CoChat
Sign Up
Cited by 2,881Long-Term Effects of COVID-19COVID-19 Clinical Research StudiesCOVID-19 and Mental Health

The DOI matches its Crossref record; 44 of the 46 references it cites were found in a registry or an index.

Characterizing long COVID in an international cohort: 7 months of symptoms and their impact

Open access
Davis, Hannah E., Assaf, Gina S., McCorkell, Lisa, Wei, Hannah, Low, Ryan J., Re'em, Yochai, Redfield, Signe, Austin, Jared P., Akrami, Athena·2021·eClinicalMedicine
DOI 10.1016/j.eclinm.2021.101019PMC PMC8280690

Published version of a preprint

Open this paper in CoChat — free accountOpen PDF

Abstract

BACKGROUND: A significant number of patients with COVID-19 experience prolonged symptoms, known as Long COVID. Few systematic studies have investigated this population, particularly in outpatient settings. Hence, relatively little is known about symptom makeup and severity, expected clinical course, impact on daily functioning, and return to baseline health. METHODS: We conducted an online survey of people with suspected and confirmed COVID-19, distributed via COVID-19 support groups (e.g. Body Politic, Long COVID Support Group, Long Haul COVID Fighters) and social media (e.g. Twitter, Facebook). Data were collected from September 6, 2020 to November 25, 2020. We analyzed responses from 3762 participants with confirmed (diagnostic/antibody positive; 1020) or suspected (diagnostic/antibody negative or untested; 2742) COVID-19, from 56 countries, with illness lasting over 28 days and onset prior to June 2020. We estimated the prevalence of 203 symptoms in 10 organ systems and traced 66 symptoms over seven months. We measured the impact on life, work, and return to baseline health. FINDINGS: For the majority of respondents (>91%), the time to recovery exceeded 35 weeks. During their illness, participants experienced an average of 55.9+/- 25.5 (mean+/-STD) symptoms, across an average of 9.1 organ systems. The most frequent symptoms after month 6 were fatigue, post-exertional malaise, and cognitive dysfunction. Symptoms varied in their prevalence over time, and we identified three symptom clusters, each with a characteristic temporal profile. 85.9% of participants (95% CI, 84.8% to 87.0%) experienced relapses, primarily triggered by exercise, physical or mental activity, and stress. 86.7% (85.6% to 92.5%) of unrecovered respondents were experiencing fatigue at the time of survey, compared to 44.7% (38.5% to 50.5%) of recovered respondents. 1700 respondents (45.2%) required a reduced work schedule compared to pre-illness, and an additional 839 (22.3%) were not working at the time of survey due to illness. Cognitive dysfunction or memory issues were common across all age groups (~88%). Except for loss of smell and taste, the prevalence and trajectory of all symptoms were similar between groups with confirmed and suspected COVID-19. INTERPRETATION: Patients with Long COVID report prolonged, multisystem involvement and significant disability. By seven months, many patients have not yet recovered (mainly from systemic and neurological/cognitive symptoms), have not returned to previous levels of work, and continue to experience significant symptom burden. FUNDING: All authors contributed to this work in a voluntary capacity. The cost of survey hosting (on Qualtrics) and publication fee was covered by AA's research grant (Wellcome Trust/Gatsby Charity via Sainsbury Wellcome center, UCL).

Overview

The claims and statistics of this paper haven’t been checked yet.

Run the check — free, no account needed. It reads the open-access text and checks the sources it cites.

A few free checks a day · papers already verified open instantly · results are public

Integrity signals

Registry

Crossref · 10.1016/j.eclinm.2021.101019

Cites

44 of 44 references verified0 citation errors · 0 not checkable

Statistics

Figures

Fig. 1 a) Survival function (Kaplan-Meier estimator), characterizing the distribution of disease duration for those who tested Negative (blue) on both diagnostic (RT-PCR/antigen) and antibody tests, those who tested Positive (orange) in either diagnostic or antibody test, and All (green) respondents. The Y axis indicates the probability that symptoms will persist longer than the time specified on the X axis. b) Probability of each symptom severity score over time. c) Average number of reported symptoms over time for those who recovered in less than 90 days (n = 154), or otherwise experienced symptoms for more than 90 days (n = 3505). a-c) In all plots, times are relative to initial symptom onset. Shaded regions represent 95% simultaneous confidence bands.
Fig. 1 a) Survival function (Kaplan-Meier estimator), characterizing the distribution of disease duration for those who tested Negative (blue) on both diagnostic (RT-PCR/antigen) and antibody tests, those who tested Positive (orange) in either diagnostic or antibody test, and All (green) respondents. The Y axis indicates the probability that symptoms will persist longer than the time specified on the X axis. b) Probability of each symptom severity score over time. c) Average number of reported symptoms over time for those who recovered in less than 90 days (n = 154), or otherwise experienced symptoms for more than 90 days (n = 3505). a-c) In all plots, times are relative to initial symptom onset. Shaded regions represent 95% simultaneous confidence bands.CC BY · PubMed Central
Fig. 2 Symptom prevalence estimates (non-neuropsychiatric symptoms). Bars represent the percentage of respondents who experienced each symptom at any point in their illness. Symptoms are categorized by the affected organ systems. When all rows in a given panel use the same denominator, the first row, labeled “All,” indicates the percentage of respondents who experienced any symptoms in that category. Error bars are bootstrap 95% confidence intervals. In Fig. 2b, Sexual dysfunction is broken up into male (Sexual dysfunction - M) and female (Sexual dysfunction - F). “Cis M” refers to cisgender males, “Cis F” refers to cisgender females, and cisgender females are further broken down by age group: “Cis F <40″ indicates cisgender females age 39 or younger, “Cis F in 40s” indicates cisgender females age 40 to 49, and “Cis F >49″ indicates cisgender females age 50 or older.
Fig. 2 Symptom prevalence estimates (non-neuropsychiatric symptoms). Bars represent the percentage of respondents who experienced each symptom at any point in their illness. Symptoms are categorized by the affected organ systems. When all rows in a given panel use the same denominator, the first row, labeled “All,” indicates the percentage of respondents who experienced any symptoms in that category. Error bars are bootstrap 95% confidence intervals. In Fig. 2b, Sexual dysfunction is broken up into male (Sexual dysfunction - M) and female (Sexual dysfunction - F). “Cis M” refers to cisgender males, “Cis F” refers to cisgender females, and cisgender females are further broken down by age group: “Cis F <40″ indicates cisgender females age 39 or younger, “Cis F in 40s” indicates cisgender females age 40 to 49, and “Cis F >49″ indicates cisgender females age 50 or older.CC BY · PubMed Central
Fig. 3 Symptom prevalence estimates for neuropsychiatric symptoms. Similar to Fig. 2, for neuropsychiatric symptoms, divided into nine sub-categories. Each bar represents the percentage of respondents who experienced that symptom. Error bars are bootstrap 95% confidence intervals.
Fig. 3 Symptom prevalence estimates for neuropsychiatric symptoms. Similar to Fig. 2, for neuropsychiatric symptoms, divided into nine sub-categories. Each bar represents the percentage of respondents who experienced that symptom. Error bars are bootstrap 95% confidence intervals.CC BY · PubMed Central
Fig. 4 Symptom time courses. Plotted time courses represent the estimated probability of experiencing each symptom at each time point, given that recovery has not yet occurred (see Methods). Times are relative to initial illness onset. Symptoms are grouped according to the affected organ systems. Shaded regions show 95% simultaneous confidence bands, estimated separately for each symptom.
Fig. 4 Symptom time courses. Plotted time courses represent the estimated probability of experiencing each symptom at each time point, given that recovery has not yet occurred (see Methods). Times are relative to initial illness onset. Symptoms are grouped according to the affected organ systems. Shaded regions show 95% simultaneous confidence bands, estimated separately for each symptom.CC BY · PubMed Central
Fig. 5 Symptom onset times. Heatmap shows the estimated probability distribution of the onset time for each symptom. White points and error bars show the mean onset time and 95% pointwise confidence intervals. Symptoms are sorted by mean onset time.
Fig. 5 Symptom onset times. Heatmap shows the estimated probability distribution of the onset time for each symptom. White points and error bars show the mean onset time and 95% pointwise confidence intervals. Symptoms are sorted by mean onset time.CC BY · PubMed Central
Fig. 6 Symptom clusters, based on temporal similarities. Plots (top row) show time courses for the symptoms in each cluster (in gray) and their mean (Cluster 1 in blue, Cluster 2 in orange, Cluster 3 in green). Time courses have been scaled separately for each symptom (by root mean squared amplitude) to visually compare their shapes. The table lists symptoms in each cluster, grouped by the affected organ systems. The heatmap (bottom row) shows time courses for all symptoms, sorted such that similarly shaped time courses are adjacent (see Methods). Columns have been scaled by their maximum amplitudes for visual comparison. Symptoms are numbered according to their table entries.
Fig. 6 Symptom clusters, based on temporal similarities. Plots (top row) show time courses for the symptoms in each cluster (in gray) and their mean (Cluster 1 in blue, Cluster 2 in orange, Cluster 3 in green). Time courses have been scaled separately for each symptom (by root mean squared amplitude) to visually compare their shapes. The table lists symptoms in each cluster, grouped by the affected organ systems. The heatmap (bottom row) shows time courses for all symptoms, sorted such that similarly shaped time courses are adjacent (see Methods). Columns have been scaled by their maximum amplitudes for visual comparison. Symptoms are numbered according to their table entries.CC BY · PubMed Central
Fig. 7 Memory and cognitive dysfunction. a) Percentage of respondents in six age groups who experienced different types of memory impairments. b) Same as (a) for cognitive dysfunction. c) Impact of memory and cognitive dysfunction on work (for those who work), for different age groups. Participants were asked to rate the impact by choosing one of the four options “Able, Mildly unable, Moderately unable, and Severely unable”. d) Overall impact of memory and cognitive dysfunction on daily life. Participants to whom the question was not applicable were excluded. Error bars show bootstrap 95% confidence intervals.
Fig. 7 Memory and cognitive dysfunction. a) Percentage of respondents in six age groups who experienced different types of memory impairments. b) Same as (a) for cognitive dysfunction. c) Impact of memory and cognitive dysfunction on work (for those who work), for different age groups. Participants were asked to rate the impact by choosing one of the four options “Able, Mildly unable, Moderately unable, and Severely unable”. d) Overall impact of memory and cognitive dysfunction on daily life. Participants to whom the question was not applicable were excluded. Error bars show bootstrap 95% confidence intervals.CC BY · PubMed Central
Fig. 8 Worsening or relapse of symptoms after physical or mental activity (post-exertional malaise). When does it start (a), how long does it last (b), and how severe is it? (c) (all patients who experienced PEM, n = 3350). Error bars are bootstrap 95% confidence intervals.
Fig. 8 Worsening or relapse of symptoms after physical or mental activity (post-exertional malaise). When does it start (a), how long does it last (b), and how severe is it? (c) (all patients who experienced PEM, n = 3350). Error bars are bootstrap 95% confidence intervals.CC BY · PubMed Central
Fig. 9 Symptom time courses for participants with COVID-19 confirmed via testing vs. rest of the population. Plots show symptom time courses (similar to Fig. 4) for respondents who were confirmed COVID-positive via diagnostic or antibody testing (orange) vs those without a positive confirmation (untested or tested negative, blue). Shaded regions show simultaneous 95% confidence bands. Symptom names are colored according to the affected organ systems.
Fig. 9 Symptom time courses for participants with COVID-19 confirmed via testing vs. rest of the population. Plots show symptom time courses (similar to Fig. 4) for respondents who were confirmed COVID-positive via diagnostic or antibody testing (orange) vs those without a positive confirmation (untested or tested negative, blue). Shaded regions show simultaneous 95% confidence bands. Symptom names are colored according to the affected organ systems.CC BY · PubMed Central
Fig. 10 Triggers and experience of relapses. a. Triggers for relapses/worsening of symptoms b. Experience of symptoms over time and relapses. Error bars are bootstrap 95% confidence intervals.
Fig. 10 Triggers and experience of relapses. a. Triggers for relapses/worsening of symptoms b. Experience of symptoms over time and relapses. Error bars are bootstrap 95% confidence intervals.CC BY · PubMed Central
Fig. 11 Remaining symptoms after six months. a) Symptoms remaining after six months. b) Symptoms remaining after six months for respondents still experiencing PEM after six months (orange), respondents not experiencing PEM after six months (green), and respondents who never experienced PEM (blue). c) Average number of symptoms over time for each group in (b). Error bars are bootstrap 95% confidence intervals.
Fig. 11 Remaining symptoms after six months. a) Symptoms remaining after six months. b) Symptoms remaining after six months for respondents still experiencing PEM after six months (orange), respondents not experiencing PEM after six months (green), and respondents who never experienced PEM (blue). c) Average number of symptoms over time for each group in (b). Error bars are bootstrap 95% confidence intervals.CC BY · PubMed Central
Fig. 12 Return to baseline and work impact. a) Distribution of Fatigue Assessment Scale scores for recovered (n = 257, blue) and unrecovered (n = 3505, yellow) population. The vertical dashed lines indicate the range for “No fatigue”[10], [11], [12], [13], [14], [15], [16], [17], [18], [19], [20], [21], “Fatigue”[22], [23], [24], [25], [26], [27], [28], [29], [30], [31], [32], [33], [34], and “Extreme” (>=35). Mean values for each distribution are also marked. b) Percentage of participants in each of the three categories. c) Distribution of scores in response to “return to pre-COVID” health baseline, where 0 indicates worst (most different from baseline) and 100 indicates best (most similar to baseline). d) Working status due to COVID-19. Error bars show 95% simultaneous confidence interval.
Fig. 12 Return to baseline and work impact. a) Distribution of Fatigue Assessment Scale scores for recovered (n = 257, blue) and unrecovered (n = 3505, yellow) population. The vertical dashed lines indicate the range for “No fatigue”[10], [11], [12], [13], [14], [15], [16], [17], [18], [19], [20], [21], “Fatigue”[22], [23], [24], [25], [26], [27], [28], [29], [30], [31], [32], [33], [34], and “Extreme” (>=35). Mean values for each distribution are also marked. b) Percentage of participants in each of the three categories. c) Distribution of scores in response to “return to pre-COVID” health baseline, where 0 indicates worst (most different from baseline) and 100 indicates best (most similar to baseline). d) Working status due to COVID-19. Error bars show 95% simultaneous confidence interval.CC BY · PubMed Central

Verification

Claims the paper makes

The claims and statistics of this paper haven’t been checked yet.

Run the check — free, no account needed. It reads the open-access text and checks the sources it cites.

A few free checks a day · papers already verified open instantly · results are public

Claims other papers make about this one

Each is checked against this paper’s own text: does it say what the citing paper says it does?

Self-check

Percentages, sums and reported differences re-computed from the paper’s own numbers.

The self-check needs the paper’s full text.

Citations

Cites

Show all 46 references

Research works printed but not linked: 3

  1. NIH workshop on post-acute sequelae of COVID-19 (Day 2) [Internet]. Available from: https://videocast.nih.gov/watch=38879
  2. Tang, 2020, 26, 1337
  3. Williams, 2021, A two-phase, single cohort study of COVID-19, Antibody Sera-Surveillance, 6

Cited by

Sorted by citations

See all 2,881

Related

Metadata

Authors
Davis, H. E., Assaf, G. S., McCorkell, L., Wei, H., Low, R. J., Re'em, Y., Redfield, S., Austin, J. P., & Akrami, A.
Affiliations
  • Patient-Led Research Collaborative (US)
  • Oregon Health & Science University (US)
Corresponding author
Akrami, Athena
ORCID
Venue
eClinicalMedicine
Publisher
Elsevier BV
Published
2021-08
Volume
38
Pages
101019
ISSN
2589-5370 (ISSN-L)
DOI
10.1016/j.eclinm.2021.101019
PMC
PMC8280690
OpenAlex
W3184339862
Cited by
2,881
References
46
Citations by year
2026: 299 · 2025: 492 · 2024: 594 · 2023: 692 · 2022: 617 · 2021: 186
FWCI
249.99
Type
journal article
Language
en
Open access
Licence
Locations
  • EClinicalMedicine (publishedVersion, cc-by)
  • PubMed Central (publishedVersion, cc-by)
  • PubMed (publishedVersion)
  • DOAJ (DOAJ: Directory of Open Access Journals) (submittedVersion, cc-by-sa)
Funders
Wellcome Trust
Topics
  • Health Sciences › Medicine › Clinical Neurology › Long-Term Effects of COVID-19
  • Health Sciences › Medicine › Infectious Diseases › COVID-19 Clinical Research Studies
  • Social Sciences › Psychology › Clinical Psychology › COVID-19 and Mental Health
Keywords
long COVID, symptom clusters, fatigue, post-exertional malaise, cognitive impairment, symptom recurrence, work disability, loss of taste and smell, outpatients

About these checks

Registry, retraction, Cites and statistics results are computed, not judged, with one exception: a printed reference that no index lists is read once by a model before it counts as a citation error. Claims are extracted by a model and kept only when their quote appears verbatim in the text; each cited source is then checked, with a model’s help, for whether it says what the claim relies on. Those verdicts can be wrong — read the claim’s quote and the cited source before relying on one.

Statistics and arithmetic are recomputed, not judged.

Confidence bands are shown only for fields calibrated on our annotated test set. None is calibrated yet, so no band is shown.

What counts as a claim ›How we measure confidence ›How we check this ›

Are you an author, or do you see a problem with this page? Request a review