Three-year outcomes of post-acute sequelae of COVID-19

SARS-CoV-2 infection is associated with long-term health effects across multiple organ systems, collectively termed Long COVID or PASC. Prior research up to 1–2 years post-infection showed that risks for some conditions abate after the first year, but many persist to 2 years, particularly among those hospitalized during acute COVID-19. Approximately one quarter of the 2-year cumulative burden of PASC accrues in the second year. However, evidence beyond 2 years is scarce, leaving uncertainty about the persistence of risks into the third year and the possible emergence of latent effects. This study aims to comprehensively characterize 3-year risks and burdens of death and PASC among individuals with COVID-19, stratified by acute care setting (non-hospitalized vs hospitalized), to inform long-term care and planning.

The paper synthesizes prior findings that Long COVID affects nearly every organ system and that risk trajectories vary by outcome and severity of acute illness. Studies at 1–2 years post-infection documented persistent risks for multiple conditions, with higher and more prolonged risks among those hospitalized during acute infection. Approximately 25% of the 2-year cumulative burden accrued in the second year. The literature also suggests potential for long-term or latent sequelae after viral infections in general (e.g., EBV and multiple sclerosis, post-polio syndrome), and emerging mechanistic insights indicate prolonged immune activation and viral RNA persistence after SARS-CoV-2 infection. Despite these findings, there has been limited characterization of outcomes at 3 years, motivating the present analysis.

Design and setting: Observational cohort study using the US Department of Veterans Affairs (VA) national electronic health databases, the largest integrated healthcare system in the US. Ethics approval was obtained with a waiver of informed consent. Cohort construction: The exposure cohort included VA users with first documented SARS-CoV-2 infection between March 1 and December 31, 2020 (ancestral variant period, pre-vaccination/antivirals), alive at 30 days post-index (T0), yielding 135,161 individuals. Acute-phase hospitalization was defined as inpatient admission within 7 days before to 30 days after a positive test; exposed individuals were stratified into non-hospitalized (n=114,864) and hospitalized (n=20,297) groups. Follow-up began 30 days after T0 and continued until death, reinfection, 1,080 days, or December 31, 2023. Controls: A control cohort of 5,206,835 VA users without a positive SARS-CoV-2 test from March 1, 2020 to March 30, 2021 was assembled. Controls were assigned a synthetic T0 mirroring the distribution of T0 in the COVID-19 cohort and required similar prior VA utilization. Follow-up mirrored the exposed cohort. Data sources: VA Corporate Data Warehouse (diagnoses, procedures, labs, medications), VA COVID-19 Shared Data Resource (PCR/antigen results), Medicare data (inpatient/outpatient), vital status files, and the Area Deprivation Index for socioeconomic context. Outcomes: Pre-specified 80 incident outcomes recognized as PASC were defined using ICD-10 codes, prescriptions, and laboratory data. Outcomes were grouped into 10 organ systems: cardiovascular; coagulation and hematologic; fatigue; gastrointestinal; kidney; mental health; metabolic; musculoskeletal; neurologic; pulmonary. Incident outcomes were counted if first occurring from 30 days post-T0 within each year-long interval. Composite outcomes included overall PASC and organ-system aggregates. Disability-adjusted life years (DALYs) were estimated using Global Burden of Disease methodology by weighting outcomes with health burden coefficients and summing the weighted occurrences. Covariates: Demographics (age, sex, race), ADI, smoking status; baseline eGFR, blood pressure, BMI; healthcare utilization (long-term care, influenza vaccination history, counts of Medicare and VA encounters, prescriptions, and lab panels); comorbidities (anxiety, cardiovascular, cerebrovascular, CKD, chronic lung disease, dementia, depression, diabetes, immunocompromised status, peripheral artery disease); calendar week and geographic region. Missing data for eGFR, BP, BMI were imputed via multivariate imputation by chained equations. Statistical analysis: Inverse probability weighting (IPW) using logistic regression propensity scores balanced baseline covariates between each COVID-19 group and the control group; weights were truncated at the 99.9th percentile. Risks were estimated in three periods: days 30–360 (year 1), 361–720 (year 2), and 721–1,080 (year 3); participants with prior occurrence of an outcome were excluded from risk sets for subsequent periods. Weighted generalized estimating equations with Poisson distribution and log link estimated incidence rate ratios (IRRs), absolute rates, excess rates per 1,000 persons, cumulative rates, and DALYs. Significance required 95% CIs excluding 1 (for IRRs) or 0 (for rate differences). Parametric bootstrapping (1,000 iterations) generated 95% CIs. Sensitivity analyses included doubly robust models, zero-inflated Poisson models, not censoring on reinfection, high-dimensional covariate adjustment, alternative hospitalization definitions, different weight truncation, complete-case analysis, Fine-Gray competing risk models, inverse probability of censoring weights, and a narrower PASC definition (73 outcomes). A negative outcome control (incident neoplasms) assessed residual confounding.

- Cohort and follow-up: 114,864 non-hospitalized and 20,297 hospitalized COVID-19 patients; 5,206,835 controls. Total person-years: 16,025,988 (non-hospitalized: 344,592; hospitalized: 60,891; controls: 15,620,505). All participants had full 3-year follow-up. Non-hospitalized COVID-19: - Death: Elevated in year 1 (IRR 1.58; 95% CI 1.53–1.62; excess 16.20 per 1,000; 95% CI 14.90–17.51), not elevated in year 2 (IRR 0.97; 0.94–1.00; excess −0.91; −2.20 to 0.38) or year 3 (IRR 1.01; 0.97–1.04; excess 0.22; −1.14 to 1.58). - Overall PASC IRRs: 1.23 (1.22–1.25), 1.16 (1.14–1.18), 1.05 (1.03–1.08) in years 1–3. - Cumulative number of sequelae: 378.7 per 1,000 (95% CI 356.6–401.1): year 1 = 212.3 (197.5–227.0), year 2 = 125.0 (107.2–142.7), year 3 = 41.2 (20.2–62.3). Percent contributions: 56.1%, 33.0%, 10.9%. - Cumulative DALYs: 91.2 per 1,000 (81.6–101.0): year 1 = 54.3 (47.9–60.7), year 2 = 27.3 (19.5–35.0), year 3 = 9.6 (0.4–18.7). Percent contributions: 59.6%, 29.9%, 10.5%. - Organ systems with elevated risk in year 3: neurologic (14.2 sequelae per 1,000; 95% CI 8.4–20.0; 7.2 DALYs; 1.0–13.4), pulmonary (11.1; 6.3–15.8; 1.5 DALYs; 0.6–2.4), gastrointestinal (7.4; 0.8–13.9; 0.8 DALYs; 0.02–1.5). Across 3 years, top DALY contributors: neurologic (32.2), cardiovascular (16.9), pulmonary (10.5), coagulation/hematologic (9.7), mental (7.2) per 1,000 persons. Hospitalized COVID-19: - Death: Elevated in all years—year 1 IRR 3.17 (3.00–3.33; excess 58.85 per 1,000; 54.37–63.33), year 2 IRR 1.44 (1.34–1.55; excess 14.16; 10.25–18.06), year 3 IRR 1.29 (1.19–1.40; excess 8.16; 4.37–11.96). - Overall PASC IRRs: 2.82 (2.76–2.89), 1.57 (1.49–1.66), 1.34 (1.24–1.45) in years 1–3. - Cumulative number of sequelae: 2,391.7 per 1,000 (2,316.0–2,472.3): year 1 = 1,696.6 (1,636.6–1,756.6), year 2 = 443.3 (375.1–511.6), year 3 = 252.8 (176.9–328.7). Percent contributions: 70.9%, 18.5%, 10.6%. - Cumulative DALYs: 766.2 per 1,000 (731.7–803.3): year 1 = 527.1 (499.5–554.7), year 2 = 149.6 (118.1–181.0), year 3 = 90.0 (55.2–124.8). Percent contributions: 68.8%, 19.5%, 11.7%. - Year 3 organ-system sequelae per 1,000: cardiovascular 41.3 (14.1–68.4; 25.5 DALYs), mental 43.1 (6.2–80.1; 20.1 DALYs), neurologic 29.9 (19.3–40.6; 11.9 DALYs), coagulation/hematologic 21.3 (0.3–42.2; 10.0 DALYs), pulmonary 28.7 (10.1–47.3; 4.8 DALYs), fatigue 11.7 (0.2–23.2; 2.6 DALYs), gastrointestinal 46.9 (24.1–69.8; 5.5 DALYs). Across 3 years, top DALY contributors: cardiovascular (173.5), mental (156.3), neurologic (138.5), coagulation/hematologic (107.5), kidney (72.4) per 1,000 persons. Hospitalized vs non-hospitalized comparison: - At each timepoint, hospitalized patients had significantly higher risks and burdens of overall PASC and across organ systems, with longer risk horizons. Cumulative DALYs at 3 years were 8.4-fold higher in hospitalized (766.2) vs non-hospitalized (91.2) individuals. Negative outcome control: - Associations with incident neoplasms were null in both groups across all years (non-hospitalized IRRs ~1.03, 0.94, 0.95; hospitalized IRRs ~0.93, 0.92, 0.93), supporting reduced residual confounding. Sensitivity analyses: - Multiple alternative specifications yielded results consistent with primary analyses.

The study demonstrates that risks of death and PASC generally decline over 3 years following SARS-CoV-2 infection, but important residual risks persist. Among non-hospitalized individuals, excess mortality risk resolved after the first year, with smaller but measurable third-year PASC burden, particularly in neurologic, pulmonary, and gastrointestinal systems. In contrast, those hospitalized during acute COVID-19 retained elevated mortality risk into the third year and substantial PASC-related health loss across multiple organ systems. These findings clarify the long-term risk horizon of PASC and underscore that acute disease severity is a key determinant of both the magnitude and persistence of long-term adverse outcomes. Potential mechanisms include greater comorbidity burden, immune dysregulation, and possible viral persistence in tissues among those with severe disease. From a public health perspective, although individual risk is lower after mild infection, the larger number of people with mild disease implies a significant population-level PASC burden. The results support strategies to prevent severe disease (vaccination, antivirals), early recognition and management of PASC, and continued monitoring to capture late-emerging effects.

Over 3 years of follow-up, risks of death and PASC attenuate but do not fully resolve, especially among individuals hospitalized during acute COVID-19, who exhibit persistent excess mortality and substantial health loss. The study extends prior 1–2 year evidence, quantifying the third-year contribution to sequelae and DALYs by organ system. Future research should evaluate longer-term trajectories beyond 3 years, include vaccinated and antiviral-treated populations and infections with later variants, elucidate mechanisms of persistent organ-specific risks, and assess interventions to prevent or mitigate PASC in both mild and severe cases.

Generalizability may be limited as the VA population is predominantly older, male, and White. Cohorts were enrolled in 2020 (pre-vaccination and pre-antivirals, ancestral variant), so risks may differ for vaccinated individuals, those treated with antivirals, or infected with later variants (e.g., Omicron). Despite extensive covariate adjustment, residual confounding and misclassification bias (including undetected infections in controls) cannot be excluded. Outcomes were limited to 80 pre-specified sequelae; uncharacterized sequelae were not captured. DALY estimates rely on GBD health burden coefficients assumed to be etiology-invariant. Absolute burden estimates are influenced by baseline rates and sample sizes differ between groups, affecting precision of estimates. Risks and burdens were estimated in discrete yearly intervals and reflect average risk within those periods. VA data, while comprehensive, may miss some external test results or diagnoses.