Article Text
PDF
Abstract
Objectives Fatigue is a pervasive clinical symptom in coronaviruses and may continue beyond the acute phase, lasting for several months or years. This systematic review and meta-analysis aimed to incorporate the current evidence for postinfection fatigue among survivors of SARS-CoV-2 and investigate associated factors.
Methods Embase, PsyINFO, Medline, CINAHL, CDSR, Open Grey, BioRxiv and MedRxiv were systematically searched from January 2019 to December 2021. Eligible records included all study designs in English. Outcomes were fatigue or vitality in adults with a confirmed diagnosis of SARS-CoV-2 measured at >30 days post infection. Non-confirmed cases were excluded. JBI risk of bias was assessed by three reviewers. Random effects model was used for the pooled proportion with 95% CIs. A mixed effects meta-regression of 35 prospective articles calculated change in fatigue overtime. Subgroup analyses explored specific group characteristics of study methodology. Heterogeneity was assessed using Cochran’s Q and I2 statistic. Egger’s tests for publication bias.
Results Database searches returned 14 262 records. Following deduplication and screening, 178 records were identified. 147 (n=48 466 participants) were included for the meta-analyses. Pooled prevalence was 41% (95% CI: 37% to 45%, k=147, I2=98%). Fatigue significantly reduced over time (−0.057, 95% CI: −107 to −0.008, k=35, I2=99.3%, p=0.05). A higher proportion of fatigue was found in studies using a valid scale (51%, 95% CI: 43% to 58%, k=36, I2=96.2%, p=0.004). No significant difference was found for fatigue by study design (p=0.272). Egger’s test indicated publication bias for all analyses except valid scales. Quality assessments indicated 4% at low risk of bias, 78% at moderate risk and 18% at high risk. Frequently reported associations were female gender, age, physical functioning, breathlessness and psychological distress.
Conclusion This study revealed that a significant proportion of survivors experienced fatigue following SARS-CoV-2 and their fatigue reduced overtime. Non-modifiable factors and psychological morbidity may contribute to ongoing fatigue and impede recovery.
PROSPERO registration number CRD42020201247.
- EPIDEMIOLOGY
- COVID-19
- Respiratory infections
Data availability statement
Data are available upon reasonable request. Data are available on request from the corresponding author.
This is an open access article distributed in accordance with the Creative Commons Attribution Non Commercial (CC BY-NC 4.0) license, which permits others to distribute, remix, adapt, build upon this work non-commercially, and license their derivative works on different terms, provided the original work is properly cited, appropriate credit is given, any changes made indicated, and the use is non-commercial. See: http://creativecommons.org/licenses/by-nc/4.0/.
Statistics from Altmetric.com
Strengths and limitations of this study
This review and meta-analysis was conducted using a significant sample size from a comprehensive search of the literature, including only confirmed cases.
Substantial unexplained heterogeneity between studies limits generalisability of our findings.
Only one reviewer screened and extracted the data from each study leaving the potential for missing articles and selection errors.
Outcome measures of fatigue were unvalidated in the majority of studies, limiting confidence in our estimates.
Total point prevalence was likely impacted by the predominance of hospitalised patients with potentially more severe disease.
Introduction
Fatigue may be characterised as tiredness or exhaustion as a result of physical or mental exertion or as a result of an illness or disease. The experience of fatigue is common and is usually short-lived but, for a small number of people, it can become long-lasting, associated with a number of impairments in daily living and quality of life.1 It is one of the most common presenting symptoms of coronaviruses.2 The current pandemic has also revealed a considerable burden of lasting symptoms,3–12 with approximately one in four people experiencing fatigue by one estimate.13 Systematic reviews indicate a pooled prevalence of post-COVID-19 fatigue to vary among 45%,14 52%15 and 64%.16 In previous epidemics, fatigue was enduring. In a follow-up of 90 SARS survivors 30 months post illness, for instance, 1 study found significantly lower vitality scores compared with Hong Kong population norms.17 A small study of Middle East respiratory syndrome patients revealed that 32.7% had clinically relevant chronic fatigue, according to their Fatigue Severity Scale (FSS) scores, at 18 months’ follow-up.18 Likewise, for a considerable number of patients with COVID-19, tiredness symptoms extend beyond 3 months and represent a larger burden of postinfection symptomology.19–41 A large study of 1142 hospitalised patients found that 61% had fatigue 7 months post COVID-19.42 Similarly, those who perceived themselves as experiencing ‘poor recovery’ had lower vitality on the 15D instrument, compared with those making a ‘full recovery’ (p<0.001) 1 year post illness.43
More severe disease, associated with being hospitalised or intensive care unit (ICU) admission, has been related to postillness fatigue.44–51 In a small cohort of 55 people, 30 days post discharge for COVID-19, each additional day of hospitalisation increased fatigue by 1.2.52 Apart from hospitalised patients, among non-hospitalised or those treated for milder disease, fatigue is persistent.53–61 In 359 patients, 63.4% reported significant fatigue up to 12 months post infection and were more likely than admitted patients to require referral for fatigue symptomology.62
Determinants of postillness fatigue include female gender63–66 and older age, although the latter relationship was not consistent. Being over 50 years was associated with fatigue severity in some studies,52 67 68 but not in others.69 70 Exercise impairments are a common feature of post-COVID-19 sequelae.71–77 Poorer performance on the 6 min walk test (6MWT) was associated with fatigue and lower vitality at 6 months despite no concomitant impairments in pulmonary functions.78 Indeed, impairments in lung functions have not thus far fully explained worse fatigue in COVID-19.78–81 Nevertheless, patients often report persistent dyspnoea, which was consistently related to their fatigue,82–85 suggestive of multidimensional functional consequences. For instance, quality of life,86 functional status87 and an increased risk for postinfection healthcare needs88 were all related to fatigue.
Anxiety, post-traumatic stress and depressive symptoms are prevalent in survivors of respiratory viral infections.85 89–94 A meta-analysis of 36 COVID-19 articles found high rates of anxiety (29%) and depressive symptoms (23%) 4–12 weeks post illness.95 The relationship between mental health outcomes and fatigue is consistent among convalescing patients with COVID-19. Depressive symptoms, for example, were associated with lower vitality96 and fatigue.79 97 In a retrospective study of 55 patients, baseline anxiety was related to higher fatigue 30 days after hospitalisation.52 Moreover, these relationships can be present at 12 months’ follow-up. Mazza et al 98 found depression (r=0.56, q=0.05) and post-traumatic stress disorder (PTSD) (r=0.52, q=0.05) were related to fatigue severity in 402 post-COVID-19 patients. Neuropsychiatric symptoms, comprising anxiety, mood swings, irritability and depression and others, predicted chronic fatigue 9 months later for those with mild/moderate disease (p=0.01).99
Summary and aims
For the majority of patients, acute fatigue diminishes during the course of a virus, but current evidence suggests some experience longer lasting symptoms, and these affect functional and psychological recovery. Meta-analyses have focused on postacute sequelae of COVID-19 or clusters of symptoms, and therefore fewer studies have investigated solely fatigue outcomes. Moreover, a proportion of these reviews were narrative in design, which did not provide a pooled estimate for fatigue. Furthermore, fatigue is reported as the most prominent factor of postinfection symptomology indicative of its importance in understanding recovery. Therefore, the objectives of this systematic review were to (a) investigate the prevalence of persistent fatigue among survivors of COVID-19, (b) integrate the findings by conducting a meta-analysis and (c) investigate current evidence for factors associated with fatigue outcomes in this context.
Methods
Search strategy
The protocol and PICO framework for this study (online supplemental file 1) was developed using the Preferred Reporting Items for Systematic Review and Meta-Analysis (PRISMA).100 Embase, PsyINFO, Medline, CINAHL, Cochrane Database of Systematic Reviews, Open Grey, MedRxiv and BioRxiv were systematically searched from January 2019 to 31 December 2021. Search terms: severe acute respiratory syndrome or severe acute respiratory adj2 syndrome or coronavirus or corona virus or corona adj1 virus or COVID-19 or COVID-19 or SARS-CoV-2 or SARS-CoV or SARSCOV2 or SARSCOV-2 or nCoV-2 or 2019-nCoV or nCoV19 or nCoV2 or COVID-19 or COVID-19 or covid and ‘chronic fatigue’ or fatigue or tired* or exhaust* or quality adj2 life or QoL or health related quality) adj2 life or HRQoL. We incorporated ‘health-related quality of life’ into our search terms in order to capture ‘vitality’, which we used as proxy for fatigue. Reference lists of the review studies were manually searched for additional articles. Full search protocols for each database are available in online supplemental file 2. Duplicate references were removed electronically and imported into Rayyan101 for screening and inclusion decisions.
Supplemental material
Supplemental material
Inclusion and exclusion criteria
Included were original articles with primary data, published in English between January 2019 and December 2021. Adult patients (≥18 years) must have had a diagnosis of SARS-CoV-2 confirmed by RT-PCR, IgM/IgG serology or clinical assessment (eg, X-ray, CT scan of the chest). ‘Probable’ or self-reported cases were excluded. All study designs were incorporated except qualitative and case reports. Main outcomes were fatigue/vitality reported as ‘postdischarge’, ‘posthospitalisation’, ‘postacute’, ‘postillness’ or ‘postonset’. Outcomes were included if measured at a median/mean time of >30 days post infection as defined. All associations with fatigue/vitality were included if reported/quantified (eg, anxiety, dyspnoea). We excluded pandemic fatigue (defined as ‘worn out’ by pandemic warnings, government safety instructions, media coverage or compliance requirements), healthcare worker fatigue in the context of their work (eg, burnout, compassion fatigue), comorbid physical disease or pregnant populations. We excluded ‘muscle fatigue’, ‘leg fatigue’ and fatigue combined with ‘malaise’ or ‘muscle weakness’. Protocols, vaccination studies, newspaper articles, conference papers, commentaries, opinions or editorials were also omitted.
Data extraction
Titles and abstracts were screened by one reviewer (KP-W). Full texts were screened by KP-W. A data spreadsheet was created to record extracted data from the included studies. Spreadsheet variables were citation, population, sample size, control group, location, virus type and diagnostic method, follow-up period, study design, inclusion/exclusion criteria, objectives, outcome variable of interest (eg, fatigue, vitality), associated variables (eg, PTSD, dyspnoea), scales/measures employed, results, power calculation (Y/N). The senior researcher (TC) reviewed 10% of the final included studies. Discrepancies were resolved via discussion and consensus. A PRISMA flow diagram is available in figure 1.
Preferred Reporting Items for Systematic Review and Meta-Analysis 2020 flow diagram.
Quality assessments
Risk of bias was assessed by the JBI Critical Appraisal Tools.102 Items related to bias included ‘Were confounding factors identified?’, which demanded a ‘yes’, ‘no’, ‘unclear’ or ‘not applicable’. An overall assessment was made by assigning a grade of low quality, moderate quality or good quality. Three researchers (KP-W, OS and TC) independently graded 73%, 14% and 13% each of the total articles and, for the purposes of inter-rater estimation, researchers graded the same 10% of the articles. Inter-rater agreement was assessed by Fleiss’ kappa, which indicated moderate agreement (k=0.534, p=0.004).
Statistical analysis
We computed pooled mean prevalence for fatigue outcomes with 95% CIs using a random effects model as high heterogeneity was anticipated. A number of studies investigated fatigue across multiple time points. Therefore, in order to maintain the independence of observations for the pooled prevalence, we selected 1 time point with accompanying prevalence from each study using 1 of the 3 methods: (a) fatigue reported at the stated mean/median time of the follow-up assessment, for example, 127 days post illness, (b) fatigue at the 3-month follow-up (being the mode for all 147 studies) or (c) for studies investigating fatigue>4 months, we selected the shortest time point. Studies with missing data were excluded from analyses. Where studies investigated both ‘fatigue’ and chronic fatigue syndrome (CFS) outcomes, we incorporated the ‘fatigue’ data only. This was because a confirmed diagnosis of CFS could not be established. To determine the trend for fatigue, 35 prospective studies, with available data for >2 follow-up times, were included in a meta-regression using the mixed effects framework for meta-analyses developed by Sera et al.103 Meta-regression coefficients were estimated using a restricted maximum likelihood estimator. To determine the proportion of fatigued participants by study design, and to increase the power, we categorised studies into two categories: ‘cross-sectional’ and ‘prospective’. The latter included longitudinal and retrospective designs. The cross-sectional category comprised the remaining designs. Two categories were used to investigate proportions for ‘ongoing symptomatic COVID-19’ (1–3 months) and ‘post-COVID-19 syndrome’ (>3 months) following The National Institute for Health and Care Excellence guidelines (nice.org.uk). The robustness of the main pooled prevalence was checked by controlling for the presence of outliers. Studies with 95% CIs falling outside the 95% CI of the total pooled effect were defined as ‘outliers’. Sensitivity analysis was performed on the mean pooled prevalence by excluding high risk of bias studies and unpublished studies. To investigate the proportion of fatigued by scale, two categories were used: (a) studies with a valid fatigue scale and (b) studies without a valid fatigue scale. Meta-analyses were conducted using R Studio, V.1.3.1073104 using packages meta, metafor, dmetar, metareg, mixmeta and irr. Heterogeneity was assessed using Cochran Q statistic. We obtained the I2 statistic with the degree of heterogeneity categorised as ‘not important’ (0%–40%), ‘moderate’ (30%–60%), ‘substantial’ (50%–90%) and ‘considerable’ (75%–100%).105 We conducted Egger’s tests and produced funnel plots to explore potential publication bias for all proportional analyses. For ‘vitality’ outcomes, lack of comparable controls and missing data precluded a means difference analysis.
Patient and public involvement
No patient was involved in this study.
Results
Search results
A total of 14 262 articles were identified using the database search protocols. Following the removal of duplicates, 13 210 articles remained for title and abstract screening. Of these, a total of 3222 were selected for full-text screening producing a final total of 178 studies and 22 systematic reviews. We identified 147 as eligible for a quantitative analysis. A summary of the 147 included articles is available in online supplemental table 1. The studies are tabulated according to categorical and continuous fatigue outcome measures. Summary table of systematic reviews is available in online supplemental file 3.
Supplemental material
Supplemental material
Study characteristics
A total of 178 articles comprising 48 466 participants and 22 systematic reviews were included.13–16 91 95 106–121 A total of 14 (8%) were preprints, 30 (17%) used a fatigue scale and 27 (15%) used a validated measure with a fatigue item(s). A total of 13 (7%) used the ‘vitality’ subscale of the 36-Item Short Form Survey (SF-36) and 108 (61%) employed a questionnaire, interview or health records. The most common countries were Italy with 25 studies and the USA with 23 studies. The UK had 19 studies and China had 14 studies. Spain had 12 and France had 9 studies. Germany had eight studies and Switzerland had seven studies. The Netherlands and Turkey had six studies each and India had five studies. Iran had four studies. Bangladesh, Denmark, Egypt and Pakistan had three studies each. Brazil, Chile, Israel, Mexico, Norway and Sweden all had two studies. Austria, Australia, Belgium, Canada, Colombia, Finland, Ireland, Hungary, Japan, Lithuania, Mexico, Nepal, Poland, Russia, Saudi Arabia and Zambia each had one study. There were 80 prospective and 11 retrospective cohort deigns. Six longitudinal studies, 29 cross-sectional, 8 case–controls, 5 case series, 36 cohorts, 3 randomised-controlled trials and 22 systematic reviews. The most frequent follow-up times were 3 months (46 studies), 6 months (22 studies), 1 month (20 studies), 12 months (12 studies) and 2 months (12 studies). All other time points had <8 studies. JBI quality assessments resulted in most studies receiving a moderate rating. Full ratings are available in online supplemental file 4. In summary, 32 were assigned a ‘high’ risk of bias, 139 received a ‘moderate’ risk assessment and only 7 were considered ‘low’ risk. Lower grades were assigned for selection bias, lack of adequate control groups, small samples, study design and methodological bias (employment of unvalidated/unreliable scales).
Supplemental material
Meta-analyses
A total of 48 466 participants were included for the meta-analysis of proportions using a random effects model. A pooled prevalence from 147 studies was found to be 41% (95% CI: 37% to 45%, I2=98%). A forest plot of this analysis is available in figure 2. Fatigue was present between 1 month and 1 year post infection with a median time of 3 months (IQR=2–6). An Egger’s test was conducted to assess possible publication bias for our proportional analysis. The results indicated funnel plot asymmetry (bias=3.35, p=0.001) (online supplemental file 5).
Supplemental material
Forest plot for proportion of fatigued.
To explore potential origins of heterogeneity and to test the robustness of our pooled prevalence, outliers were controlled for. A 1% difference was found once n=84 outlier studies were removed; 42% (95% CI: 40% to 45%, I2=67%), although heterogeneity was reduced to ‘substantial’. Given the range of postinfection assessment periods, the effect of time on fatigue was investigated by a linear mixed effects model meta-regression. The outcome variable was the proportion of individuals reporting fatigue, with ‘months’ (number of months since infection) and ‘hospitalisation’ (whether someone was hospitalised) as predictors. A total of 35 studies with available fatigue data and multiple time points (≥2 follow-ups) were included. We found an effect of time, with the proportion of fatigued participants decreasing by 5.7% per month (95% CI: 1% to 10%, p=0.05). There was no effect of hospitalisation and no interaction between hospitalisation and time (table 1).
Results of linear mixed-effect meta-regression of time and hospitalisation
We conducted two subgroup analyses to explore the origins of heterogeneity arising from study methodology and investigate between group differences. No significant difference in fatigue was found between n=67 cross-sectional studies (44%, CI: 38% to 50%, I2=97.6%) and n=80 prospective studies (39%, CI: 33% to 45%, I2=98%), p=0.272.
A higher proportion of fatigued participants was found in n=36 studies using a scale (51%, 95% CI: 43% to 58%, I2=96.2%) compared with n=111 studies using an unvalidated questionnaire (38%, 95% CI: 33% to 43%, I2=98%), p=0.004. To assess fatigue occurring at (a) 1–3 months (‘ongoing symptomatic COVID-19’) and (b) >3 months (‘post-COVID-19 syndrome’), two random effects subgroup analyses were conducted. Between 1 and 3 months, the proportion of fatigued was 41% (95% CI: 36% to 47%, k=86, I2=98.3%). At >3 months, the proportion was 41% (95% CI: 34% to 48%, k=61, I2=97.4%). Sensitivity analysis was performed by excluding n=30 quality assessments (graded ‘low’) and removing unpublished results from the main analysis (n=8). Results found the pooled prevalence to be 40% (95% CI: 36% to 45%, I2=98.3%) and 41% (95% CI: 37% to 46%, k=139, I2=98%), respectively, indicating little impact on the main results. Egger’s tests indicated publication bias for both time categories and sensitivity. Plots are available in online supplemental files 6–15.
Supplemental material
Supplemental material
Supplemental material
Supplemental material
Supplemental material
Supplemental material
Supplemental material
Supplemental material
Supplemental material
Supplemental material
Factors associated with fatigue
Not all studies investigated or reported factors associated with fatigue. For some, the available data for each risk factor were too few to conduct a quantified analysis. Studies also used diverse outcome measures or non-validated scales. In addition, some risk factors were reported but not accompanied by quantified data making comparisons between studies problematic. Consequently, reported associations were arranged in tabular form illustrating the direction of the association with fatigue (table 2). A positive symbol (+) indicated a positive association, a negative symbol (−) indicated a negative association and a zero (0) indicated no significant association between the investigated variable and fatigue.122 The follow up time point in the prospective cohort designs is shown in brackets. Where a risk factor was examined with another (eg, ICU admission with age), one set of results was included. Full details of the associations are available in online supplemental file 16.
Supplemental material
Variables associated with fatigue
Non-modifiable factors
Older age was reported in 30 studies with mixed results. A total of 6 reported an association with or an increased likelihood of fatigue (OR=1.02) in participants>50.52 66–68 123 124 Two reported higher fatigue in >60 year olds125 and >40 year olds.83 Some, however, reported that younger age related to fatigue126–129 or no difference in fatigue severity between <65 and >65 year olds.130 The remaining 17 studies did not find a relationship to fatigue.69 70 79 80 84 85 96 98 99 131–138 However, studies reporting non-significant results had small-to-modest sample sizes and were therefore potentially underpowered. Gender was investigated by 46 studies. A total of 30 studies reported a significant association with fatigue. More women were fatigued, 42 52 63–66 68 96 98 99 123 125 128 130 133 136 139–152 they (54.3%) reported more severe/moderate fatigue than men (29.6%)86 129 and had significantly lower vitality scores (M=81.80) compared with men (M=83.25).124 However, 16 used an unvalidated instrument potentially affecting results. Those finding no significant difference70 79 80 83 84 131 132 135 137 138 153 154 had small sample sizes and only three used a fatigue scale.
Physical factors
The key physical factors associated with fatigue were dyspnoea, pulmonary functions, exercise capacity, comorbidities and ICU admission. An association between breathlessness and fatigue was found in three studies79 84 85 and those with fatigue had a higher prevalence of breathlessness in four other studies.82 83 129 155 At 3–6 months post infection, two did not find a relationship,80 96 suggestive of improvements over time. Staudt et al (2021) found that ‘respiratory symptoms’ on the St George’s Respiratory Questionnaire (SGRQ) were related to fatigue in multivariate analyses at 10 months post infection (OR=1.06, p=0.05). However, only two used a dyspnoea scale or a fatigue scale. All had small sample sizes, therefore potentially underpowered. Pulmonary functions were reported in five studies. Forced expiratory volume in one second (FEV1) related to higher vitality in 1 (r=0.0.23, p<0.05),78 but non-significant in the others.79 80 155 These studies assessed survivors≥3 months, suggesting results are indicative of functional improvements overtime. Exercise capacity was generally poor in survivors156 and seven studies examined its relationship with fatigue, with mixed results. Better exercise performance was associated with vitality (r=0.526, p<0.001),78 but not with 4 m gait speed test85 or 6MWT.79 Two others found improved fatigue following a physical rehabilitation programme.97 157 At 3 months post infection, fatigue was cited as the reason for halting a cardiopulmonary performance test or limiting exercise in three studies.158–160 Myopathy was associated with fatigue in another small study of 20 people161 suggestive of poor conditioning contributing to limited capacity. Generally, fatigue had an inverse relationship with exercise capacity in the early months. Where the relationship remained beyond 3 months,78 patients were overweight/obese, which possibly affected performance. Also, all studies had small sample sizes limiting generalisability.
Physical comorbidities such as hypertension, asthma and diabetes were related to fatigue in nine studies.52 63 68 126 128 136 146 148 162 Four found no relationship.132 133 137 147 A large study of 4755 participants found hypertension increased the likelihood (OR=1.27, p=0.05) of persistent fatigue>6 months.148 Yomogida et al 68 reported that having at least one comorbidity increased the risk for fatigue (OR=4.39, p<0.001). Moreover, worse physical health was related to fatigue (OR=10.48)65 163 164 implying general poorer functioning among survivors.165
For those admitted to ICU, some experienced high fatigue (eight studies),83 129 131 and lower vitality,166 167 or had an increased likelihood for fatigue (OR=4.63).52 128 168 Four studies found no association between ICU admission and worse fatigue or vitality.42 169–171 Patients who received mechanical ventilation had lower vitality (M=50, 95% CI: 44 to 57) than a sex-matched and age-matched group (M=68, 95% CI: 67 to 69).172 Similarly, more intubated patients had fatigue (38.1%) than non-intubated (29.9%).173 One study found the proportion of fatigued participants was higher in the ward group (74%) compared with ICU (33%).143 Disease severity also had an inconsistent impact on fatigue, with most studies finding no association with severe acute disease or fatigue prevalence in severity categories.80 86 93 130 136 137 153 174–180 Six studies found a significant association with critical illness or a significantly higher proportion of fatigued in severe illness.123 135 145 181–183 Two studies found a relationship between severity of acute illness and vitality,184 185 although both had small samples and were single-centre designs. Interestingly, moderately severe COVID-19 related to fatigue (OR=2.1) in one study.186 Even after a longer hospital stay, the relationship with fatigue was inconsistent with two finding significance,52 124 while four did not.69 98 137 149 Taken together, these results indicate an uncertain contribution of critical illness to fatigue, although the non-significant results chiefly occurred>6 months. However, the classification of disease severity varied between studies and countries making comparisons difficult.
Psychological factors
A relationship with anxiety was found up to 6 months post infection in three studies.52 83 149 The fatigued had higher anxiety (56.3%) compared with non-fatigued (24.6%, p<0.001).83 149 In contrast, no significant interaction between anxiety and fatigue at 1 month related to later fatigue.187 Similar results were found for depression. Previous depression was associated with lower vitality (−12.05, p=0.005) in one study96 and a higher proportion of fatigued had depressive symptoms in four other studies (p=0.004).83 90 155 188 Other studies found consistently moderate positive correlations (r=0.470).98 171 189 or increased fatigue scores (b=0.89, p=0.05) in those with depressive symptoms.52 The relationship continued up until 12 months.79 98 Four studies found that those with PTSD symptoms were fatigued90 129 and PTSD was associated with fatigue at 6 and 12 months after infection.98 Barizien et al 132 found higher scores on the PTSD Checklist for Diagnostic and Statistical Manual of Mental Disorders (DSM-5) in those with fatigue (M=31, IQR=18) compared with those without fatigue (M=18, IQR=19, p<0.001). Generalisability of these results, however, is likely limited due to modest sample sizes and single-centre designs. In addition, only three studies used a valid fatigue scale.
Discussion
This review investigated the prevalence of persistent fatigue in survivors who had a confirmed diagnosis of SARS-CoV-2, using a mean of >30 days post infection. We found a considerable proportion of patients continued to experience fatigue up to 12 months after their initial illness, which was associated with some non-modifiable factors including female gender, age and modifiable factors such as anxiety, depression and post-traumatic stress. Our findings support other research indicating that fatigue is an important symptom in persistent postacute sequelae.14 112 150 190–196 Rates of fatigue may depend on when it was measured and, in this respect, we found overall rates of fatigue decreased by 6% per month. Fatigue did not differ by hospitalisation status, indicating that the contribution of severe disease was not related to fatigue recovery for most people. This is consistent with previous reviews, which did not find support for the effects of critical illness on fatigue outcomes.117 197 Respiratory impairments, a key clinical indicator, were associated with worse vitality post recovery (r=0.290, p=0.026),78 although at 10 months, FEV1 was not associated79 implying that, as lung function improved, fatigue diminished. Indeed, rehabilitation aimed at improving functioning by incorporating aerobic exercises, improved vitality scores.97 167 198 Some survivors, however, continued to experience dyspnoea, which was associated with their fatigue,83–85 despite normal pulmonary tests.80 159 Similarly, reduced exercise capacity, as a result of critical illness, is thought to contribute to reduced HRQoL and fatigue outcomes in recovered patients.199 However, our review did not find a consistent relationship between exercise performance and worse fatigue in those who had more severe disease. It is possible that these limitations are related to diminished muscle function199 and deconditioning. Rehabilitation programmes have led to improved vitality157 198 and lower fatigue.97 157 A 9-week telerehabilitation study of 115 participants, incorporating 2/3 aerobic exercises per week to improve physical capacity, reported significantly increased vitality scores from pre=40.7(SD=21.7) to post=58.5(SD=21.2), p=0.001.167 While deconditioning could explain fatigue, persistent fatigue may be related to other variables including psychological factors.
Depression and anxiety were found to be correlated with fatigue in our review.52 171 Moreover, these relationships were found some distance from the initial infection.98 155 In a prospective study of 402 participants using a fatigue scale, Mazza et al found that both anxiety (r=0.48) and PTSD (r=0.52) were moderately correlated with fatigue at 12 months, post illness. These findings accord with critical illness studies200 and systematic reviews suggesting that symptoms of depression, anxiety, PTSD and fatigue persist long after discharge.197 For COVID-19, we cannot be certain of the longevity of psychological factors or their relationship to fatigue because the body of evidence is too small, but current literature indicates the relationship remains up to 6 months.83 132 This fits with previous COVID-19 research indicating those with chronic fatigue were more likely to have psychiatric morbidity 4 years following SARS infection.201 Similarly, those with psychiatric illness reported higher fatigue than those without (p<0.05) in survivors of SARS.202
Theoretical implications
The associations of fatigue persistence were multidimensional. Factors such as dyspnoea and comorbidities (eg, hypertension) were likely risk factors for fatigue in the shorter term, whereas psychological factors appeared more likely to be associated with fatigue longer term. The psychological risk factors could have been related to adverse effects of the pandemic as well as infection.203 204 Taken together, these factors, alongside other mechanisms such as skeletal muscle deficits,205 could lead to poorer global functioning and lower engagement in activities or exercise. Lower scores on objective walking tests and reduced physical functioning were associated with fatigue in some studies. We have summarised diagrammatically the factors associated with post-COVID-19 fatigue (see figure 3).
Diagram of post-COVID-19 fatigue findings. ICU, intensive care unit.
Practical implications
Our review suggests post-COVID-19 fatigue is complex, affecting multiple domains of physical and psychological well-being. While there were small improvements in fatigue over time, our review indicates that fatigue remains a significant problem for patients beyond their anticipated recovery time.206 Pulmonary rehabilitation programmes have shown promise.97 167 198 Our results also suggest that psychological interventions may benefit some survivors. Given fatigue is one of a number of post-COVID-19 symptoms,207–210 an integrated management approach has been suggested.211 Care pathways should identify those most at risk for long-term symptoms such as women and older people with comorbidities.
Future directions
Few studies have examined correlates between fatigue, physical and pulmonary functioning, psychological and social functioning in hospitalised and outpatients. Some research focuses on symptom ‘clusters’ or ‘post-COVID-19 syndrome’212–215 limiting understanding of fatigue processes specifically. Future studies should interrogate risk factors further to help inform the development of clinical interventions to address persistent fatigue. Furthermore, fatigue is the principal symptom for postillness patients, but there is little research into what mechanisms may ameliorate distress resulting from infection, and thus protect against long symptoms. Severity of the illness, for instance, was not conclusive in our study and nor was length of hospital stay, pointing to the importance of individual differences.
Limitations
The generalisability of our results should be applied with caution due to a number of limitations. First, we found considerable, unexplained between-study heterogeneity. Measurement error was not found to explain the inconsistency. However, diverse tools were used to measure fatigue in different populations. Non-validated questionnaires were unlikely to capture fatigue dimensions accurately given most only had 1–2 fatigue-related items. Moreover, scoring and cut-offs were under-reported, contributing to variability. Included studies could not adequately exclude ‘pandemic-related fatigue’ in their selections or definitions. Therefore, we recognise that our results could not completely exclude such fatigue and its potential influence on participants in the included studies. Some studies used particular populations, including older age or only those admitted to ICU, meaning they were not representative. Furthermore, our sample comprised primarily of hospitalised patients with potentially more severe disease. This was complicated by different admission and discharge protocols across countries, with some admitting all confirmed patients regardless of disease severity. This could explain why there was no difference between hospitalised and non-hospitalised survivors. We also encountered missing data, which reduced the reliability of our results. Moreover, Egger’s tests suggested all but one analyses were asymmetric representing a high likelihood of publication bias. Small study effects were likely to affect precision. Larger studies, with more precise CIs, are likely to be a more reliable indicator of fatigue proportions. Moreover, sample bias probably occurred due to recruitment from single-centre post-COVID-19 clinics216–218 for persistent symptoms and therefore could be expected to have higher fatigue than controls or population norms. Different admission and discharge protocols and lung function reference ranges vary between countries.219 Our results, therefore, should be viewed with this in mind. Methodologically, our study had only one reviewer for screening and data extraction, and we did not contact authors for missing data meaning our study was at higher risk for excluding relevant data. Other limitations include the inclusion of non-peer reviewed articles and those limited to English. For the meta-analysis, given the multiple assessment times, we incorporated one median follow-up time obtained from each study, which may not denote actual fatigue prevalence. Despite these limitations, we incorporated a substantial sample size likely to be a reasonable estimate of fatigue in this population.
Conclusion
This large review provides a broad illustration of fatigue outcomes and complements the body of information for persistent symptoms in those recovering from COVID-19. We report that fatigue decreases over time, but recovery pathways are potentially impeded by a number of risk factors, independent of disease severity or hospitalisation. Our study indicates the need for long-term clinical and psychological rehabilitation support for survivors of COVID-19.
Data availability statement
Data are available upon reasonable request. Data are available on request from the corresponding author.
Ethics statements
Patient consent for publication
Ethics approval
Not applicable.
Acknowledgments
The authors thank Carolina Carvalho for her contribution to the quality assessment analysis.
References
Supplementary materials
Supplementary Data
This web only file has been produced by the BMJ Publishing Group from an electronic file supplied by the author(s) and has not been edited for content.
- Data supplement 1
- Data supplement 2
- Data supplement 3
- Data supplement 4
- Data supplement 5
- Data supplement 6
- Data supplement 7
- Data supplement 8
- Data supplement 9
- Data supplement 10
- Data supplement 11
- Data supplement 12
- Data supplement 13
- Data supplement 14
- Data supplement 15
- Data supplement 16
- Data supplement 17
Footnotes
Contributors KP-W contributed to the study design, data collection, data analysis and draft manuscript preparation. IG contributed to the design, data analysis and manuscript review. OS contributed to the data analysis, quality assessments and manuscript. RAE contributed to the study design and manuscript review. FG contributed to the study design and manuscript review. TC contributed to the study design, manuscript and supervision, and is the guarantor of this work.
Funding The authors have not declared a specific grant for this research from any funding agency in the public, commercial or not-for-profit sectors.
Competing interests FG has received support or honoraria from, Lundbeck, Otsuka and Sunovion and has a family member with previous professional links to Lilly and GSK. FG is in part supported by the National Institute for Health Research’s (NIHR’s) Biomedical Research Centre at South London and Maudsley NHS Foundation Trust and King’s College London, the Maudsley Charity and the NIHR Applied Research Collaboration South London at King’s College Hospital NHS Foundation Trust. The views expressed are those of the author(s) and not necessarily those of the NIHR or the department of health and social care. RAE has received support or honoria from Boeringher Ingelheim and is a member of the ERS Group 01.02 Pulmonary Rehabilitation. TC is the author of several self-help books on chronic fatigue for which she has received royalties. TC (KCL) has received ad hoc payments for workshops carried out in long-term conditions. TC is on the Expert Advisory Panel for Covid-19 Rapid Guidelines. She is also in receipt of grants from NIHR and St Thomas’ Charity. TC collaborates with The Post-hospitalisation Covid-19 Study. TC is the Director of the Persistent Physical Symptoms Service. There are no other relationships or activities that could have influenced the submitted work.
Patient and public involvement Patients and/or the public were not involved in the design, or conduct, or reporting, or dissemination plans of this research.
Provenance and peer review Not commissioned; externally peer reviewed.
Supplemental material This content has been supplied by the author(s). It has not been vetted by BMJ Publishing Group Limited (BMJ) and may not have been peer-reviewed. Any opinions or recommendations discussed are solely those of the author(s) and are not endorsed by BMJ. BMJ disclaims all liability and responsibility arising from any reliance placed on the content. Where the content includes any translated material, BMJ does not warrant the accuracy and reliability of the translations (including but not limited to local regulations, clinical guidelines, terminology, drug names and drug dosages), and is not responsible for any error and/or omissions arising from translation and adaptation or otherwise.