Nearly three-fifths of Americans had been infected by the COVID-19 virus at least once as of February, according to new estimates from the Centers for Disease Control and Prevention (CDC). The results, which are based on seroprevalence research involving blood samples from all 50 states, indicate that infection prevalence varied widely across age groups: It was about 75 percent for children 11 and younger, 74 percent for 12-to-17-year-olds, 64 percent for 18-to-49-year-olds, 50 percent for 50-to-64-year-olds, and 33 percent for Americans 65 or older.
This study suggests that roughly 192 million Americans had been infected as of February, more than twice the number of cases that had been reported at the time. Based on the number of COVID-19 deaths recorded at the end of February, that estimate implies an overall U.S. infection fatality rate (IFR) of about 0.5 percent, which is substantially lower than the estimates used in early epidemiological models that projected as many as 2.2 million COVID-19 deaths in the United States, more than twice the current total. At the same time, the implied IFR is much higher than the estimates suggested by early seroprevalence studies in California and Florida.
Is 0.5 percent in the right ballpark? That depends on how well the CDC study measured the prevalence of infection. But it is also important to keep in mind that infection fatality rates can vary widely over time as the mix of patients changes, treatment improves, and vaccination becomes increasingly common; across age groups, since the risk for older people is vastly higher than the risk for younger people; and across locations with different demographics, patient characteristics, and health care capacities.
The blood samples for the CDC study were drawn for diagnostic purposes unrelated to COVID-19, and the researchers looked for anti-N antibodies, which are produced in response to infection but not in response to the vaccines approved for use in the United States. The New York Times reports that the study "used a test sensitive enough to identify previously infected people for at least one to two years after exposure."
The researchers note four limitations: "First, convenience sampling might limit generalizability. Second, lack of race and ethnicity data precluded weighting for these variables. Third, all samples were obtained for clinical testing and might overrepresent persons with greater health care access or who more frequently seek care. Finally, these findings might underestimate the cumulative number of SARS-CoV-2 infections because infections after vaccination might result in lower anti-N titers, and anti-N seroprevalence cannot account for reinfections."
Those limitations suggest that the total number of infections may be higher than the CDC's estimate. The Times reports that "some scientists said they had expected the figures to be even higher, given the contagious variants that have marched through the nation over the past two years." If the gap between reported cases and total infections is bigger than the CDC's results suggest, that would imply a lower overall IFR.
In any case, a nationwide IFR estimate for a particular period of time obscures factors that have a big impact on the danger posed by COVID-19. In light of those factors, any single IFR estimate is apt to be misleading. Instead of trying to estimate the one "true" IFR, it makes more sense to recognize that there are many IFRs, contingent on time, location, and demographic variables.
In a January 2021 Bulletin of the World Health Organization article, Stanford epidemiologist John Ioannidis reported that the IFRs implied by seroprevalence studies "tended to be much lower than estimates made earlier in the pandemic." But he also noted that "the infection fatality rate is not a fixed physical constant," and "it can vary substantially across locations, depending on the population structure, the case-mix of infected and deceased individuals and other, local factors."
Although it has long been clear that COVID-19 fatality rates are strongly correlated with age, the magnitude of the differences remains astonishing. According to the CDC's "best estimate," the IFR for people 65 or older is 9 percent, 4,500 times the IFR for children and teenagers (0.002 percent). A Lancet analysis published this month found that "age-specific IFR estimates form a J shape, with the lowest IFR [0.002 percent] occurring at age 7 years." The estimated IFR "increas[es] exponentially" with age: from about 0.06 percent for a 30-year-old to 1 percent for a 60-year-old and 20 percent for a 90-year-old.
Parents of children who are not yet eligible for vaccination may be reassured by the CDC's estimate that 75 percent of kids younger than 12 already had been infected by February. "That so many children are carrying antibodies may offer comfort to parents of those aged 5 and under," the Times says, "since many may have acquired at least some immunity through infection." But the most reassuring thing about the risk that COVID-19 poses to children in that age group is that it has always been tiny: According to the Lancet study, the IFR ranges from 0.002 percent for 5-year-olds to 0.005 percent for 1-year-olds.
That same study found that "all-age COVID-19 IFR varied by a factor of more than 30 across countries and territories during the pre-vaccine era." The countries with the highest rates as of July 15, 2020, were Portugal (2.1 percent), Monaco (1.8 percent), Japan (1.8 percent), Spain (1.7 percent), and Greece (1.6 percent). When the researchers adjusted for age demographics, Portugal and Spain were still in the top five, but the other three countries were replaced by Peru, Oman, and Mexico.
"Because IFR is strongly related to age," the authors report, "population age structure accounted for nearly three-quarters of variation in IFR estimates for in-sample countries on July 15, 2020." But even when that factor was taken into account, "many North American and European countries continued to have high IFRs despite having greater access to health-care resources." The researchers say possible explanations include "high SARS-CoV-2 transmission rates in the care home population of some locations" and "a higher prevalence of comorbidities that increase the severity of COVID-19 disease."
The IFRs implied by CDC seroprevalence research conducted around the same time likewise varied widely across states, ranging from 0.1 percent in Utah to 1.4 percent in Connecticut. As with the international comparisons, age demographics probably explain much of the variation (the median age in Utah is substantially lower than the median age in Connecticut), but other factors (such as preexisting medical conditions) may also be important.
The Lancet study, which covered 190 countries and territories, also found that IFRs fell over time. Adjusted for age demographics, they ranged from 0.17 percent to 1.16 percent on April 15, 2020, and from 0.12 percent to 0.77 percent on January 1, 2021. The median IFR fell from 0.54 percent to 0.35 percent during that period. The age-standardized IFR for the United States, according to these estimates, fell from 0.73 percent to 0.43 percent.
IFRs were dropping well before vaccines were widely available, which may reflect a combination of shifting patient characteristics, improved treatment, and naturally acquired immunity. "The evidence suggests that a range of improvements in clinical management have contributed to substantive improvements in clinical outcomes that are likely to decrease the IFR over time," the researchers say.
Vaccination, which dramatically reduces the risk of life-threatening symptoms, can be expected to push IFRs down further, although age still seems to be the most important predictor of infection outcomes. As Reason's Elizabeth Nolan Brown noted in September, "a Financial Times analysis found the COVID-19 mortality risk is about equal for vaccinated 80-year-olds and unvaccinated 50-year-olds, while an unvaccinated 30-year-old has less chance of dying than a vaccinated 45-year-old."