Long-term health impacts of COVID-19 among 242,712 adults in England

The study investigates the long-term health impacts of COVID-19 in adults in England, focusing on the persistence of symptoms (Long COVID) and associated effects on health and quality of life. Given varied estimates of symptom persistence due to heterogeneous study designs and a lack of representative population data, the authors aim to describe symptom duration among individuals with symptomatic SARS-CoV-2 infection, identify factors associated with persistence beyond 12 weeks and with recovery thereafter, and compare current self-reported health, quality of life, and symptom profiles among those with ongoing symptoms, those who have recovered, and those never infected. The work leverages the large, community-based REACT programme to include non-hospitalised individuals and a comparator cohort.

Prior research reports substantial variability in Long COVID prevalence due to differences in design, setting, follow-up, and definitions. A meta-analysis of 194 studies found pooled prevalence of persistent symptoms at mean 126 days of 52.6% among hospitalised and 34.5% among non-hospitalised patients, with common symptoms including fatigue, breathlessness, impaired sleep, pain, and impaired usual activity. The Long-CISS (Scotland) study reported 8% of symptomatic participants not recovered by 6 or 12 months and higher odds of numerous persistent symptoms compared with never-infected controls, notably changes in smell/taste, breathlessness, chest pain, palpitations, and confusion. The Dutch Lifelines COVID-19 cohort identified core symptoms of Long COVID including chest pain, breathing difficulties, painful breathing, muscle pain, heavy limbs, loss of smell/taste, temperature dysregulation, tingling extremities, and general tiredness. Previous REACT-2 estimates suggested 21.6% had symptoms at 12 weeks among those with evidence of prior infection but lacked a negative control group.

Design: Cross-sectional follow-up survey of adults (≥18 years) previously enrolled in the REACT-1 (PCR testing; 19 rounds, May 2020–Mar 2022) and REACT-2 (antibody testing; six rounds, Jun 2020–May 2021) programmes who consented to recontact and data linkage (n=2,494,309). Invitations: 800,000 adults invited via email between 1 Aug–1 Dec 2022; oversampling of individuals with prior persistent symptoms (≥12 weeks), those PCR-positive in REACT-1, and unvaccinated IgG-positive in REACT-2, plus a random sample of remaining eligible adults. Participation: 282,780 registered (35.3% of invited); 276,840 completed (97.9% of registered). Exclusions: Those with COVID-19 episode date within 12 weeks of survey completion were excluded from relevant analyses (n=24,142). Final analytic sample for main comparisons was n=242,712. Data collection: Online questionnaire on current health, functionality, and 29 symptoms; detailed COVID-19 history (PCR/LFD positives, symptom onset and resolution dates, severity), and validated instruments: EQ-5D-5L (quality of life), Dyspnoea-12 (for those with breathlessness), three PEM items from the DePaul Symptom Questionnaire (for those with fatigue), GAD-7 and PHQ-9 (mental health). Group definitions: Participants classified into (1) No COVID, (2) Asymptomatic or resolved short COVID <4 weeks, (3) Resolved short COVID ≥4 to <12 weeks, (4) Resolved persistent COVID ≥12 to <52 weeks, (5) Resolved persistent COVID ≥52 weeks, (6) Ongoing persistent COVID ≥12 to <52 weeks, (7) Ongoing persistent COVID ≥52 weeks. Confirmed COVID-19 defined as symptomatic infections with a positive PCR/LFD (self-reported, REACT-1 positives, unvaccinated REACT-2 IgG positives, or Pillar 2 records); asymptomatic infections defined as positive test with no reported symptoms. Re-infections counted if ≥90 days apart. Variant periods: Wild-type (pre-Dec 2020), Alpha (Dec 2020–Apr 2021), Delta (May 2021–mid-Dec 2021), Omicron (late Dec 2021 onwards). Data linkage: Linkage to UKHSA Pillar 2 PCR/LFD results and NHS Digital vaccination records using NHS number and identifiers. Statistical analysis: Descriptive statistics (counts, percentages; medians/IQRs or means/SDs). Comparisons across COVID-19 categories used chi-square tests, ANOVA or Kruskal–Wallis as appropriate. Logistic regression models (mutually adjusted for age, sex, ethnicity, IMD, comorbidities, smoking status, initial severity, vaccination status at infection, and dominant variant) estimated adjusted odds ratios (aOR) and 95% CIs for: (a) persistent symptoms lasting ≥12 and ≥52 weeks versus asymptomatic or symptoms resolved <4 weeks; (b) current symptom odds in those with ongoing persistent symptoms versus all others. Time-to-event analysis: Kaplan–Meier plots for time to symptom end date. Among those with symptoms ≥12 weeks, Accelerated Failure Time (AFT) models with log-normal distribution estimated adjusted time ratios (aTR) for factors associated with rate of recovery beyond 12 weeks (aTR >1 indicates slower recovery). All tests two-sided, significance p<0.05; analyses in STATA 15.

Participation and sample: 276,840/800,000 (35.3%) completed; 266,854/276,840 (96.4%) reported infection status; 157,668/266,854 (59.1%) had tested positive; 24,142 excluded due to <12 weeks follow-up; analytic sample in Table 1 n=242,712. Symptom duration: Among 130,251 symptomatic, test-confirmed infections, median duration 1.3 weeks (mean 5.4 weeks; range 1 day–3.0 years; IQR 6 days–2 weeks). Proportions with persistent symptoms: 10.2% >4 weeks, 7.5% ≥12 weeks, 5.2% ≥52 weeks. Persistence over time: Of those with persistent symptoms at 12 weeks, 69% still symptomatic at 52 weeks (31% recovered within a year). Risk factors for persistence (mutually adjusted logistic regression): Female vs male: aOR 1.42 (95% CI 1.35, 1.50) for ≥12 weeks; 1.49 (1.38, 1.62) for ≥52 weeks. Comorbidities vs none: one comorbidity aOR 1.31 (1.19, 1.44) for ≥12; 1.52 (1.31, 1.76) for ≥52; two or more aOR 1.46 (1.27, 1.75) for ≥12; 2.35 (1.85, 2.97) for ≥52. Initial severity vs mild: moderate aOR 1.76 (1.63, 1.89) for ≥12; 1.47 (1.32, 1.64) for ≥52; severe aOR 4.87 (4.52, 5.25) for ≥12; 3.55 (3.19, 3.96) for ≥52. Ethnicity: Asian vs White aOR 0.80 (0.69, 0.93) for ≥12; 0.71 (0.57, 0.88) for ≥52. Variant period vs Wild-type: Alpha aOR 0.60 (0.56, 0.64) for ≥12; 0.59 (0.54, 0.64) for ≥52. Delta aOR 0.38 (0.35, 0.41) for ≥12; 0.32 (0.29, 0.36) for ≥52. Omicron aOR 0.12 (0.11, 0.13) for ≥12 (insufficient follow-up for ≥52). Lower deprivation associated with reduced odds; slight suggestion of lower Long COVID risk in older vs younger adults. Recovery among those with ≥12-week symptoms (AFT models): Slower recovery: female aTR 1.14 (1.06, 1.23); comorbidities aTR 1.24 (1.08, 1.42) for one; 2.05 (1.58, 2.66) for ≥2. Faster recovery: other ethnicity aTR 0.63 (0.45, 0.89) and mixed aTR 0.75 (0.57, 0.99) vs White; least deprived areas aTR 0.78 (0.69, 0.88) vs most deprived; not-current smokers aTR 0.73 (0.62, 0.86) vs current smokers; later variants vs Wild-type: Alpha 0.79 (0.72, 0.86), Delta 0.89 (0.79, 0.99), Omicron 0.69 (0.61, 0.78). Current symptoms and quality of life: Among those with ongoing persistent symptoms, most common current symptoms were mild fatigue (66.9%), difficulty thinking or concentrating (54.9%), joint pains (54.6%). Compared to all others, the largest symptom odds increases were loss/change of smell (aOR 9.31 [8.64, 10.04]), loss/change of taste (aOR 8.47 [7.85, 9.15]), shortness of breath (aOR 6.69 [6.29, 7.12]), severe fatigue (aOR 6.19 [5.66, 6.77]), difficulty thinking/concentrating (aOR 4.97 [4.68, 5.27]), chest tightness/pain (aOR 4.71 [4.37, 5.08]), and poor memory (aOR 4.40 [4.15, 4.66]). Health-related quality of life and mental health were worse in ongoing persistent symptoms: higher number of current symptoms (median 7–8 vs 2–4), greater reduction in daily activities (a lot: up to 31.7%), lower EQ-VAS (means ~64.7–66.5 vs ~78), lower EQ-5D utility (0.75–0.78 vs 0.84–0.89), higher PHQ-9 ≥10 (up to 45.9%) and GAD-7 ≥10 (up to 26.1%). Post-exertional malaise was more common in ongoing persistent symptoms groups. Those who had recovered after long symptom duration had health status broadly similar to those with short illness or no COVID-19.

The study demonstrates that while most symptomatic SARS-CoV-2 infections in adults in England are short-lived, a significant minority experience persistent symptoms meeting Long COVID definitions, with a subset continuing beyond a year. Factors most strongly associated with persistence include female sex, comorbidity burden, greater initial severity, socioeconomic deprivation, and infection during the Wild-type period, while later variants are associated with lower risk and faster recovery, potentially reflecting population immunity and changes in viral characteristics. Symptom reporting was high across all groups, but certain symptoms—loss or change of smell/taste, shortness of breath, severe fatigue, and cognitive difficulties—were notably more specific to ongoing post-COVID conditions. Importantly, individuals who recover, even after prolonged symptoms, report health and quality-of-life metrics comparable to those without COVID-19 history or with short-duration illness. These results address the research questions by quantifying duration, identifying risk factors and recovery correlates, and characterizing current symptom and quality-of-life impacts in a large, community-based sample with appropriate comparison groups.

COVID-19 illness is generally of short duration, but a substantial minority of adults experience persistent and burdensome symptoms, with some lasting beyond a year. Female sex, comorbidities, higher initial severity, socioeconomic deprivation, and infection during the Wild-type period were associated with higher risk of persistence, while later variants were associated with lower risk and faster recovery. Individuals who recover after persistent symptoms typically return to health and quality-of-life levels similar to those without COVID-19 or with short-duration illness. These findings, derived from one of the largest community-based cohorts with never-infected and recovered comparators, inform healthcare planning and patient support. Future work should refine causal understanding of variant, immunity, and vaccination effects, develop targeted interventions for high-risk groups, and explore symptom trajectories qualitatively (e.g., ongoing in-depth interviews with REACT participants) and longitudinally.

Potential misclassification due to undetected infections in comparator groups, especially when universal testing was unavailable; selection bias from a 34.6% response rate and overrepresentation of females, older adults, White ethnicity, and less deprived groups; recall and reporting biases given retrospective symptom onset and resolution dates, and self-reported measures; time-varying confounders (behavior, seasonality, public health measures, knowledge/expectations) possibly influencing associations, particularly by variant period; measurement limitations of PHQ-9 for Long COVID due to somatic items potentially overestimating depression; absence of population-weighted prevalence estimates due to complex sampling and oversampling of positives and those with persistent symptoms; inability to map all affiliations perfectly; and insufficient follow-up time for assessing very long persistence after Omicron period.