Long COVID risk by pre-infection symptoms and functional status: A retrospective cohort study of data from the All of Us Research Program.

IMPORTANCEOver seven million U.S. adults experience "long COVID", or persistent health issues after COVID-19. Multiple guidelines recommend the inclusion of functional status in long COVID diagnostic criteria, but more evidence is needed to guide this recommendation. This study explores the adjusted odds of developing long COVID by pre-infection symptoms and functional status, and the feasibility of estimating functional status using health records data. DESIGNRetrospective cohort study in a multicenter national longitudinal cohort of U.S. adults with history of COVID-19, using health records and survey responses through July 2022 (All of Us CDR 7.0). EXPOSURE(S)Pre-infection (-5 years) incidences of (a) at least one symptom common in long COVID, and (b) functional status, indicated by All of Us baseline survey responses and diagnostic/procedure/billing codes. Disease and demographic data covariates were included in the adjusted models. RESULTSN = 65,464 met inclusion criteria (n=40,655 had post-infection occurrences of at least one symptom (long COVID group), while n=24,809 had none). Adjusted odds ratios within 99% confidence intervals [99% CI] of developing long COVID increased with lower pre-infection self-reported mental health ("Good" compared to "Excellent" AOR=1.14 [1.04,1.25], P>0.000), and more pre-infection symptoms (compared to the median of four, people with zero had much lower odds (AOR=0.15 [0.04,0.61], P=0.008). Adjusted odds were not significantly affected by any single pre-infection symptom, self-rated physical ability, or clinical documentation of functional impairment. CONCLUSIONSGreater pre-illness symptom burden and lower self-rated mental health increased the odds of long COVID symptoms, after adjusting for demographics, variant, functional status, and individual symptoms. Long COVID represents a change from baseline functioning and health, even in people with pre-infection incident symptoms and functional impairments. This estimation of pre-infection functional status using harmonized electronic health records data demonstrated the feasibility of these data in developing the diagnostic utility of functional status changes in long COVID.