Drooid Logo
Back to story perspectives

Full Breakdown

AI-Driven Study Reveals 10 Million Undiagnosed Long COVID Cases

6/2/2026, 1:37:44 AM

Context and Key Figures

Long COVID is defined as symptoms lasting at least three months after infection. The World Health Organization estimates 6 % of COVID patients develop chronic symptoms; the CDC reports 18 % of U.S. adults surveyed have experienced long COVID. MGB researchers, led by Hossein Estiri, applied AI to 460,000 medical records, publishing in *JAMA Network Open*. Experts include Ziyad Al-Aly (Washington University), MIT’s Beth Pollack, UMass Memorial’s Sandeep Jubbal, and Beth Stelson of the Patient-Led Research Collaborative.

Data & Statistics

The AI analysis found 16 % of COVID-positive patients develop long-term symptoms, equating to at least 10 million undiagnosed cases in the United States. Prior estimates range from 6 % (WHO) to 18 % (CDC). A 2025 simulation estimated $9,900–$11,600 per case annually, totaling over $8 billion in U.S. costs from 2025-2027, with more than 90 % from lost productivity.

Why It Matters: Healthcare and Economic Consequences

The undiagnosed population creates a blind spot, limiting targeted treatment and expanding chronic disease burden. Productivity losses affect employers nationwide, magnifying economic strain beyond direct medical costs. A shortage of long-COVID clinics and no definitive diagnostic test further impede care.

Official Statements & Responses

Estiri said the AI uncovered a major blind spot and warned that the burden of new chronic diseases is accumulating. Al-Aly called long COVID a serious global problem but called the study’s data unhelpful. Pollack called it a persistent, multi-system illness that deeply impacts quality of life. Jubbal noted diagnosis remains symptom-based because no test exists.

Criticism & Opposition

Al-Aly argued the analysis omitted key variables, possibly inflating prevalence. Other experts warned that diagnostic uncertainty and the tendency to attribute symptoms to mental-health issues highlight methodological limits.

On-the-Ground Patient Perspective

Beth Stelson, a long-COVID patient in the Patient-Led Research Collaborative, reports fluctuating symptoms, severe exhaustion, and ongoing functional limits six years after onset, describing the illness as relapsing with an ebb-and-flow pattern.

Conflicting Estimates & Gaps

Prevalence estimates differ: WHO 6 %, CDC 18 %, MGB 16 %. No accepted diagnostic test exists, and long-COVID clinics are scarce. Funding cuts in 2025, including the HHS Office of Long COVID Research closure and NIH program terminations, leave a national strategy void.

Verbatim Quotes

  • “We built an AI to map the post-pandemic reality, and it uncovered a 10 million-person blind spot in the American healthcare system,” — Hossein Estiri, MGB
  • “I think Long Covid is a serious national and global problem that demands attention from governments and international bodies,” — Ziyad Al-Aly, Washington University
  • “Not having a diagnostic test is a big limitation,” — Sandeep Jubbal, UMass Memorial Health
  • “While we've come miles in recognizing that long COVID is a medical disorder,” she said, “there is still some psychologizing that happens with the illness.” — Beth Stelson, Washington University

What’s Next: Policy and Research Directions

Stakeholders urge standardized diagnostic criteria, more long-COVID clinics, and renewed federal funding for longitudinal research. Embedding AI-derived case detection into public-health surveillance could improve tracking and resource allocation.