Short-term Effects of a Three-week Inpatient Post-COVID-19 Pulmonary Rehabilitation Program - a Prospective Observational Study

Background: The clinical course of coronavirus disease 2019 (COVID-19) varies distinctly. Particularly after severe or critical courses, i.e., after hospitalization, clinical manifestations frequently persist after the acute phase. However, symptoms may also persist after initially milder courses that can be treated in an outpatient setting. For patients who remain symptomatic after COVID-19, pulmonary rehabilitation (PR) is recommended. However, only few studies investigated the effectiveness of PR, especially considering different disease courses. The main objective was to evaluate the feasibility, safety, and ecacy of post-COVID-19 PR. Methods: A total of 120 post-COVID-19 patients who were referred to the Bad Reichenhall Clinic between April 2020 and January 2021 were asked to participate in this prospective observational study. PR was tailored to each patient’s individual needs and was based on the current recommendations. The primary outcome dyspnea was assessed with numerical rating scales and the modied Medical Research Council (mMRC) dyspnea scale. Secondary outcomes included other symptoms such as cough and sputum; physical capacity; lung function; fatigue; quality of life (QoL); depression; and anxiety. Furthermore, patients rated the overall effectiveness of PR and their subjective change in health status. Results: A total of 108 patients (mean age 55.6±10.1 years, 45.4% female) were included and were assigned to 3 groups depending on the referral mode (A: severe acute; B: severe after interval, C: mild after interval). At the end of PR, we detected improvements in the intensity of exertional dyspnea, physical capacity, QoL, fatigue, and depression in the overall group, with large effect sizes (Cohen’s d>0.8). Moderate effect sizes (0.5 ≤ d<0.8) were found for resting dyspnea and the mMRC-dyspnea scale. Cohen’s d>0.4 was found for vital capacity, forced expiratory volume in one second, partial oxygen pressure, and anxiety. Signicant but rather small effect sizes (0.2<d<0.4) were found for cough, sputum, pain, and other lung function parameters. 16.0% of group B, and 12.5% of group C, and a moderate improvement (+ 4 to + 5 points) was reported by 46.7%, 52.0% and 37.6% of groups A, B, and C, respectively. A weak but signicant correlation was found between the GROC values for the duration of treatment and the reduction of dyspnea on exertion (r = 0.259, p = 0.019), the changes in mMRC values (r = 0.252, p = 0.029), and the increase in the 6MWD (r = 0.323, p = 0.003). A stronger correlation was found between the GROC values for the duration of treatment and the reduction in fatigue (r = 0.429, p < 0.001).

1. Physical training (O) was based on the recommended exercise program modalities by the German Respiratory Society [17] and consisted of two obligatory main components: a) endurance training was scheduled as 3-4 supervised units per week for 30-60 minutes each time if this was tolerated by the patient; otherwise, the duration and frequency were adjusted individually. Exercise intensity was controlled by a pulse oximeter, modi ed 0-10 BORG scale [21], and heart rate. b) Strength training was scheduled for 2-3 supervised sessions per week of 45-60 minutes each if tolerated; otherwise, the duration and frequency were adjusted individually. Whole-body vibration training (F) was performed 7 times per week and was used only if there was no clinical evidence of thromboembolic complications and if no elevated D-dimer levels were identi ed. Inspiratory muscle training (F) was provided for patients with inspiratory muscle weakness and was scheduled for 7 sessions per week for 21 minutes each, of which 1 session per week was supervised.
2. Respiratory physiotherapy (O) consisted of 2 units of group respiratory physiotherapy per week for 45 minutes each session. If necessary, patients also received the following optional components: (F) a) individual breathing training by physiotherapists, (b) physiotherapy seminar on coughing techniques, and (c) mucolytic inhalation therapies (e.g., NaCl inhalation).
3. General physiotherapy (F), e.g., physiotherapeutic pain management, hand and foot function training, gait training, or fascia training. 4. Patient information about COVID-19 (O) in the form of a 45-minute presentation by a respiratory physician. 5. Routine medical diagnostics (O): Medical admission and discharge examination, comprehensive lung function and laboratory diagnostics, blood gas analysis, 6-minute walk test [22] with measurement of oxygen saturation, and cardiological function diagnostics (electrocardiogram obligatory, echocardiography if indicated). If necessary, further imaging examinations; psychological, psychiatric, or orthopedic examinations; or other specialist consultations were carried out.
6. Close medical supervision (O): All patients received regular medical visits, during which all therapy components, including drug therapy, were reviewed, and adjusted if necessary. Oxygen therapy at rest and especially during exertion, e.g., as part of exercise therapy, was available. Noninvasive ventilation could be continued during PR and monitored accordingly. 7. Psychosocial support, in the form of a psychologically guided self-help group on COVID-19 (O). Social counseling (F) and/or individual clinical psychological counseling (F) were offered if necessary. Neuropsychological diagnostics of cognitive performance were conducted if patients showed signs of cognitive impairment. Furthermore, relaxation seminars (F), such as progressive muscle relaxation or autogenic training, were offered. 8. Nutritional counseling (F) was offered, for example, for patients with overweight, diabetes, or other comorbid disorders. 9. Occupational therapy (F), e.g., prescription and consultation regarding necessary aids, memory training, or training regarding activities of daily life.

Primary outcome
The primary outcome dyspnea was assessed based on the current dyspnea sensation and dyspnea in the last 7 days. On the one hand, we used an 11-point numeric rating scale [23] (NRS) to assess the present severity of dyspnea sensation at rest and on exertion. The scales have two dimensions (symptom intensity and symptom unpleasantness) and range from 0 ("no symptoms") to 10 ("worst imaginable symptom severity"). The minimal clinically important difference is considered to be 1 to 2 points [24,25].
On the other hand, the modi ed Medical Research Council (mMRC) dyspnea scale [26][27][28] was used to rate the impact of dyspnea on daily activities over the last week on a 5-point scale, ranging from 0 ("I only get breathless with strenuous exercise") to 4 ("I am too breathless to leave the house or I am breathless when dressing"). The mMRC is considered to have low sensitivity to change, and a reduction of 1 point can be considered clinically relevant [29].

Assessment of secondary outcomes and further variables
Physical capacity The 6-minute walking distance (6MWD) was measured using a track length of 30 m according to the European Respiratory Society's and American Thoracic Society's technical standards [22]. The deviations from the healthy reference values were calculated according to Enright and Sherrill [30].
Lung function tests, blood gases, and laboratory blood tests Forced expiratory volume in one second (FEV1), vital capacity (VC), residual volume (RV), total lung capacity (TLC), total speci c airway resistance (sRtot), maximal inspiratory pressure (PImax), and the single-breath transfer factor of the lung for carbon monoxide (TLCO) were determined using spirometry and body plethysmography (MasterScreen Body, CareFusion, Hoechberg, Germany) and MasterScreen Diffusion System, Jaeger-Viasys, CareFusion), and respiratory muscle testing was performed using a Jaeger mouth occlusion pressure device in accordance with recommendations of the national guidelines [31,32]. Capillary blood gas samples to assess the partial oxygen pressure (PaO2) and partial carbon dioxide pressure (PaCO2) were taken at rest while breathing ambient air (ABL 800, Radiometer, Willich, Germany). Laboratory chemistry tests, including D-dimers, brain natriuretic peptide (BNP), lactate dehydrogenase (LDH), and C-reactive protein (CRP), were performed by a certi ed external laboratory (SYNLAB Holding, Augsburg, Germany).

Cardinal symptom
The patients were asked to name their "most important symptom". To ease the process, we presented a list of symptoms that are frequently associated with COVID-19. Further symptoms could be speci ed by the patients.

Fatigue
The German version of the Brief Fatigue Inventory (BFI [33]) was used to assess fatigue. This questionnaire assesses the severity of and impairment from fatigue with ten questions. The 9 subscales and the total score range from 0 to 10, with higher values indicating higher severity/impairment. Scores of 4 to 6 correspond to "moderate fatigue", and scores of 7 to 10 correspond to "severe fatigue" [34].

Quality of life
Generic health-related quality of life (QoL) was measured using the 5-level EuroQol questionnaire (EQ-5D-5L) and the EuroQol visual analog scale (VAS) [35]. The descriptive system of the EQ-5D-5L comprises ve dimensions (mobility, self-care, usual activities, pain/discomfort, and anxiety/depression). Each dimension has ve levels: "no problems", "slight problems", "moderate problems", "severe problems", and "extreme problems". The VAS is an instrument to describe patients' current subjectively experienced QoL on a scale ranging from 0 ("worst possible health condition") to 100 ("best possible health condition"). The minimal clinically important difference in chronic obstructive pulmonary (COPD) patients is 8 points [36].

Depression and anxiety
The Patient Health Questionnaire-9 (PHQ-9) [37] and the General Anxiety Disorder-7 (GAD-7) [38] of the German version of the Patient Health Questionnaire [39] were used as screening tools to assess symptoms of depression and anxiety. All items are scored on a 4-point Likert scale (0 = "not at all", 1 = "several days"; 2 = "more than half of the days"; 3 = "nearly all days"). In addition to using the sum of scores (ranging from 0 to 27 for the PHQ-9 and 0 to 21 for the GAD-7), we classi ed individuals with values of ≥ 10 as being at risk for a clinically relevant depressive disorder (PHQ-9) or a clinically relevant anxiety disorder (GAD-7).
Estimation of the overall effectiveness of rehabilitation from the patient's perspective At T 2 , patients were asked to rate the effectiveness of the PR on an 11-point Likert scale, ranging from 0 ("the rehabilitation was ineffective for me") to 10 ("the rehabilitation was highly effective for me").

Global rating of change in subjective health
The global rating of chance in subjective health (GROC) [40] was assessed using one item that compares current subjective health with subjective health at the beginning of inpatient rehabilitation and one item that compares current subjective health to subjective health at the time before SARS-CoV-2 infection.

Statistical analyses
Statistical analyses were performed using IBM SPSS (V26) and R (V3.6.3). For the analyses of differences between the subsamples at baseline, we used independent-samples t-tests for continuous parameters and chi-square tests for categorical variables. For the analyses of changes in the outcomes between T 1 and T 2 , we used repeated-measures analysis of variance (RM-ANOVA). The evaluations were performed pairwise only so that a value was calculated only if both values (T1 and T2) were available. To assess within-group effect sizes, Cohen's d [42] was calculated by dividing the differences in means between T 2 and T 1 by the standard deviation of the change scores. Values of 0.2, 0.5, and 0.8 are considered small, moderate, and large, respectively. We calculated 95% con dence intervals for within-group mean differences as well as for Cohen's d.
Predictors of therapy success were calculated using linear regression analyses. For these analyses, we calculated the changes in clinically relevant variables according to Δ = M T2 -M T1 . Intercorrelations between the deltas were calculated using Pearson correlations for parametric variables and Spearman correlations for nonparametric variables.

Sample
Between April 28th, 2020, and January 08th, 2021, 120 patients were enrolled in PR due to COVID-19. One patient was excluded from the study because of cognitive impairment, ve patients were excluded because of insu cient German language skills, and six patients did not agree to participate in the study. The remaining 108 patients, consisting of 49 women (mean age: 53.67 ± 10.67 years; mean body mass index (BMI): 30.74 ± 6.64) and 59 men (mean age: 57.20 ± 9.21 years; mean BMI: 29.70 ± 5.54), gave their informed consent and were enrolled. Further baseline descriptive data are displayed in Table 1. One patient dropped out because he withdrew his consent for further study participation. Two patients dropped out because of medical conditions that were not related to PR. The rest of the sample completed the rehabilitation program without complications. The mean treatment duration of PR was 26.3 ± 5.9 days (range: 5-42 days). The initial acute COVID-19 diagnosis was based on PCR diagnostics in all but one case, for whom the diagnosis was based on antibody detection. All patients were PCR negative upon admission to the clinic.
Comorbidities were common: Only 2 of the 108 patients (1.9%) had no comorbidities. A total of 55.6% of patients had at least one cardiovascular disease (see Fig. 1), most frequently hypertension (52.8%). A total of 31.5% of patients had at least one upper or lower airway disease; among them, 2 patients (1.85%) had COPD, and 18 (16.7%) had asthma. Only 1 patient was a current smoker, 63 (58.3%) had never smoked and 44 (40.7%) were ex-smokers.

Subsamples
The German health care system provides an administrative referral path to rehabilitation. Therefore, we clustered the patients according to the mode of admission into three groups (see Table 1). The rst group (group A, 'acute severe') consisted of patients who were admitted to PR either directly after hospital discharge or within the rst month after discharge from the hospital. This "follow-up rehabilitation" initiated by hospital physicians while the patient is still in the hospital is an integral part of the German rehabilitation system. These patients were still heavily burdened and still in the process of recovering. The second group (group B, 'severe after interval') consisted of patients who had been hospitalized for at least 3 days but were admitted to the PR after more than one month following discharge from the hospital. These rehabilitations were usually not initiated by the hospital but rather through family physicians or specialists in private practice. The last group (group C 'mild after interval') included patients who had been treated in an outpatient setting or who had been monitored for a maximum of one night in a hospital. These rehabilitations were initiated by physicians in private practice.
Sample characteristics are presented in Table 1. Patients in group C were more likely to be younger (p = 0.045) and female (p = 0.001) and to require less often oxygen therapy during the acute phase (p < 0.001). There were no signi cant differences between the three subgroups regarding BMI (p = 0.265). Differences regarding outcome measures are displayed in the respective sections below.
Cardinal symptoms at the beginning of PR The most prevalent symptom was dyspnea on exertion. Fifty percent of the whole sample named this symptom the most burdensome. This nding was consistent in all groups. Descriptively, however, there were differences: In group A, 56% reported dyspnea on exertion as the most important symptom; in group B, 50% reported dyspnea on exertion as the most important symptom; and in group C, 40% reported dyspnea on exertion as the most important symptom. Dyspnea at rest, on the other hand, was reported in less than 1% of patients. Other symptoms that were frequently mentioned were sadness regarding health status (16.7%), faintness/lack of energy (15.7%), and pain in various parts of the body (7.8%). Chi-square tests of independence did not nd any signi cant difference between the three subgroups (p = 0.298 to p = 0.918).

Primary outcome: dyspnea
The quantitative description of the intensity of dyspnea, as measured by the NRS and the mMRC, is displayed in Table 2, while the NRS results concerning unpleasantness are displayed in Table S1 of the supplement. A graphical illustration of the signi cant improvements in clinically relevant variables is displayed in Fig. 2.  Exertional dyspnea was signi cantly more pronounced than resting dyspnea at baseline in all groups. In the overall group, moderate to large pre-post changes were observed for intensity in exertional dyspnea. A clinically relevant improvement of at least 2 points (MCID) was found in 66.1% of patients. 71.5% of the 50% of patients who reported NRS scores of > 6 (median) at T 1 exhibited clinically relevant improvements at T 2 . There were small to moderate correlations between the improvement in exertional dyspnea and improvements in QoL, fatigue, and anxiety (see Table 3).  For the mMRC scores, we also found moderate pre-post changes. Slightly more than 50% bene ted from a clinically relevant improvement (MCID = 1 level) without differences between the three subgroups. In total, 59% of the 50% of patients who were most severely affected (mMRC > 2) bene ted. There were small to moderate correlations between an improvement in mMRC scores and improvements in QoL (as measured by the VAS), fatigue, and anxiety (see Table 3).
Small to moderate changes were observed in the intensity of dyspnea at rest, with small correlations between an improvement in resting dyspnea and improvements in 6MWD and anxiety. We found no signi cant differences between the 3 subgroups in this regard, either at the start of PR or in the pre-post difference after PR. Moderate changes were also found for unpleasantness of dyspnea at rest or on exertion, which was slightly more pronounced at rest in group C (see Table S1 of the supplemental material).

Secondary outcomes: objective parameters
Physical capacity      Table S2 of the supplemental material).
Secondary outcomes: Patient-reported outcomes NRS cough, sputum, and pain Patients reported only low levels of cough and phlegm at T 1, with small to moderate changes in all groups (see Table S1 of the supplemental material). The experience of pain was most pronounced in group C, with small reductions in the overall group.

Fatigue
A signi cant decrease in fatigue was observed in groups A and B, with large effect sizes. Fatigue was more pronounced in group C at T 1 , and the decrease over the course of PR did not reach statistical signi cance. To illustrate the ndings regarding the differences in fatigue in the 3 subgroups, the relative frequencies of fatigue severity are displayed in Fig. 3.
The overall decrease in fatigue was signi cantly intercorrelated with decreases in dyspnea, as measured by both the NRS and the mMRC. Furthermore, signi cant correlations were found for decreases in depression, anxiety, and impairments in QoL.

Depression and anxiety
At T 1 , the mean values of the PHQ-9 were just below the cutoff between mild and moderate depressive symptom severity (10 points Table 3).

Rating of the overall effectiveness of PR from the patient's perspective and global rating of change in subjective health
Patients rated the effectiveness of PR as high, ranging from 7.07 in group C to 8.93 in group A (see Table 6). Signi cant differences could be detected between group A and group C, with group A reporting higher effectiveness of PR.

Safety and feasibility of PR in patients after COVID-19
During PR, complications were recorded in a systematic and standardized way. Of the 108 patients, one patient had to be transferred to a psychiatric hospital due to an acute psychosocial crisis that was not related to PR. Another patient had to quit PR because of disc prolapse. Another patient terminated study participation prematurely because he withdrew his consent for further study participation. All other 105 patients participated in at least 90% of the physicianprescribed therapies without related side effects.
Prediction of successful PR concerning the primary outcome of dyspnea  Table 3, the decrease in exertional dyspnea was also not correlated with an improvement in the 6MWD or improvements in parameters of the lung function tests or the laboratory measures.
rather small effect sizes (0.2 < d < 0.4) were found for cough, sputum, pain, other lung function parameters (TLC, TLCO, PI max), and results of the laboratory blood tests (D-dimers, CRP).
There was a high burden of comorbidities. In only 2 cases, no secondary diagnoses were mentioned. This gure is higher than what is reported in the literature for the overall group of COVID-19 patients [44,45] but is in line with cross-sectional studies in post-COVID-19 rehabilitation settings [46][47][48]. This difference could be explained by the fact that pre-existing comorbidities negatively in uence the course of COVID-19, and patients with a more severe course are more likely to undergo post-COVID-19 rehabilitation. The most frequently mentioned comorbidities were cardiovascular diseases and obesity. This nding is in line with existing data [45]. In addition, orthopedic and neuromuscular comorbidities were frequent, and some ICU-associated, psychological, and internal diseases were observed. These comorbidities require the multimodal, multiprofessional, and interdisciplinary approach of PR. Previous studies regarding rehabilitation following COVID-19 have reported similar frequencies of comorbidities [48][49][50][51]. This nding supports the necessity of a multimodal and interdisciplinary therapeutic approach for post-COVID-19 rehabilitation.

Effectiveness of PR
Comparing the three subgroups, for some outcomes, stronger effects were found in groups A and B. For example, exertional dyspnea improved in groups A and B, with large effect sizes (d = 0.922 and d = 0.845, respectively). Group C (mild after interval) also improved signi cantly but with a rather low to moderate effect size (d = 0.485). Similar trends were found for some lung function parameters (VC, FEV1, TLC, PaO2). Furthermore, patients in group C rated both the e cacy of the PR and the personal improvement lower than the other two subgroups. In conclusion, these results indicate that patients with severe forms of COVID-19 show greater improvements over the course of PR, especially if the beginning of treatment occurs soon after the acute phase. These results are in line with the ndings obtained by Al Chikhanie et al. [52]. However, it is important to emphasize that signi cant improvements were also found in group C, with some strong (6MWD, EQ-5D-5L-VAS) and moderate effect sizes (PHQ-9). Cohen's d values > 0.4 were also found in group C regarding the intensity of exertional dyspnea, dyspnea in daily life, and impairments in QoL. This indicates that despite a lower effect size, patients with an initially mild course of the disease still bene t from PR even after a long duration, as described by Glöckl et al.

Primary outcome: dyspnea
Exertional dyspnea was mentioned as the "most important symptom" by the majority of patients in all three groups (groups A, B, and C in 65%, 50%, and 40%, respectively). Regarding the intensity of exertional dyspnea, as measured by NRS, all 3 subgroups bene ted signi cantly, with large (groups A and B) or moderate (group C) effect sizes. These improvements were in the same range as those in our recent study on PR for asthma patients [53]. However, the decrease in NRS scores was not associated with improvements in lung function or 6MWD, but there were signi cant correlations with the reduction in fatigue and especially anxiety (r = 0.348). Possibly, the self-con dence in one's own performance regained through training during PR and the resulting reduction in the fear of exertion contributes to a reduction in exertional dyspnea.
Although most lung function parameters (VC, TLC, FEV1, TLCO, PaO2) in group C were normal on average and were higher than those in groups A and B, group C scored their resting dyspnea higher at the start of PR and still scored higher at the end of PR. The more pronounced fatigue and the tendentially lower training effects of group C might have had an in uence in this regard.
Regarding the mMRC dyspnea scale, which records shortness of breath in everyday life in the last week, there was a signi cant improvement, with a moderate effect size. This is worth noting because the mMRC dyspnea scale is considered to have low sensitivity to change. The improvement in the mMRC dyspnea scale is descriptively above the mean improvement seen in a large study on PR for COPD patients [53]. However, similar to the reduction in NRS values, the decrease in mMRC scores was not correlated with an increase in lung function but with a decrease in fatigue and anxiety.
Very few post-COVID-19 rehabilitation studies have reported results on dyspnea scores. A dyspnea assessment using the mMRC dyspnea scale was reported for a subgroup in the study by Glöckl et al. [43] [57,60]. The improvement in patients in group C, who started PR a mean of 143 days after the acute phase and whose pulmonary function parameters were normal except for a reduced PImax value, was well above the MCID of patients with COPD [61-63] or idiopathic pulmonary brosis [64]. There were no signi cant correlations between the improvement in the 6MWD and lung function parameters except for a correlation with the improvement in PImax. Thus, we assume that the improved exercise capacity might not be a consequence of improved lung function but rather the result of adaptive mechanisms of the cardiovascular system and the musculature due to exercise training.

Lung function tests
The lung function pattern, which was more restrictive at baseline, improved in the overall group with a moderate effect size, with signi cant improvements in groups A and B. Similar lung function improvements were also seen in some other post-COVID-19 PR studies [43,46,52]. Maximal inspiratory pressure also improved in the total group, with a small effect size, with the most severe limitation and the smallest nonsigni cant improvement seen in group C at T 1 .
However, the changes in VC, TLC, and FEV1 did not correlate with improvements in exertional dyspnea or physical capacity.
Quality of life, fatigue, depression, and anxiety In line with previous research [7,10], we detected not only physical but also mental impairments and subjectively experienced impairments in health-related QoL in the sample at T 1 . In all subgroups, patients reported impairments in health-related QoL and symptoms of fatigue, depression, and anxiety. Over the course of PR, these parameters improved signi cantly, with mostly large effect sizes. Patients reported lower levels of impairment in health-related QoL and a decrease in fatigue, depression, and anxiety. In comparison to our own results for depression and anxiety in rehabilitants with asthma or COPD at the beginning and at the end of PR [60, 65], the results can be considered comparable. Interestingly, both measures of QoL were signi cantly correlated with a reduction in fatigue. Furthermore, the increase in the EQ-5D-5L-VAS showed a moderate correlation with the reduction in anxiety. However, there were no signi cant correlations between either of the two measures of QoL and the decrease in depression. Considering these numbers and the fact that one-third of the patients reported either sadness regarding their own health status or faintness/lack of energy as their most burdening symptom at T 1 , we assume that fatigue and anxiety are of particular importance in the subjectively experienced burden of disease regarding COVID-19. Noticeably, improvements in fatigue, anxiety, and QoL were the only variables that showed statistically signi cant correlations with reductions in exertional dyspnea. Even though these results align with each other, it is currently hard to fully interpret the ndings because of a lack of studies that examine the role of anxiety and/or fatigue in PR following COVID-19. Therefore, we highly recommend a closer examination in this regard to further explain our ndings.
Regarding the overall subjectively experienced current health status, our data revealed a signi cant improvement, with large effect sizes for all groups. Threequarters of the patients achieved an improvement in the VAS score above the MCID of 8 points.
Given the signi cant reductions in depression and particularly anxiety in all the subgroups, we endorse the conclusion of Demeco et al.
[66], who recommended post-COVID-19 rehabilitation programs not only for physical reasons but also for psychological reasons.
Regarding fatigue, we found a signi cant difference between group C and groups A and B. Group C reported the highest values of fatigue at T 1 and achieved no signi cant changes over the course of PR. These results suggest that patients with initially rather mild courses of disease and leading fatigue symptoms may need more speci c therapeutic approaches for the treatment of persisting fatigue symptoms. It is, however, important to point out the small sample size of 15 patients with fatigue scores for both measurements. Therefore, these results must be interpreted with adequate caution.
Safety and feasibility of PR in patients after COVID-19.
The rehabilitation program proved to be safe; in particular, no complications occurred, and all but two patients were able to complete PR. Moreover, the program proved to be feasible, since all patients were able to perform more than 90% of the prescribed therapies.

Limitations
There are some limitations that we must point out. First and foremost, the data were from an observational study rather than a randomized controlled trial.
Therefore, we cannot state with certainty to what extent rehabilitation caused the improvements. Factors such as the natural course of recovery or regression toward the mean may also have had an in uence on the data. Consequently, our results should be interpreted with caution. Furthermore, because of the sample size, some effect sizes may not be precisely estimated. This was most prominently observable in the analyses of group C, which only included 21 patients. Therefore, the analyses of this group in particular should be generalized with caution.
It is important to remark that most of the patients assigned to the clinic can be characterized as having an interest in PR and being compliant with treatment recommendations. Furthermore, patients with a positive prognosis for rehabilitation are in general more likely to be admitted to PR. Therefore, the results cannot be easily extrapolated to the total group of all post-COVID-19 patients.
The study relies on data from one clinic only. Although another recently published study of German PR with a comparable patient population found comparable results [43] further research is needed to validate our ndings. Considering the aforementioned limitations, randomized trials are needed.

Conclusions
In summary, PR seems to be safe, feasible, and effective in patients after acute COVID-19, improving clinically relevant outcomes. This was true for all three groups analyzed, with a trend toward greater e cacy after 'severe courses' of COVID-19 and an earlier start of rehabilitation after the acute phase of the disease. However, it must be emphasized that signi cant and clinically relevant effects were also seen after longer intervals of latency following milder courses. Therefore, adequate PR should be recommended for all post-COVID-19 patients who remain symptomatic.  Changes in clinically relevant outcomes over the course of the rehabilitation (absolute value of Cohen's d and 95% con dence interval) Notes: 6MWD: 6-minute walking distance; BFI: Brief Fatigue Inventory; CRP: C-reactive protein; EQ-5D: 5-level EQ-5D questionnaire; FEV1: forced expiratory volume in one second: GAD-7: Generalized Anxiety Disorder-7; mMRC: modi ed Medical Research Council; NRS1: Numeric Rating Scale for resting dyspnea; NRS2: Numeric Rating Scale for exertional dyspnea; PaO2: partial pressure of O2; PHQ-9: Patient Health Questionnaire 9; PImax: maximal inspiratory pressure; TLC: total lung capacity; TLCO: diffusion capacity of the lungs for carbon monoxide; VAS: visual analog scale; VC: vital capacity