Revised estimates of types and duration of long COVID symptoms based on claims records from 245 million US patients.

A testing-based claims analysis of 244.7 million US patients finds long COVID symptom attributability drops sharply versus unrestricted controls: 83.5% fewer at 30–119 d, 98.2% at 120–359 d, and 100% at 360–719 d.

• Why it matters: Unrestricted test-positive versus uninfected controls can be biased by selective testing, misattributing long-term effects of non-COVID conditions to COVID-19.
• What they did: Using 14.4 billion insurance claims, the study analyzed 614 clinical outcomes over 2 years, comparing test-positive to test-negative patients among tested individuals.
• The result: The testing-based design reduced attributable symptoms (43 vs. 262 at 30–119 d; 5 vs. 286 at 120–359 d; 0 vs. 369 at 360–719 d) and avoided unrelated outcomes being attributed to COVID-19, implying attributability beyond one year is uncertain.