A new estimate of COVID-19 prevalence in Texas, the second-most populous state, suggests that the true number of infections is more than four times as high as the official tally. While the Houston Chronicle presents that as bad news, it implies a statewide infection fatality rate (IFR) substantially lower than the most recent national estimate from the Centers for Disease Control and Prevention (CDC). The implied Texas rate is also much lower than the estimates used in last spring’s worst-case scenarios, which projected as many as 2.2 million COVID-19 deaths in the United States.
According to an analysis by the University of Texas at Austin COVID-19 Modeling Consortium, based on cellphone mobility data and hospitalization numbers, one in six Texans has been infected by the virus that causes the disease. That amounts to 4.75 million people, compared to a confirmed case tally of 1.1 million. When you take underreporting into account, a spokesman for the Texas Department of State Health Services told the Chronicle, the model’s estimate is likely to be “generally in the ballpark” of the true number.
“If you thought things were bad when Texas topped 1 million COVID-19 cases, guess what?” the Chronicle says. “Researchers estimate at least four times as many people have caught the virus.” The modeling consortium’s associate director, Spencer Fox, likewise says the infection estimate shows that “the speed at which things can get out of hand is a lot quicker than people expected.”
That is one way of looking at it: Other things being equal, the chance of encountering a carrier rises as the number of active infections goes up. But if more than three-quarters of infections have gone undetected in Texas, that suggests they did not cause symptoms serious enough for people to seek testing, which is more reassuring than alarming. It also means the IFR (deaths as a share of all infections) is much lower than the case fatality rate (deaths as a share of confirmed infections), which is currently 1.8 percent in Texas. Based on the current statewide tally of COVID-19 deaths (about 20,300), the IFR would be roughly 0.4 percent, meaning one patient will die for every 250 people who are infected.
That rate suggests that COVID-19 (in Texas, at least) is much more deadly than the seasonal flu but not nearly as deadly as people initially feared. The projections that the CDC made in March, which predicted that as many as 1.7 million Americans could die from COVID-19 without intervention, assumed an IFR of 0.8 percent. Around the same time, researchers at Imperial College produced a highly influential worst-case scenario in which 2.2 million Americans died, based on an IFR of 0.9 percent. The CDC’s most recent “best estimate” of the nationwide IFR in the United States, based on data from other countries, is 0.65 percent.
Based on antibody screening of blood drawn for routine diagnostic tests unrelated to COVID-19, the CDC has produced state-by-state infection estimates. While the patients whose blood was used in those studies may not be representative of the general population, the CDC’s estimates indicate that the IFR varies widely from one state to another. As of mid-August, for example, the implied IFR was at least 10 times higher in Connecticut than in Idaho, Nebraska, Oregon, Tennessee, or Utah.
Possible explanations for these interstate differences include age demographics, the prevalence of preexisting medical conditions, the quality and capacity of local health care systems (including the extent to which they are strained by the pandemic), and population density, which not only makes it easier for the virus to move from person to person but may result in larger virus doses and more dangerous infections. Another factor could be the timing of each state’s epidemic, since the development of more effective treatments may have improved outcomes for people infected more recently.
Based on samples drawn in July, the CDC estimated that 1.6 million people had been infected in Texas by mid-August. Combined with the contemporaneous death count, that implied an IFR of about 0.66 percent—very close to the CDC’s nationwide estimate. The gap between that implied IFR and the one suggested by the University of Texas at Austin model may be partly due to differences in methodology: The CDC estimated the number of infections based on the prevalence of antibodies in blood samples, while the new estimate is based on a less direct (although perhaps more representative) approach. But the difference may also reflect factors that have made COVID-19 less deadly over time, including a younger, healthier mix of patients and improved treatment.
In addition to the interstate differences, the lethality of COVID-19 varies dramatically by age group. According to the CDC’s most recent national estimates, for example, the IFR for people in their 70s (5.4 percent) is 1,800 times as high as the IFR for people 19 or younger (0.003 percent). The CDC’s IFR estimate is 0.02 percent for 20-to-49-year-olds and 0.5 percent for 50-to-69-year-olds.
[This post has been revised to correct the comparison with the seasonal flu.]