JAMA Netw Open
Long COVID hits one in six, twice what codes capture
Clinical takeaway: Expect long COVID to surface under other diagnostic labels in patients with prior infection. Most cases represent chronic conditions, not transient postviral symptoms.
The mismatch between true long COVID prevalence and what diagnostic codes register has been recognized but poorly quantified. The ICD-10 code designed for post-COVID conditions identifies fewer than 1% of COVID-19 survivors, and broader code-based approaches catch about 7%. This retrospective cohort study aimed to size the surveillance gap and characterize what is being missed.
Across 457,950 adults with prior COVID-19, an AI phenotyping algorithm identified long COVID in 16.28% of cases, more than double the rate from the best code-based surveillance, with regional rates ranging from 13.64% in Western Pennsylvania to 22.69% in Southern California. Extrapolated to the roughly 103 million documented US COVID-19 cases, the authors estimate about 15 million Americans are living with chronic post-COVID conditions, though they caution that figure depends on the availability of longitudinal records.
Of the 883 ICD-10 codes associated with long COVID manifestations, two-thirds represented chronic or potentially chronic conditions. Among identified cases, 89.31% had developed at least one chronic condition requiring sustained clinical management, representing 14.54% of all COVID-19 patients in the cohort. Cumulative prevalence increased steadily through mid-2024 across all four regions, indicating long COVID continues to accumulate even four years into the pandemic and well after widespread vaccination.
Researchers analyzed electronic health records across four US regions: 12 hospitals in New England, 1 in Southeast Texas, 5 in Southern California, and 40 in Western Pennsylvania. The phenotyping algorithm identifies symptom sequences appearing three or more months after COVID-19 and persisting at least two months, while excluding cases explained by preexisting conditions. Prior validation showed 80% precision.
The authors argue that surveillance has missed the shift from acute postviral syndrome to accumulating chronic disease. Capacity planning, disability programs, trial recruitment, and reimbursement models all rest on case counts that miss most affected patients.
"These patients are not absent from clinical care; they are absent from the diagnostic code that would identify them as long COVID patients," said lead author Jiazi Tian, MSc, a data scientist in the Clinical Augmented Intelligence Group at Mass General Brigham. "The cardiologist seeing new dysautonomia, the endocrinologist seeing new metabolic disease, the neurologist seeing unexplained cognitive complaints — some of these presentations are long COVID arriving without the label that would connect them to a COVID-19 infection."
Source: Tian J. JAMA Netw Open. 2026 May 27. Long COVID persistence and surveillance gaps across 58 US hospitals