Long COVID Recovery Varies by Symptoms: About Half Improve by Year Three

A three-year US follow-up study

A study published in the Annals of Internal Medicine found that symptoms can continue to improve years after a person develops Long COVID, although improvement does not necessarily mean a complete recovery. Researchers analyzed 1,019 adults in the RECOVER-Adult cohort who still met the study’s definition of likely Long COVID 12 months after their first SARS-CoV-2 infection.

The participants were recruited at 83 sites across 33 US states, Washington, DC, and Puerto Rico. They completed symptom surveys every three months. Researchers scored those surveys with the Long COVID Research Index, or LCRI. Adults with a score of 11 or higher one year after infection were included in the main analysis.

This design is important to interpreting the findings. The study did not follow a random sample of everyone who had COVID-19. It focused on adults who still crossed the study’s threshold for likely Long COVID at the one-year mark. The results therefore describe the course of symptoms in that group, not the probability that any person who contracts COVID-19 will develop persistent symptoms or recover within a set period.

Half met the sustained-improvement measure by year three

The primary outcome was not a participant saying they felt cured. It was an LCRI score below 11 at two consecutive study visits. By 24 months after infection, an estimated 36.2% of participants had met that measure. By 36 months, the estimate had risen to 50.1%.

Researchers also assessed two secondary outcomes. By 36 months, an estimated 56.4% of participants had reduced their LCRI score by at least half and rated their quality of life as good or better. Another 26.9% had an LCRI score of zero, meaning they no longer reported symptoms measured by the index.

Those percentages describe different outcomes. The 50.1% figure represents sustained improvement below the study threshold; it does not mean that half the participants were completely symptom-free. The 26.9% figure represents complete remission of the symptoms included in the LCRI. Treating the two figures as equivalent would overstate what the study found.

The primary 36-month estimate also has a confidence interval, from 46.1% to 54.0%. The zero-score estimate ranged from 23.3% to 30.5%. These ranges reflect statistical uncertainty around the cohort estimates. They do not provide a personal forecast for an individual patient.

Improvement differed across symptom patterns

The estimated likelihood of sustained improvement varied substantially by the predominant symptom pattern. It was about 60% among participants whose symptoms were mainly changes in smell or taste. It was about 19% among those with symptoms across multiple body systems. The difference suggests that an overall average can conceal very different experiences among people grouped under the broad label of Long COVID.

The researchers also reported differences by age and infection period. Estimated improvement was higher among adults aged 18 to 45 than among those aged 46 to 64. Participants whose first infection occurred during the Omicron period also had a higher estimated probability of improvement than those infected before Omicron. The study did not find a clear difference between women and men.

These are adjusted statistical associations within an observational cohort. They do not establish that age, a particular variant, or a specific symptom caused improvement. They also cannot determine why one person recovers while another with a similar symptom pattern continues to experience limitations.

The symptom groupings should be understood as research categories used to compare reported patterns. They are not proof that every participant in a category has the same underlying biological process. A person’s medical history, other conditions, symptom severity, and ability to function in daily life may all differ from those of other participants.

Study limitations and implications for care

This was a prospective observational study, not a clinical trial. It tracked symptoms over time but did not test whether a medication, rehabilitation program, or lifestyle change caused improvement. The authors said the findings are interpretable in a population with likely Long COVID and do not establish causal relationships between individual characteristics and symptom improvement.

The LCRI was developed as a research measure within the RECOVER initiative. Questions remain about how well one score can capture a complex condition that may affect several body systems. External experts have also questioned whether the index’s weighting of symptoms such as loss of smell or taste and chronic cough could influence comparisons among symptom groups. The index has not been independently validated across all populations.

Symptoms can fluctuate as well. A person may have a period of improvement and later experience a relapse. Two consecutive scores below a research threshold therefore do not guarantee that symptoms will never return. Complete remission under the index is a defined research outcome, not a guarantee of permanent recovery or a substitute for a clinical assessment.

The study adds a longer view of Long COVID: some people continue to improve well beyond the first year, while many others do not meet the study’s improvement measures even after three years. The authors emphasized the need for placebo-controlled clinical trials to establish which treatments are effective. Observing improvement over time cannot identify a treatment that caused it.

For patients, the findings should not be read as a promise that everyone will recover naturally, nor as evidence that persistent symptoms cannot improve. They show that the course can be prolonged and variable. Decisions about care still require an individual assessment of symptoms, daily functioning, medical history, and changes over time. The study offers population-level evidence, but it cannot provide a reliable timetable for any one person.