IntroductionBackground and rationale

During the COVID-19 pandemic, China adhered to stringent ‘zero COVID’ policies for nearly 3 years with strict lockdowns and restrictive measures. As of 7 December 2022, China cancelled strict restrictions, which led to a huge surge of COVID-19 infection, the first wave of COVID-19, from December 2022 to January 2023. Although substantial studies have delved into the acute effects of COVID-19, research on the short-term and long-term health consequences remains inadequate. Some studies have predicted that at least 65 million people worldwide may suffer from long COVID on the basis of documented cases around the world and a conservative estimated incidence of 10% of infected people.1 An investigation in China reported that 76% of infected patients in Wuhan continued to experience at least one symptom after 6 months after hospital discharge.2 Despite growing concerns about the long-term effects of COVID-19, definitive clinical guidelines have not yet been established. Given that the COVID-19 infection in China is predominantly Omicron, findings of international research may not fully apply. With the high rate of COVID-19 infection during the pandemic and the high incidence of sequelae, the need to understand and respond to long COVID is becoming increasingly urgent.3

According to the National Institute for Health and Care Excellence guideline, symptoms related to COVID-19 are classified into three categories: acute COVID-19 (symptoms and abnormalities lasting up to 4 weeks), ongoing symptomatic COVID-19 (symptoms and abnormalities present from 4 to 12 weeks beyond the acute phase) and post-COVID-19 syndrome (symptoms and abnormalities continuing beyond 12 weeks and not attributable to other diagnoses), but there is no agreed-upon definition.3 4 The concept of the postacute COVID-19 has been proposed to include persistent symptoms or development of sequelae beyond 3 or 4 weeks from the onset of symptoms of acute COVID-19.5 6 Therefore, this study will investigate participants with confirmed mild or moderate COVID-19 infection at the different stages of COVID-19: acute infection phase (within 4 weeks) and 3, 12 and 24 months after infection. Long COVID, also termed ‘the post COVID-19 condition’ or ‘persistence of post-COVID-19 syndrome’ by the WHO, is commonly used to describe symptoms that continue or develop after the acute phase of COVID-19.3 7 Patients with long COVID exhibit a heterogeneous range of physical and emotional symptoms.8 Common symptoms of long COVID include sensory issues (altered taste and/or smell), neurological symptoms (fatigue, having trouble with concentration and ‘brain fog’) and cardiorespiratory abnormalities (shortness of breath).9 Neuroimaging has provided objective evidence for the coexistence of recoverable and long-term unrecovered changes on the brain due to long COVID-19.10 Long-term follow-up MRI research showed grey matter volume changes, white matter abnormalities, microbleeds and functional connectivity changes in patients with COVID-19.11–13 Such nervous system changes may result in mental and neuropsychological abnormalities, including emotional distress (eg, anxiety and depression) and cognitive dysfunction (eg, memory, concentration and executive dysfunction).14–16

To the best of our knowledge, there is a lack of dynamic observation studies in China that track the progression of long COVID population from the acute to the recovery phase. By proposing this large-scale observational research combining multimodal MRI data, haematological indices and outcomes of neuropsychological scales and questionnaires, we aim to trace the progression of physical and mental health over follow-up time, explore the underlying biological mechanism and identify predictive neuroimaging and haematological biomarkers associated with the COVID-19 effects. A better understanding of long COVID will provide information to facilitate the identification, diagnosis and treatment of this condition, ultimately helping reduce its long-term adverse effects.

Study objectives

This investigation is not only a longitudinal prospective follow-up study on long COVID but also a cross-sectional study, comparing patients with mild or moderate COVID-19 infection and healthy controls (HCs). The main specific objectives of the study are as follows:

Establish a Chinese data set of long COVID, including the multimodal MRI data of the brain and spine, haematological indices and outcome of neuropsychological scales and questionnaires.

Investigate the demography and development and features of main symptoms in patients with COVID-19 over follow-up time.

Explore the progression of neuroimaging damage, haematological abnormalities, neuropsychological disturbances and main symptoms in different COVID stages and their interrelationships.

Predict progression of long COVID using MRI and haematological biomarkers and understand more about its biological mechanism.

Methods and analysisStudy design

This is a multicentre observational and prospective follow-up cohort study and will be performed at nine public hospitals. At baseline, patients will recall their preinfection status (ie, time point 0 (T0)) condition and complete some questionnaires, including preinfection symptoms. At the acute phase (ie, within 4 weeks, time point 1 (T1)), 3 months (ie, time point 2 (T2)), 12 months (ie, time point 3 (T3)) and 24 months (ie, time point 4 (T4)) after the onset of COVID-19, participants with confirmed mild or moderate COVID-19 infection will complete the following programmes: (a) submit blood samples at the local laboratory, (b) undergo multimodal MRI scan of the brain and spine, and (c) fill in the neuropsychological scales and questionnaires. The uninfected HCs will complete the same programmes as the infected group mentioned above at the time of inclusion. A Chinese data set of long COVID will be established, including multimodal MRI data, haematological indices and outcome of neuropsychological scales and questionnaires.

Participants

The study population consists of (a) patients with COVID-19 admitted to one of the nine recruiting hospitals in China (ie, The First Affiliated Hospital of Xi'an Jiaotong University, No.215 Hospital of Shaanxi Nuclear Geology, Yanan Traditional Chinese Medicine Hospital, Shangluo Central Hospital, Xi'an QinHuang Hospital, Yulin No. 2 Hospital, Hanzhong Central Hospital, Ankang Central Hospital, Baoji High-Tech Hospital) during the first wave of COVID-19 in China (from December 2022 to January 2023) and (b) individuals who have never been infected with COVID-19 during the first wave of COVID-19 in China, referred to as HCs.

Inclusion criteria

Patients infected with COVID-19

Confirmed diagnosis of mild or moderate COVID-19 (WHO criteria).

Infected during December 2022 or January 2023.

From 18 to 65 years of age and right-handedness.

Informed consent.

HCs

Exclusion criteria

Chronic lung illness, central nervous system diseases, neurological or psychiatric disorders (eg, mental disorder, degenerative diseases of the central nervous system, tumours, trauma) or left-handedness.

Contraindications for MRI scanning (eg, metallic or pacemaker implants, claustrophobia or pregnancy, etc).

Recruitment process

The study will be carried out in nine hospitals among healthcare workers (physicians, nurses, employees) and recruit subjects from the community through media, social networks and advertisements. The prerequisite for patients with COVID-19 to be taken into account to our research is that they are infected with COVID-19 confirmed by quantitative real-time PCR (qRT-PCR) or antigen detection test. Since the emergence of COVID-19 in 2019, many people in China have been tested by qRT-PCR multiple times due to national policies. So as for HCs, one inclusion criterion is no history of COVID-19 infection. This requires that HCs have no clinical symptoms of COVID-19 during the first wave and that all of their qRT-PCR tests have negative results. All participants will sign an informed consent form.

Study procedures

After providing informed consent, participants (patients at four time points (T1, T2, T3, T4) and HCs at the time of inclusion) will undergo the following programmes: (a) blood draw on an empty stomach in the morning by trained nurses, (b) multimodal MRI scan of the brain and spine performed at the local radiology departments according to standard procedures (including an MRI safety check) and (c) neuropsychological scales and questionnaires carried out by trained research assistants on the same day (figure 1). If blood testing, MRI scanning and neuropsychological scales and questionnaires cannot be completed on the same day (eg, due to hospital scheduling constraints), separate appointments will be scheduled at a maximum of 3 days apart. Blood samples will be sent immediately to the local clinical laboratory for analysis using professional equipment.

Figure 1

The time flow of the study design. The dark figure represents patients with COVID-19, and the light figure represents healthy controls. T1, time point 1; T2, time point 2; T3, time point 3; T4, time point 4.

Outcome measures and measurement toolsDemographics and disease-related characteristics

Demographic data, including age, sex, body mass index, education, occupation, lifestyle (smoking, alcohol consumption and exercise habits), basic disease and vaccination status, will be collected from patients and HCs via a basic information questionnaire at the time of inclusion. Patients will additionally provide information on disease-related characteristics, including COVID-19 onset date, initial symptoms, disease course, fever history, clinical classification and medication use.

MRI protocol

Participants will receive a 3T multimodal MRI scan to assess nervous system damage. Considering that olfactory dysfunction is acknowledged as the primary symptoms of COVID-19 by the WHO and one of the common symptoms of long COVID, the MRI scan protocol adds sequences to acquire olfactory bulb MRI. The scanning protocol consists of the following sequences:

Brain: T1-weighted MRI, T2-weighted MRI, fluid-attenuated inversion recovery (FLAIR) imaging, magnetic resonance image complication, resting-state functional MRI, intravoxel incoherent motion, diffusion tensor imaging, diffusion kurtosis imaging, perfusion-weighted imaging, arterial spin labelling, time of flight-magnetic resonance angiography (TOF-MRA).

Olfactory bulb: T2-weighted MRI.

Spine: T1-weighted and T2-weighted MRI.

These sequences allow optimal evaluation of the nervous system damage on the basis of the findings from prior literature. Primary imaging diagnosis will be reported by two certified radiologists based on all structural images. Subsequently, most sequences will be analysed using specialised software by research. The variables generated by analysing each sequence and the purpose are shown in table 1, and the corresponding neuroimaging damage will be described compared with the HCs.

Table 1

Neuroimaging analysis

Blood test

Participants will receive blood test on an empty stomach, which includes the following items:

IgM and IgG antibody against SARS-CoV-2, analysed by chemiluminescence immunoassay to identify SARS-CoV-2 infection.

Cytokine (interleukin-1beta (IL-1β), IL-2, IL-4, IL-5, IL-6, IL-8, IL-10, IL-12, IL-17, interferon-alpha (IFN-α), interferon-gamma (IFN-γ), tumour necrosis factor alpha (TNF-α)), detected by flow cytometry (FCM).

Serum myocardial enzymes (lactic dehydrogenase and creatine kinase).

Erythrocyte sedimentation rate.

Complete blood cell count with differential, detected by FCM.

C reactive protein and hypersensitive C reactive protein, detected by FCM.

Kidney and liver function test.

Procalcitonin.

Some subcentres perform additional tests such as coagulation tests (eg, prothrombin time, activated partial thromboplastin time, thrombin time, fibrinogen), D-dimer, routine urinalysis and scatological analysis, and brain natriuretic peptide. All test results are provided by the clinical laboratory, following standard clinical criteria.

Neuropsychological scales and questionnaires

To assess cognitive, emotional and behavioural dysfunction, participants will complete neuropsychological scales (approximately 45 min) based on current literature. Patients and HCs will complete a series of scales to investigate the health (condition and disease symptoms), neurological symptoms (pain and fatigue), emotional distress (depression, anxiety and traumatic stress), sleep (quality and satisfaction) and cognitive function (eg, attention, memory and executive function) at appointed time points. At the beginning of the participation, patients will recall the preinfection (ie, T0) health condition and complete relevant scales. As the research progresses, new scales will be added if needed. Table 2 shows an overview of the investigated domain, assessment instrument and the time point of investigation. The following scales will be administered:

The EuroQol-5D-5L is a self-completion questionnaire for describing and valuing health condition, based on a descriptive system that defines health in five dimensions: mobility, self-care, usual activities, pain/discomfort and anxiety/depression.17 Each dimension has three response categories: no problems, some problems and extreme problems. The questionnaire respondents also mark their overall health on a 0–100 hash-marked, vertical visual analogue scale on the day of the investigation.18

Table 2

Outcome measures and measurement tools per time point

The Patient Health Questionnaire-9 (PHQ-9) is a reliable and valid clinical tool of making criteria-based diagnoses of depressive disorders and measuring depression severity. The PHQ-9 score corresponding to the level of depression severity is 1–4 (minimal), 5–9 (mild), 10–14 (moderate), 15–19 (moderately severe) and 20–27 (severe).19

Impact of Event Scale – Revised (IES-R) is a useful instrument in the assessment of traumatic stress and concerned more with measuring the core constructs of intrusion, avoidance and arousal that characterise traumatic stress.20 The IES-R scores higher than 24 are of concern, and the higher the scores, the greater the concern for post-traumatic stress disorder and associated health and well-being consequences.

The items of the Fatigue Impact Scale (FIS) can be divided into three parts to assess physical, cognitive and psychosocial functioning, with higher scores indicating more severe fatigue.21

The Multidimensional Fatigue Inventory (MFI), a 20-item self-report instrument, reflects the following dimensions: general fatigue, physical fatigue, mental fatigue, reduced motivation and reduced activity. Higher scores indicate more severe fatigue.22

The Fatigue Assessment Scale (FAS) includes 10 fatigue items covering physical fatigue and mental fatigue, five questions, respectively. Scores between 0 and 22 indicate no fatigue, between 22 and 34 mild-to-moderate fatigue, and between 35 and 50 severe fatigue in sarcoidosis.23

The Pain-Visual Analogue Scale is a well-known visual analogue scale for the assessment of pain intensity and functions best for the subject’s subjective feeling of the intensity of pain right now, present pain intensity.24

Digit Span Test focuses on auditory attention and working memory with the digit span forward and backward.25 The participant is provided with a series of digits (eg, 6, 9 and 4) and is asked to repeat them in the same (forward digit span test) or in the reverse order (backward digit span test). If the respondent is correct, a longer list is provided. The outcome is the greatest number of digits that the participant is capable of repeating correctly. Higher scores indicate better performance.

Auditory Verbal Learning Test assesses the verbal memory.26 A list of 12 words is read by the examiner three times and participants recall them following each presentation, which is called ‘immediate recall’. The ‘short delayed recall’ and ‘long delayed recall’ require the participants to recall the list after 5 min and 20 min, respectively, and the number of correctly recalled words is recorded. Finally, the subjects are instructed to recognise the learnt words from a list of 24 words which consists of 12 learnt words from the above list and 12 unlearnt words. The more the number of words, the better the immediate memory or delayed recall ability.

The Montreal Cognitive Assessment (MoCA) tests global cognitive function.27 Executive functions in multiple aspects are evaluated using an alternating task adapted from the Trail Making B task, a phonemic fluency task, and a two-item verbal abstraction task. A three-dimensional cube copy and a clock-drawing task are used to assess visuospatial abilities. Language is evaluated using a three-item confrontation naming task with low-familiarity animals, repeating syntactically of two complex sentences task, and the aforementioned fluency task. The short-term memory recall task consists of two learning trials of five nouns and after approximately 5 min, participants need to recall them. Working memory, attention and concentration are evaluated using digits forward and backward, a sustained attention task and a serial subtraction task. At last, orientation to time and place is assessed. Higher scores (0–30) indicate a better performance. Using a cut-off score 26, the MoCA has a sensitivity of 90% to detect MCI.27

The Trail Making Test A and B and Stroop Test assess cognitive processing speed and executive functioning skills.28 29 For each participant, the time it takes to complete the connection of sequential numbers (TMT-A) and interconnect sequential numbers and letters (TMT-B) is recorded. The TMT-B is replaced by Stroop Test if the participants are unfamiliar with alphabetical order. Shorter completion time of both parts indicates better cognitive function.

The Beck Anxiety Inventory (BAI) is a useful measure of assessing anxiety that was constructed to avoid confounding with depression.30 Higher scores are associated with anxiety.

The Beck Depression Inventory (BDI) is among the most used self-rating scales for measuring depression and is popular in clinical applications and research. The BDI measures the intensity of depression by the main symptoms of the depressive syndrome and reliably discriminates between depressives and nondepressives.31 Sum scores range from 0 to 63. BDI scores less than 13 indicate minimal depression, scores between 14 and 19 indicate mild depression, scores between 20 and 28 indicate moderate depression and scores at least 29 indicate severe depression.32

The Pittsburgh Sleep Quality Index (PSQI) consists of seven self-rated component scores: sleep quality; sleep latency; sleep duration; habitual sleep efficiency; sleep disturbance; daytime dysfunction and use of sleeping medication.33 PSQI total scores more than 5 have good diagnostic sensitivity and specificity in distinguishing good sleep quality from poor sleep quality.34

The Insomnia Severity Index assesses nighttime and daytime insomnia symptoms, consisting of difficulty in falling asleep and sleep maintenance, early awakening, satisfaction with sleep patterns, interference with daily function, perceived prominence of impairment attributed to insomnia and concerns about sleep problems. Each item is rated on a 5-point Likert scale, with a total score ranging from 0 to 28 and a higher score denotes more severe insomnia. The scale is frequently used as a follow-up method in several clinical research. Higher scores suggest more severe insomnia.35 36

All neuropsychological scales and questionnaires will ultimately be assessed by experts in the cognitive and psychological domains.

At the first time point, symptoms of COVID-19 were assessed using a simple questionnaire. To be more scientific, this study specifically developed the long COVID symptoms and severity score (LC-SSS) at the second time point to assess the severity of commonly recorded symptoms associated with long COVID by self-report. The scores of each symptom are added together to get a total score for each category (Gastrointestinal, ENT (Ear, Nose, Throat), Neurological, Pain, Cardiopulmonary, Psychological) while the total score of this questionnaire is the sum of six categories. A score of zero indicates an absence of long COVID symptoms. The structural validity of the LC-SSS was good with a comparative fit index of 0.969 and Cronbach’s α of 0.93, and test–retest reliability was also satisfactory (intraclass correlation coefficient 0.732), indicating excellent internal consistency and that the LC-SSS provides a robust and valid tool for assessing long COVID.37 The detailed contents of the long COVID Symptoms and Severity Score are displayed in online supplemental file 1.

All team members will undergo training and standardisation sessions to ensure consistency in the interpretation and diagnosis of the outcomes across different sites or time points.

Sample size

As the first wave of COVID-19 in China is a particular public health emergency and the first time point of the study was at the time of the COVID-19 pandemic, the specific infection rate of COVID-19 in the general population of China during the first wave was not known at the beginning of the study. Consequently, no formal sample size calculation was conducted prior to implementing the study protocol. At the first time point, 501 participants (418 patients and 83 HCs) from nine recruiting hospitals have been observed. A power analysis for independent two-sample t-tests using G*Power was applied.38 Some showed good performance, for instance, the statistical power of PHQ-9, FIS, MFI, FAS and BAI was 0.99, 0.96, 0.99, 0.99 and 0.99, respectively, showing that the number of participants allowed for detecting differences between difference between groups (Cohen’s d=−0.82, –0.46, −0.76, –0.59 and −0.54, respectively). The final sample size of n=501 was pragmatically selected, balancing the competing demands of maximising statistical power while simultaneously minimising institutional burden.

Statistical analysis

To answer objective 2, follow-up analyses will be used to evaluate the differences in main symptoms and demographic at four time points between 418 patients with COVID-19 (mild or moderate) and 83 HCs. Independent t-test will be used for normally distributed continuous data, Mann-Whitney U test for non-normally distributed data and χ2 test for categorical variables. Multivariate analyses will be used to correct for potential confounders, including comorbidities. Using standard scores allows the normalisation of scores from different questions, enabling statistical comparability. This is useful for comparing results across different individuals or studies because it eliminates the influence of the original score ranges. Demographic analysis will also be performed using the above methods. To answer objectives 3–4, correlation analysis and multivariate regression analysis will examine the relationships among MRI, haematology, neuropsychology and symptoms in different stages. Potential predictors are defined as variables with a marginally significant association (p<0.10) with the outcome variable. Only these variables will be included in the subsequent regression analyses to determine the most significant predictors. Logistic regression will be applied for categorical variables and linear regression for continuous variables. Furthermore, to address the initial imbalance between groups, propensity score-matching analysis will be used to balance the groups in certain analyses, determining the patient-to-control matching ratio based on research objectives.39 This statistical approach will help mitigate the effects of the sample size discrepancy. Generally, a two-tailed p value of <0.05 is considered statistically significant.

Patient and public involvement

Given the recognised practical and ethical justifications for involving patients in research, the study was designed in partnership with patient and public involvement (PPI) to improve the quality and relevance of research and finally benefit research outputs. The PPI group consisted of patients, collaborators, advocates and members of the public. These contributors were involved from the initial phase and advised on trial design. Collaborators contributed their experience in the use of consensus methods, and patients shared their personal experiences of the COVID-19 sequelae, advising on patient-orientated outcomes. After discussing the primary and secondary outcomes together, the PPI group will review the content and coauthor the manuscript and finally support in disseminating the results.

Discussion

The first wave of COVID-19 in China was from December 2022 to January 2023. The dominant variants in China at that time were BA.5.2 and BF.7 which are transmissible and spread rapidly. For instance, given the transmission dynamics of BF.7, one modelling analysis estimated that 76% of the Beijing population may have been infected with COVID-19 by 22 December 2022, and 92% of the Beijing population may have been infected with COVID-19 already by the end of January 2023.40 In other words, a very high percentage of the population was infected with COVID-19 from December 2022 to January 2023. Significantly, an increasing number of patients presented with postacute nervous system sequelae, even those without neurological manifestations in the acute phase.10 Early detection and prevention of COVID-19-related neurological dysfunction are crucial, and the need to understand and respond to long COVID is increasingly pressing. The available literature and mounting evidence suggest potential persistent effects of COVID-19 on many aspects, but the relative long-term health consequences of COVID-19 are still not fully understood. Additionally, most of the long COVID research only focus on one single time point after the onset of COVID and lack continuous follow-up. Observation of dynamic brain changes is conducive to a better understanding and early prevention of long COVID. Therefore, according to the current literature, the longitudinal prospective follow-up study will investigate patients infected during the first wave of COVID-19 in China at four time points (acute infection phase (within 28 weeks) and 3, 12 and 24 months after infection) to fully understand long COVID, which is the critical strength of this study. Additionally, this study aims to investigate the long-term sequelae of COVID-19 and the relationships among the COVID-19-associated neuroimaging damage, haematological abnormalities, neuropsychological disturbances and symptoms. To reduce the sampling error and improve the reliability of the results, patients come from nine hospitals and the sample size is large enough.

Recently, MRI research observed the effects of COVID-19 on the nervous system in the acute/subacute phase and showed various forms of nervous system damage, which consisted of diffuse cerebral white matter hypodensities/hyperintensities, leukoencephalopathy, cortical FLAIR signal abnormalities, acute and subacute infarcts, microhaemorrhages and leptomeningeal contrast enhancement.41–43 However, little is known of the longer follow-up and longitudinal consequences of COVID-19 on the nervous system.10 Longitudinal follow-up studies are required to explore the long-term neurological effects of the COVID-19 pandemic and shed light on its related mechanisms. So far, the potential pathogenesis of long COVID remains unclear. The proposed biological mechanisms for COVID-19-related nervous system damage include viral invasion of the nervous system, hyperinflammation, vasculitis, hypoxia and hypercoagulation.44 45 Common haematological abnormalities such as decreased platelets, lymphocytes, haemoglobin, eosinophils and basophils have been observed in patients with COVID-19 since the early stage and are associated with the clinical outcome and disease severity.46 The hyperinflammatory state of patients with COVID-19 is reflected by high levels of IL-2, IL-4, IL-6, IL-10, TNF-α, IFN-γ, C reactive protein and erythrocyte sedimentation rate compared with HCs, with IL-6 being the most commonly reported cytokine elevated in patients with COVID-19.47–49 Some specific laboratory indicators suggested the deterioration of disease, such as leucocytosis, lymphopenia, platelet, C reactive protein, alanine aminotransferase, aspartate aminotransferase, albumin, creatinine, procalcitonin, D-dimer, creatine kinase and lactate dehydrogenase.50 Combining neuroimaging and haematology can provide more insight into the aetiological mechanism of COVID-19. Therefore, the study will collect blood samples from HCs and patients (four time points), and it is possible to compare the differences in haematology cross-sectionally and longitudinally or to dynamically observe the changes in haematology in the population infected with COVID-19.

This study has several limitations. First, the main limitation is that the small size of the HC group as the uninfected population was rare during the first wave of COVID-19 in China, much less than the infected population, due to the high infection rate. Some patients with COVID-19 are asymptomatic or have only mild symptoms, which misleads them that they are uninfected and influences the results. To avoid the situation, the uninfected population will be verified by the IgM and IgG antibody against SARS-CoV-2 test. Additionally, some of the HCs are reluctant to go to the hospital to participate in our research considering the high risk of infection. Second, since the study is time consuming and requires patients to participate four times, the loss rate of follow-up may be high, which may lead to data bias. Therefore, during the recruitment process, the study will make great efforts to improve the compliance and reduce the loss rate of follow-up. The chosen study sample is limited to the first infection wave of COVID-19 in China, which may potentially reduce the generalisability of the results. Third, the majority of patients treated COVID-19 with medication during the period of illness, such as febrifuge, traditional Chinese medicine, cough medicine and painkiller. It is conceivable that the type of medicine will lead to different or reduced neurological and neuropsychological consequences. The reduction in the inflammatory response relieves symptoms and reduces illness severity, thereby reducing the risk of damage to the nerve system. To reduce the bias caused by the different types of medicine, the study will record the medicine used by patients during the period of illness and subsequent analyses will take the bias due to medicine use into account. Finally, due to concerns about nosocomial transmission, the study did not collect MRI, haematology and other data during the period of illness, so the study lacks data on the neurological damage caused by the virus in the early stages of the outbreak.

Despite these limitations, this investigation is real-world research which traces the evolution of neuroimaging, haematological indices, neuropsychological performance and symptoms in different stages after COVID infection and explores their relationships, providing detailed insight into the complex relationships among the COVID-19-associated nervous system damage, haematological abnormalities, neuropsychological disturbances and disease symptoms. This longitudinal prospective follow-up study will explore neuroimaging and haematological biomarkers of long COVID and understand more about its biological mechanism, which is conducive to predicting the progression of long COVID and identifying high-risk patients developing long COVID to aid in prognosis and early intervention. This may help the physician not only to estimate the patient’s burden of symptoms but also monitor the evolution of the disease over time. Additionally, this work can identify key areas for further research and potential target for therapeutic strategies. Finally, the study will establish a Chinese data set of long COVID, including the multimodal MRI data of the brain and spine, haematological indices and outcome of neuropsychological scales and questionnaires.