Recruitment
From May 13, 2020, to January 11, 2022, more than three thousand patients were evaluated in the newly established post-COVID ambulatory at the Department of Pulmonary Diseases and Tuberculosis, University Hospital Olomouc, Czech Republic. The availability of the post-COVID ambulatory was publicly announced on the website of the University Hospital Olomouc, and patients could book appointments without the need for a recommendation from a general practitioner or another specialist. The patients were invited to participate in a longitudinal observational study of COVID-19 survivors. Those who agreed to participate in the study signed an informed consent form. Non-participants received the same level of medical care as study participants, but their clinical data were neither collected nor included in the dataset.
The inclusion criteria for the study were: a minimum age of 15, and a previous COVID-19 infection. A COVID-19 infection was confirmed by one or more of the following: a history of a positive SARS-CoV-2 PCR test, a history of a positive SARS-CoV-2 antigen test coupled with typical symptoms, or positive levels of SARS-CoV-2 IgM antibodies in non-vaccinated individuals who experienced typical symptoms. Notably, the presence of subjective symptoms was not a requirement for inclusion.
No primary exclusion criteria were defined, as the impact of different comorbidities on the recovery after COVID-19 was one of the interests of the study. However, the patients with a probable refutation of SARS-CoV-2 infection were excluded post hoc. This probable refutation was defined by an absence of a positive PCR or antigen test for SARS-CoV-2 and undetectable SARS-CoV-2 antibody levels in serum.
Course of the study
The study was delineated by two primary visits. The initial visit (V1) was scheduled in a mean of 12 weeks (range = 1–68, SD ± 7.55) following the confirmed infection. The second follow-up visit (V2) was scheduled based on clinical necessity, within a mean interval of 21 weeks (range = 3–50, SD ± 7.96) between the two visits. In a limited number of subjects, additional check-ups were necessary; however, but the scope of these additional examinations was limited based on clinical requirements.
Both V1 and V2 consisted of an extensive patient interview, documented in free-form by one of the investigating pneumologists (N = 14). The focus of these interviews was on the symptoms of acute COVID-19, the persistence of residual symptoms, the emergence of new symptoms during the follow-up, comorbidities, medications, occupational history, and other details customary for a comprehensive pneumological examination. Patients underwent a primarily cardio-respiratory focused physical examination. The findings of this objective examination were recorded in free-form according to the standards of the department.
The patients received a chest X-ray (CXR) in posteroanterior projection, which was subsequently interpreted by a trained radiologist (N = 25). The radiologist's description, alongside any comments from the investigating pneumologist, were documented in the report. Chest HRCT was performed in selected individual based on clinical necessity. Pulmonary function testing (PFT) - including spirometry, body-plethysmography, and lung diffusion capacity for carbon monoxide - was conducted by experienced spirometry technicians using the MasterScreen by Jaeger®. Data retrieval was accomplished through the SentrySuiteTM Version 2.19 by CareFusion. The results, along with their interpretation by the investigating physician, were recorded.
Additionally, a set of blood tests was performed at both V1 and V2. These tests included differential blood count, C-reactive protein (CRP), kidney and liver function tests, routine coagulation parameters, and antibodies against SARS-CoV-2 (IgM and IgG classes) in non-vaccinated subjects. These blood tests were initially performed for all patients, but their use was later (from October 2021) restricted to cases with clinical necessity. A limited group of consecutive patients also had blood samples taken for a complex analysis of cellular immunity.
Furthermore, the participants were administered a questionnaire provided by the Department of Psychiatry, University hospital Olomouc, including the Beck Anxiety Inventory.
Glucocorticoid treatment
The treatment with glucocorticoids (CS) was considered in patients with persistent pathological findings on CXR and/or HRCT and/or impaired pulmonary function parameters, namely decreased DLCO. The presence of symptoms alone was not considered to be an indication for glucocorticoid therapy. In patients considered for glucocorticoid therapy, additional criteria such as age, frailty, comorbidities, and other individual risk factors were evaluated by the examining physician. Finally, selected patients were prescribed oral glucocorticoids (prednisolone or an equivalent dose of methylprednisolone).
The initial dose and the tapering regimen were individualized, with doses lower than 0.5 mg/kg of prednisolone being utilized in patients with higher estimated risk of adverse effects. The mean prescribed initial dose of prednisolone was 27.6mg (SD ± 10,64), and the mean duration of glucocorticoid therapy was 13.3 weeks (SD ± 10,06).
This approach was in line with the national positional document on the treatment of pulmonary impairment in patients recovering from COVID-19.7
Construction and analysis of the dataset
Data sources and processing
The data curation and wrangling process began shortly after the launch of the study, with four main data sources being utilized: the medical documentation, the database of laboratory test results, the database of radiological studies and the Czech national registry of SARS-CoV-2 vaccination and tests. The data were extracted, encoded, and stored in a Microsoft Excel spreadsheet. The data mining from the medical documentation was done manually, while the database sources were processed semi-automatically. The data were cleaned and finally validated by the main author prior statistical analysis.
Biometric data, timeline
The medical documentation of each enrolled patient was searched manually, including the reports from the two main visits (V1 and V2) and all interim check-ups. Biometric data were gathered (sex, age, weight, height), body mass index (BMI) was calculated, and a BMI category25 was determined ranging from underweight to class III obesity.
The date of the first positive SARS-CoV-2 test (or onset of symptoms) was extracted from the medical documentation; the dates were later manually validated against the Czech national registry of SARS-CoV-2 vaccination and tests. The dates of V1 and V2 were recorded to allow the establishment of a timeline for each patient. The reasons for missing follow-up were recorded.
Lastly, the continuation or lack of follow-up after V2 was noted to facilitate future extension of the dataset.
Comorbidity score
The presence or absence of selected comorbidities in the patient’s medical history prior suffering from COVID-19 was recorded, including arterial hypertension (AHT), diabetes mellitus (DM), ischemic heart disease (IHD), pulmonary embolism (PE), dyslipidemia (DLD), hypothyroidism (HT), bronchial asthma (BA), chronic obstructive pulmonary disease (COPD), sarcoidosis, interstitial lung disease (ILD). Other relevant diagnoses were recorded as plain text for future processing. The comorbidity score (ComS) was established by adding one point for each of the ten selected comorbidities.
Severity of acute COVID-19, categorization of post-COVID syndrome
The severity of acute COVID-19 was classified into one of three categories: mild, defined as a mild ambulatory course without pneumonia; moderate, involving pneumonia but not requiring oxygen therapy; and severe, involving pneumonia that required oxygen therapy, ventilatory support, intensive care, or critical care.
Furthermore, the patients were grouped into one of the four post-COVID syndrome categories (A-D) proposed by Skala et al. These classifications were based on the presence or absence of subjective respiratory symptoms and objective signs of respiratory pathology (as detailed in Table 1).
Subjective symptoms and their dynamics
The data reporting on selected subjective symptoms experienced during the acute phase of COVID-19 (ageusia or anosmia, gastrointestinal symptoms, asymptomatic course) and post-COVID symptoms (dyspnoea, cough, fatigue, fever, anosmia, ageusia, chest discomfort or pain, headache, undulating course of subjective symptoms) were extracted and recorded. The overall change in the perceived severity of subjective symptoms between V1 and V2 was estimated based on the verbal description of patient’s symptoms, as recorded by the examining physician, and a score subjective improvement score (SIS) was recorded using a Likert type scale ranging from 0 (worsening, or no improvement) to 10 (complete resolution of symptoms). To limit the combined observer bias, the scoring for all patients was conducted post-hoc by a single pneumologist. In cases where a lack of information prevented the establishment of the SIS, the data point was left blank, and the reason for this absence of data (no symptoms at the initial visit, missing follow-up, or insufficient documentation) was recorded.
Radiological findings and their dynamics
The findings on CXR, and HRCT where necessary, were sorted into categories ranging from 0 to 5. This classification was referred to as the "post-COVID radiological score" (PRS) and was based on the extent of post-COVID residual changes (refer to Table 2). The scoring was based upon the written descriptions of the CXR and HRCT studies provided by a trained radiologist. In instances where the description did not permit accurate categorization, the radiological study was reviewed by a pneumologist proficient in reading chest X-rays and CT scans, and the category was then assigned.
The radiological findings at V1 and V2 were compared side by side and the radiological improvement score (RIS) was estimated on a Likert scale ranging from 0 (indicating worsening or no improvement) to 10 (indicating a complete resolution of the pathological radiological finding). To reduce combined observer bias, the scoring of the initial radiological findings and their rate of improvement was carried out by a single pneumologist skilled in interpreting radiological studies.
Pulmonary function testing, blood tests
The results from blood tests and data from pulmonary function tests (PFT) were collected in a semi-automated routine. The PFT data encompassed predicted values, actual measured values, and relative values (expressed as a percentage of predicted) where applicable. The difference between the values obtained at V2 and V1 was calculated as both an absolute and relative change over the entire period between the two main visits, and as an amount of change per week of follow-up.
Glucocorticoid therapy
The data regarding the use of CS were gathered from the source documentation by a pneumologist blinded to the subjective and radiological improvement scores, PFT and laboratory test results. The data included the glucocorticoid molecule used, initial dosage expressed as an equivalent dose of prednisolone, and the duration of the administration in weeks. Patients on chronic CS medication (rheumatoid arthritis, solid organ transplant recipients, etc.), and patients with newly initiated CS therapy for reasons other than post-COVID respiratory sequelae (bronchial asthma exacerbation, newly diagnosed sarcoidosis, etc.) were marked, but they were still included in the analysis.
Statistical analysis
The data regarding biometric parameters, subjective symptoms and their improvement, radiological findings and their improvement, selected PFT parameters and their dynamic, the severity of COVID-19, post-COVID syndrome category, and the data regarding glucocorticoid steroid therapy, the initial dose and duration of therapy were statistically described and hypotheses were tested by means of univariate analysis. Shapiro-Wilk test and visual examination of the density plots were used to detect violations of normality. Parametric and non-parametric hypothesis tests were used as applicable (Student’s t-test, Mann-Whitney U test, Chi-square test, Kruskal-Wallis test with Dunn’s post hoc tests, Spearman correlation, ANOVA). All tests were conducted at the alpha level of 0.05. Logistic regression was employed to investigate potential predictors of a lack of subjective and radiological improvement, defined as SIS = 0 and RIS = 0, respectively. The predictors evaluated in the logistic regression included age, sex, COVID-19 severity, use of glucocorticoids, comorbidity score, and obesity (defined as BMI >= 30). All statistical tests were performed by an experienced statistician using the IBM SPSS Statistics software package.
At the time of writing this paper, efforts to complete and expand the Olomouc post-COVID dataset are ongoing, involving the authors and associated collectives. The status and availability of different variables within the dataset vary, therefore some parts of the following analyses were conducted on different subsets of subjects. As such, whenever the analyses were not performed on the entire sample, the specific number of observations is indicated.
Ethics approval and informed consent
All subjects agreed to join the study and signed informed consent. The local ethical committee approved the study (Ethical Committee of University Hospital Olomouc and Faculty of Medicine and Dentistry of Palacky University Olomouc, Czech Republic), with decision number 98/21 (date 7/JUN/2021).