; we’re always looking at data that is one or two years back.”
Why real time COVID-19 data are important even now
As population-wide immunity to SARS-CoV-2 increases through infections and vaccinations, it’s becoming more important to know who benefits most from booster doses—which may require rethinking the current boosters-for-all approach. Now, says Dr. Paul Offit, director of the vaccine education center and professor of pediatrics at the Children’s Hospital of Philadelphia, it’s time to get smarter about targeting boosters to those who need them the most. To do that, public-health officials need to know who is getting severely ill from COVID-19 infections and getting hospitalized, and who is dying from the disease. That would help doctors to focus on making sure those groups of people are vaccinated, boosted, and given access to antiviral medications that can mitigate symptoms.
Depending on what better data find, it could also mean pulling back on boosters for those who aren’t receiving dramatically increased protection because their immune systems are relatively healthy, Offit says. “By chasing every variant and boosting everyone, we are on some level acting like the boy who cried wolf, and risking that when there is a wolf”—a strain of COVID-19 resistant to our current immune protection—”people won’t listen [and get boosted when they really need to],” says Offit.
Up-to-date data would also help us better nail down exactly who is at highest risk from COVID-19, and how to treat them. Cheng’s group has revealed an intriguing connection between hypertension and COVID-19 infections; after the first Omicron wave in late 2021 and early 2022, she and her team analyzed patients who were hospitalized for COVID-19, and found that after controlling for other factors, hypertension was enough to land some people in the hospital with more serious COVID-19 complications. It’s known that SARS-CoV-2 infects cells by using a receptor that is also involved in regulating blood pressure, ACE2, and that could explain why people with genetic differences that put them at increased risk of high blood pressure might also be at higher risk of more severe cases of COVID-19. Other studies have explore what role blood pressure medications can have in changing how infectious SARS-CoV-2 might be. But understanding these interactions will only be possible if more robust data on people who require hospital care are collected. “We are still at the tip of the iceberg,” says Cheng about the understanding of how COVID-19 is affecting other health conditions.
“We have never seen anything like this virus before, so we’d love to understand how this virus is different from all of the other viruses we have seen in our lifetime,” says Cheng. “That way we can be better prepared to counsel, treat, and manage patients as we move forward living with COVID-19.”
Correction January 31
The original version of this story misstated the type of data reviewed by Seattle King County health officials. The county reviewed hospitalization, not death data.
More Must-Reads From TIME