A new study of more than 244 million Americans suggests COVID-19 causes far fewer lasting health problems than earlier research indicated, although the authors stress that long COVID remains a serious concern for many patients.
Researchers led by a team at Stanford University published the analysis in PNAS, the Proceedings of the National Academy of Sciences. They examined 14.4 billion health insurance claims from 244.7 million patients in the United States, covering the first four months after each person tested positive or negative for COVID-19 between March 2020 and March 2021. It is one of the largest attempts yet to measure how common long COVID is in the United States.
The main difference in this study was the comparison group. The team compared people who tested positive with people who tested negative by PCR test, instead of comparing them with a group that included people who were never tested. CIDRAP, a news service at the University of Minnesota, reported the findings on 9 November.
How the long COVID study was designed
The researchers tracked 614 negative health outcomes across three periods after infection: the early period, one to four months after infection; the middle period, five to 12 months; and the late period, 13 to 24 months. For each period, they counted how many of those outcomes could be tied to the infection itself.
Many people were tested for COVID-19 while seeking care for other illnesses, the researchers found. If those illnesses later caused health problems, and those people were counted in the comparison group, the problems could be wrongly blamed on COVID-19. Enrolling patients at the time of their first PCR test was meant to reduce that bias, and the approach still picked up the known acute symptoms of COVID-19.
What the numbers showed
In the early period, the researchers linked 43 symptoms to long COVID when they used test-negative controls. That was 83.5 percent less than the 262 outcomes linked to the virus when the comparison group included people who had not been tested. The gap shows how much the choice of comparison group can change the result.
The difference grew wider over time. In the middle period, five poor health outcomes were linked to COVID-19 in the test-based design, compared with 286 in the analysis that included untested people. In the late period, no negative health outcomes were linked to long COVID in the test-based design, compared with 369 in the analysis that included people who had not been tested.
Why earlier long COVID estimates ran high
Earlier studies often compared test-confirmed COVID-19 patients with groups that included untested controls, the authors wrote. That approach, they said, leads to health problems being attributed to COVID-19 infection when they might instead be due to other factors or to the conditions that prompted the visit and the test in the first place.
The new findings do not mean long COVID has gone away. The authors wrote that the condition remains a serious concern, but it may involve fewer symptoms, and ones of shorter duration and lower overall severity, than previously thought. That distinction matters for patients who live with symptoms for months, and for doctors trying to work out which problems the virus caused.
What the findings do not change
Separate research underlines that recovery can still be slow. A study published in the Annals of Internal Medicine and tracked by CIDRAP found that only about one in four adults with long COVID had reached full remission of symptoms three years after infection, while about half had seen a meaningful improvement by that point.
The Stanford team used the claims to track long COVID through the first year of the pandemic, before vaccines were widely available and before later variants appeared. The authors did not claim the results apply in the same way to infections in later years, and they did not study people who were never tested at all. The paper also relied on insurance claims, which record the care people received rather than every symptom they felt.
For now, the study offers a narrower picture of the burden left by COVID-19. It suggests that careful comparison groups matter as much as large datasets when researchers try to measure long COVID.