Article Text

How long is Long-COVID? Symptomatic improvement between 12 and 18 months in a prospective cohort study
  1. Robert M Barker-Davies1,2,
  2. O O’Sullivan1,3,
  3. D A Holdsworth4,5,
  4. P Ladlow1,6,
  5. A Houston1,
  6. R Chamley4,5,
  7. A Greenhalgh1,
  8. E D Nicol7,8 and
  9. A N Bennett1,9
  1. 1Academic Department of Military Rehabilitation, Defence Medical Rehabilitation Centre, Loughborough, UK
  2. 2School of Sport Exercise and Health Sciences, Loughborough University, Loughborough, UK
  3. 3Academic Unit of Injury, Recovery and Inflammation Sciences, University of Nottingham, Nottingham, UK
  4. 4Academic Department of Military Medicine, Royal Centre for Defence Medicine, Birmingham, UK
  5. 5Oxford Heart Centre, Oxford University Hospitals NHS Foundation Trust, Oxford, UK
  6. 6Department for Health, University of Bath, Bath, UK
  7. 7Department of Cardiology, Royal Brompton Hospital, Birmingham, UK
  8. 8School of Biomedical Engineering and Imaging Sciences, King's College London, London, UK
  9. 9National Heart and Lung Institute, Imperial College London, London, UK
  1. Correspondence to Group Captain A N Bennett, Academic Department of Military Rehabilitation (ADMR), Defence Medical Rehabilitation Centre, Loughborough, LE12 5BR, UK; alexander.n.bennett{at}


Introduction COVID-19 infection can precede, in a proportion of patients, a prolonged syndrome including fatigue, exercise intolerance, mood and cognitive problems. This study aimed to describe the profile of fatigue-related, exercise-related, mood-related and cognitive-related outcomes in a COVID-19-exposed group compared with controls.

Methods 113 serving UK Armed Forces participants were followed up at 5, 12 (n=88) and 18 months (n=70) following COVID-19. At 18 months, 56 were in the COVID-19-exposed group with 14 matched controls. Exposed participants included hospitalised (n=25) and community (n=31) managed participants. 43 described at least one of the six most frequent symptoms at 5 months: fatigue, shortness of breath, chest pain, joint pain, exercise intolerance and anosmia. Participants completed a symptom checklist, patient-reported outcome measures (PROMs), the National Institute for Health cognitive battery and a 6-minute walk test (6MWT). PROMs included the Fatigue Assessment Scale (FAS), Generalised Anxiety Disorder-7 (GAD-7), Patient Health Questionnaire-9 (PHQ-9) and Patient Checklist-5 (PCL-5) for post-traumatic stress.

Results At 5 and 12 months, exposed participants presented with higher PHQ-9, PCL-5 and FAS scores than controls (ES (effect size) ≥0.25, p≤0.04). By 12 months, GAD-7 was not significantly different to controls (ES <0.13, p=0.292). Remaining PROMs lost significant difference by 18 months (ES ≤0.11, p≥0.28). No significant differences in the cognitive scales were observed at any time point (F=1.96, p=0.167). At 5 and 12 months, exposed participants recorded significantly lower distances on the 6MWT (ηp2≥0.126, p<0.01). 6MWT distance lost significant difference by 18 months (ηp2<0.039, p>0.15).

Conclusions This prospective cohort-controlled study observed adverse outcomes in depression, post-traumatic stress, fatigue and submaximal exercise performance up to 12 months but improved by 18-month follow-up, in participants exposed to COVID-19 compared with a matched control group.

  • COVID-19

Data availability statement

Data are available upon reasonable request.

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:

Statistics from

Request Permissions

If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.


  • A prolonged syndrome including fatigue, exercise intolerance, mood and cognitive problems, termed Long-COVID, has been recognised to follow COVID-19 infection. This prospective cohort study sought to describe the profile of Long-COVID in a young, physically active working population, the majority of whom were managed in the community in contrast to literature on hospitalised cohorts with prevalent comorbidities.


  • Adverse outcomes in depression, post-traumatic stress, fatigue and submaximal exercise performance were observed for up to 12 months but improved by 18-month follow-up in participants exposed to COVID-19 compared with a matched control group.


  • Employers should be mindful of the necessary timelines during which clinical rehabilitation advice should be followed to avoid potential symptom exacerbation in those experiencing typical Long-COVID symptoms.


COVID-19 infection has been associated with a syndrome of prolonged symptoms typically characterised by fatigue, exercise intolerance, mood disturbance and cognitive problems termed Long-COVID (LC).1–3 LC can be described as an umbrella term with symptom clusters identified affecting multiple systems.4–6 LC has implications for quality of life and employment, particularly for those in physically demanding occupations such as military service.

A recent meta-analysis demonstrated pooled prevalence of fatigue, depression, anxiety, memory loss and joint pain around 20% at 1-year follow-up post-COVID-19.7 Some surveys have reported around 30% of workers identifying at least mild restrictions or inability to return to previous work between 3 and 12 months post-COVID-19.8 9 Many studies have been conducted on hospitalised populations where high prevalence of new disability was anticipated and observed.5 7 10 However, most patients contracting COVID-19 have been manged in the community. A UK community survey of over half a million participants found 38% of those exposed to COVID-19 declared at least one LC symptom at 12 weeks.6 Thirty-six per cent of a cohort of non-hospitalised post-COVID-19 patients presenting to a UK community rehabilitation service reported reduced hours or not returning to work with symptoms at a median of 223 days’ duration.11 A large general population cohort study demonstrated incomplete recovery in 42% of those managed in the community 18 months after symptomatic COVID-19.12 While those enduring the most severe acute illness are likely to experience sequelae, patients who experience a milder acute illness should also remain vigilant for LC symptoms requiring further rehabilitation.3 13

Return to work (RTW) requires special consideration. Physical as well as mental activity and stress have been identified as triggers for LC symptom deterioration.2 Symptom clusters identified under the LC umbrella include post-exertional symptom exacerbation (PESE), dysfunctional breathing pattern and dysautonomia.4 14 15 For these conditions, accelerated RTW has potential for further morbidity. Guidelines have advocated a symptom-guided approach,4 13 the rationale of which being that a lack of pacing can drive symptom deterioration. Conversely, titration of activity to exercise response has been shown to reduce PESE in a population with established LC.16 A longitudinal cohort study has identified the ability of employees with LC to undertake modified duties as a significant positive predictor of RTW.17 Overall, caution has been advocated against accelerated RTW.18 Given the relatively novel status of COVID-19, medium to long-term outcomes are an active area of research, with the hope of guiding and prognosticating outcomes. This will aid occupational planning.

This study tested the hypothesis that continued improvement in the symptomatology, exercise capacity and cognitive performance of those exposed to COVID-19 would result in no difference to matched controls 18 months post-exposure to COVID-19.


Study design

The study design was a prospective observational cohort study.


Assessments were conducted at the Defence Medical Rehabilitation Centre (DMRC), Stanford Hall at each time point for assessment.


Participants were recruited to the Military COVID-19 Observational Outcome in a Viral Infectious Disease Study as previously described.19–21 This paper represents the final follow-up of that cohort 18 months after acute infection. This report has been compiled in accordance with Strengthening the Reporting for Observational Studies in Epidemiology (STROBE) guidelines. The STROBE diagram for the study is represented in figure 1. Of 113 participants initially recruited at 5 months,20 post-initial illness 88 attended for 12-month follow-up and 70 at 18 months. The most frequent reason for dropout was participants not engaging with the research team following discharge from routine clinical follow-up from the rehabilitation centre (n=14) (figure 1).

Figure 1

Strengthening the Reporting for Observational Studies in Epidemiology diagram from participant identification to 18-month follow-up.

Participants who had been exposed to COVID-19 were recruited from a clinical pathway for hospitalised and community-managed symptomatic patients.22 The pathway was designed to offer a clinical service to deliver occupational guidance and clinical rehabilitation for those with ongoing post-COVID-19 symptoms. Symptom status was determined from a checklist based on early post-COVID-19 to data to identify the 10 most common symptoms according to a large smartphone app-based symptom tracking study.23 In addition, a further 26 symptoms were identified from the Stanford Hall Consensus Statement for post-COVID-19 rehabilitation.13 The full checklist is in online supplemental file 1. Controls were recruited from local military units. Controls were matched for age, sex, height and job role. Controls who developed COVID-19 during the study period were removed from subsequent analysis (figure 1). Eligibility for recruitment was assessed, and agreed, by two senior consultants based on positive COVID-19 antigen testing, history and imaging. It is summarised in online supplemental file 2. Written informed consent was obtained from all participants. At the point of study design in April 2020, no evidence was available on COVID-19 to inform sample size by power calculation. A convenience sample of around 50–100 exposed participants was sought to compare with a smaller group of 20–30 controls.

Outcome measures

At each time point, participants completed the symptom checklist (online supplemental file 1) and patient-reported outcome measures (PROMs) via surveys using the research and electronic data capture (REDCap) platform.24 The National Institute for Health (NIH) cognitive assessment battery and a 6-minute walk test (6MWT) were also completed as described below.

Participants completed PROMs relating to depression (Patient Health Questionnaire-9, PHQ-9); anxiety (General Anxiety Disorder scale-7 questions, GAD-7); post-traumatic stress disorder (PTSD, National Centre for PTSD checklist, PCL-5); quality of life (QoL, European QoL 5 domains, EQ5D), Alcohol Use Disorders Identification Test (AUDIT) and fatigue (Fatigue Assessment Scale, FAS). The Functional Activity Assessment (FAA) which is validated as a self-report scale for service personnel to describe RTW status was also completed.25

Cognitive assessments were performed using the NIH cognitive toolbox cognition battery for age 12+ years on an iPad (Apple, California, USA) supervised by a doctor, with the fluid, crystallised and total composite scores analysed.

The 6MWTs were performed by exercise rehabilitation instructors according to the American Thoracic Society guidelines, including the body composition data recorded in table 1.

Table 1

Patient demographics

Statistical analysis

To address the research question of the current study and avoid potential errors associated with small subgroup numbers, statistical analysis was pooled across groups (as per online supplemental file 2) as follows. The COVID-19-exposed group were compared with controls (table 1). Within the COVID-19-exposed group, further comparisons were made between hospitalised and community-managed participants, and recovered and symptomatic participants (table 2). Symptomatic participants were defined as any participants who self-reported with any of the six most prevalent chronic symptoms identified at 5 months (anosmia, fatigue, shortness of breath (SoB), chest pain, joint pain and exercise intolerance).

Table 2

Patient demographics

At 5 months after acute illness, differences were identified between COVID-19 and control groups. PROMs data were non-normally distributed so non-parametric tests were applied. Only variables for which a medium effect size (ES) (r>0.3) was observed were carried forward to analysis at 12 and 18 months from acute illness.26 Analyses at 5 months found all PROMs except the AUDIT and EQ5D scores surpassed the threshold to be investigated further. Contrary to Ladlow et al,19 GAD-7 and PCL-5 were found to have significant differences at 5 months,19 the groups compared were not identical, as a result of participant dropouts from the patient groups seen all the way to 18-month follow-up. As per previous findings in this cohort, an association between illness and raised body mass, body mass index (BMI) and waist circumference was expected.20 Waist circumference was therefore controlled for using analysis of covariance in analysis of the 6MWT. Composite t-scores from the NIH cognitive toolbox tests were normally distributed, and were analysed using two-way analysis of variance (group x time). Change in the ordinal FAA scale over time was analysed using the Freidman test. Analysis of the data was per protocol, supervised use of the REDCap system24 during inpatient visits for questionnaire completion resulted in complete datasets for all participants. Multiple imputation was not used for the participants lost to follow-up as per figure 1. All statistical calculations were made using SPSS, V.27 (IBM).

Public and patient involvement was performed prior to ethical approval and study recruitment with a series of 4 focus groups of 8–12 patients attending Defence Medical Rehabilitation Centre Stanford Hall for residential COVID-19 rehabilitation. These focus groups tested the face validity of participant information and proposed testing protocols. As a result of this feedback, participant information provided at the point of recruitment was simplified and included greater use of illustrations.


Demographics and cohort subgroupings are detailed in tables 1 and 2, respectively. Age, sex and height were matched, but participants exposed to COVID-19 had significantly greater body mass, BMI, waist and hip circumference at all time points to controls (p≤0.01) (table 1). Within the COVID-19-exposed group, hospitalised patients had significantly greater BMI and waist circumference at all time points (p≤0.01) (table 2). Symptomatic participants had significantly greater waist and hip circumference than those recovered at 18 months (p≤0.02) (table 2).

Overall, symptom prevalence reported by checklist declined over the observation period. Specifically related to the participants who remained symptomatic at the 5-month time point, prevalence of all symptoms at the 18-month follow-up was consistent with the control group, barring fatigue and SoB (reported by 35% and 14%, respectively) (online supplemental file 3). Regarding hospitalised patients, the prevalence of fatigue remained close to 50% over the period of observation. Prevalence of SoB (reported by 68% at 5 months) remained high (20%) for a pre-morbidly physically active population, at 18 months post-acute illness. For the hospitalised group, the remaining symptom prevalence was similar to the control group at the 18-month follow-up (online supplemental file 3).

At 5 and 12 months, exposed participants presented with higher PHQ-9, PCL-5 and FAS scores than controls (ES ≥0.25, p≤0.04) (figure 2B–D). By 12 months, GAD-7 had improved so that there was no difference to controls at or beyond this time point (ES <0.13, p=0.29) (figure 2A). By 18 months, PHQ-9, PCL-5 and FAS showed improvement, with no difference to controls (ES ≤0.11, p≥0.28). Underlying data and statistical values for figure 2 are in online supplemental file 4.

Figure 2

Patient-reported outcome measures and 6-minute walk test (6MWT) distance up to 18 months post-COVID-19 exposure compared with controls. (A) Generalised Anxiety Disorder-7 (GAD-7), (B) Patient Health Questionnaire-9 (PHQ-9), (C) Patient Checklist-5 (PCL-5) for post-traumatic stress, (D) Fatigue Assessment Scale (FAS) and (E) 6MWT. *Significantly poorer score for those exposed to COVID-19 compared with controls.

At 5 and 12 months, exposed participants recorded significantly lower distances (ηp2≥0.126, p<0.01), which by 18 months was not significantly less than controls (ηp2<0.039, p>0.15) (table 3). Although the magnitude of the deficit reduces over the course of 18 months to non-significant levels, clinically meaningful mean differences persist. The difference in point estimates is >70 m compared with a minimally clinically important difference of 14–30.5 m calculated by receiver operator curve analysis from systematic review.27 To elicit reasons for this, a subgroup analysis was conducted to understand whether acute severity or recovery status may factor into this difference. With regard to recovery status, a medium ES (ηp2≥0.06) was observed between groups (ηp2=0.093), suggesting that sustained illness may be a greater factor than acute severity in terms of long-term functional deficits.

Table 3

Six-minute walk test distance in metres with subgroup analysis

There was no significant difference in cognition (fluid composite minus crystallised composite scores) between groups (F=1.96, p=0.167) and no significant group×time interaction (F=3.18, p=0.08).

Prior to COVID-19, 59% of those subsequently exposed to COVID-19 were fully deployable with 18% not deployable. By 18 months post-COVID-19, 49% were deployable in some capacity (29% fully and 20% limited) and 51% remained not deployable (figure 3). At 5 months post-COVID-19, 54% were fit for their trade (FAA 1–2). There was a significant main effect for improvement in FAA over time (Χ2=17.7, p<0.001) with Dunn’s multiple comparison test indicating significant improvement at 18 months (p<0.01) but not 12 months (p=0.45) compared with 5 months post-COVID-19. By 18 months, 88% were fit for their trade (FAA 1–2) (figure 3).

Figure 3

Occupational status of participants following COVID-19. Top row: deployability as per formal occupational grading. Bottom row: Functional Activity Assessment (FAA) for which there was significant improvement over time (Χ2=17.7, p<0.001) with Dunn’s multiple comparison test indicating significant improvement 18 months (p<0.01) but not 12 months (p=0.45) compared with 5 months post-COVID-19.


The main finding of this study is improvement in classic LC symptoms with no statistical differences persisting compared with age, sex, height and job-matched controls 18 months post-exposure. The findings follow a trend in improvement within the existing cohort study. In line with previous research, an association was found between markers of obesity and COVID-19 hospitalisation19 20 as well as incomplete recovery.5 6 In the current study, obesity was controlled for in the analysis of submaximal exercise performance. Further planned subgroup analysis sought to determine whether acute severity, as defined by hospital admission, or established prolonged symptomatology at the time of recruitment (5 months post-exposure) had a bearing on long-term outcomes. The current study would suggest that established (LC) symptoms at 5 months may have more bearing on exercise capacity than acute severity, though both of those groups declare similar symptom prevalence throughout. Previous analysis of this cohort at an earlier time point demonstrated that those managed in the community and who feel recovered have no demonstrable exercise performance or symptomatic deficits19 20 which was maintained at 18 months.

Clinically, these findings have two key implications. First, in a physically active population, such as these military participants, a subjective feeling of recovery, following community-managed COVID-19 infection, is sufficient to sanction the return to high levels of physical exercise and arduous employment. Second, those with continued symptoms may require a significantly longer period of rehabilitation. While this should be guided by symptomatic response to exercise, the current study suggests an 18-month period may be required for those with established LC. These findings are commensurate with large post-COVID-19 cohort studies which also demonstrate minimal symptomatic improvements between 6 and 12 months.5 9 Though in one of those studies, in agreement with the current study, there was subsequent significant improvement by 2 years.1

Employers therefore should be mindful of these timelines during which period workers may be at risk of complications. For example, we identified dysautonomia to be prevalent among our cohort.14 Typical military activities such as prolonged standing, work in warm environments and exposure to dehydration can provoke orthostatic intolerance.15 PESE has also been associated with more severe fatigue in an LC cohort.28 Further, the cohort described with PESE reported reduced working ability.28 It is recommended therefore, particularly for those in physically demanding occupations, that persistent symptoms are reviewed by a medical officer, mindful of the potential differential diagnosis and with access to specialist rehabilitation services.18 Fatigue in the current study, as mirrored across the literature, was the most prevalent symptom.7 Fatigue requires careful management with emerging evidence supporting pacing and symptom-guided management.16 A balance needs to be struck between exacerbation of the fatigue condition and deconditioning that occurs among other negative health consequences of physical inactivity.29

This cohort represents a population for whom physical fitness is a requirement of their occupation. Around half of this cohort remained not deployable at 18 months compared with 18% pre-COVID-19. This finding may reflect a lag in the administrative process for upgrading which is frequently revisited 6 monthly in the UK Armed Services. As a result, the more responsive FAA was used, which encouragingly showed 88% of personnel to declare themselves fit for their primary role in this post-COVID-19 cohort by 18 months. These findings have generalisability to young physically active cohorts with minimal comorbidities. This contrasts with much of the current post-COVID-19 literature based on hospitalised cohorts.5 7 10

The strengths of this work are the inclusion of both community as well as hospitalised patients, use of a young working aged cohort, recruitment of participants with proven, normal pre-morbid fitness, prospective design, including prospective recruitment of a matched control group and in-person visits at multiple time points. LC has a broad definition with an extensive list of possible symptoms which could be attributed to alternate conditions.30 We sought to narrow this definition and used an iterative process to select the most common LC symptoms to define those not recovered from COVID-19 at 5 months. Further, we excluded participants with a history of cardiac or respiratory conditions from the study.

The small sample size and dropout rate are weaknesses. Small sample size was a result of a pragmatically designed study, in a novel disease with no evidence to guide sample size at the point of study design, embedded within a new clinical rehabilitation service designed to provide timely insight for occupational guidance. The dropout rate is likely to underestimate, rather than overestimate, improvements in LC symptoms and function. Reasoning for this is that none of the patients lost to follow-up remained under prolonged care of DMRC. Further, engagement in military activities including overseas deployment was anecdotally noted to hamper efforts to arrange follow-up visits. While this was at cost to the study protocol, the overall benefit to Defence serves Defence Rehabilitation’s mission statement to ‘force-generate’ personnel fit to deploy on operations, and demonstrates full recovery from COVID-19 and RTW. The control group lost 10 participants who subsequently contracted COVID-19. At the point of study design in April 2020, this could not have been predicted. The study findings do rely on this small group and further studies with larger numbers are required.

To conclude, this study demonstrates improvement of typical LC symptoms occurring between 12 and 18 months post-illness using a prospective cohort-controlled method. Anxiety, depression, fatigue, post-traumatic symptoms and exercise performance, which were significantly worse at 5-month follow-up compared with a matched non-infected control group, lost significant differences by 18 months. For patients exposed to COVID-19 with persistent symptoms, occupationally focused medical review is recommended and access to specialist rehabilitation services may be required.

Data availability statement

Data are available upon reasonable request.

Ethics statements

Patient consent for publication

Ethics approval

This study involves human participants and the Ministry of Defence Research Ethics Committee granted favourable opinion to this study (1061/MoDREC/20). Participants gave informed consent to participate in the study before taking part.


To all the participants, administrative staff and support teams at DMRC Stanford Hall and OUH, we acknowledge, and thank, your hard work, dedication and valuable input.


Supplementary materials


  • Twitter @ollieosul

  • Contributors DAH, ANB, EN, PL, OO'S and RB-D conceived the study. DAH, EN and ANB secured funding to deliver the research. RB-D, OO'S, AG, PL and RC acquired data at DMRC. AH and RB-D provided statistical analysis and produced graphs. RB-D, with support from PL, OO'S, AH and ANB, drafted the manuscript. All authors reviewed the manuscript. ANB acted as chief investigator and guarantor for the study.

  • Funding A grant was received from the Defence Medical Services Research Steering Group. No specific funding number was assigned.

  • Disclaimer The Defence Medical Services Research Steering Group had no role in the preparation of this manuscript.

  • Competing interests None declared.

  • Provenance and peer review Not commissioned; internally 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.