Introduction

The Severe Acute Respiratory Syndrome Coronavirus 2 (SARS-CoV-2) has had a profound and unprecedented global impact. Wuhan, a city in China, became the origin of this global pandemic that began in 2019 and then became widespread throughout the world [1]. According to WHO, 778 million people have been affected by this virus as of early 2026 [2]. To this day, seven million lives have been lost because of this virus [2]. Notably, people who survived the disease are still not able to evade the consequences of SARS-CoV-2. According to an estimate, nearly 10–15% of people who suffered from the disease experienced a condition known as post-COVID syndrome or long-haul COVID [3, 4]. This syndrome is a collection of complications affecting multiple organ systems including nervous system (ageusia, headaches), cardiovascular (myocardial hypertrophy, coronary artery disease, myocardial infarction), respiratory (respiratory failure, pulmonary embolism, pneumonia), immune (Kawasaki, rheumatoid arthritis), and hematological system (vascular hemostasis, blood coagulation) [5].

Even though the pandemic has been in the past for more than two years, post-COVID complications are still considered a burden to the healthcare industry [6]. In addition to causing psychological distress, this condition results in the substantial loss of economic and healthcare resources, including significant healthcare expenditure and reduced labor productivity [6,7,8]. COVID-19 pandemic can lead to a secondary mental illness paradigm [8]. The mental health of people after COVID-19 may be impacted by fear of infection, worry, rage, post-traumatic stress disorder, stigma, avoidant behaviors, boredom, and frustration [8]. Limited socialization, isolation issues, and other psychosocial and economic stressors are exacerbated by hypochondriac concerns [6]. In the six months following COVID-19 infection, significant neuropsychiatric morbidities have been found recently [9]. These risks were higher in, but not limited to, patients who had severe COVID-19 [10, 11].

One year following discharge, survivors of critical illness exhibit persistent psychological impairment, including considerable levels of anxiety, depression, and post-traumatic stress disorder (PTSD) [7]. The majority of patients with severe acute respiratory distress syndrome (ARDS) had neurocognitive impairment at one year, which included problems with attention, concentration, memory, and mental processing speed [8]. When compared to admissions for other causes, patients with severe ARDS and extended mechanical ventilation had significantly lower quality of life after being discharged from the critical care unit [8].

Nearly half of people with post-COVID conditions experience an improvement in their symptoms 7 to 15 months following testing [10]. However, 19.3% of infected persons have chronification of symptoms, which is defined as the persistent presence of symptoms at each follow-up and beyond 12 months following the infection [10]. The underlying mechanisms of post-COVID are not yet understood [10]. The four main theories for explaining the conditions that follow COVID-19 are viral persistence, dysbiosis, an autoimmune reaction, or a dysregulated inflammatory response [10]. All post-COVID complications require extensive assessment and investigations, which pose a significant financial burden on the global economy [10, 11]. The crisis of such magnitude requires efficient and effective strategies to minimize the economic as well as health burden.

The post-COVID-19 condition also imposes a significant global economic strain, with annual direct and indirect costs, arising from healthcare expenditures and lost labor productivity, estimated to range from US$140 to US$600 billion [11]. These substantial implications underscore the urgent requirement for scalable and cost-effective rehabilitation plans. Beyond the fiscal impact, the clinical burden is characterized by persistent physical impairments that severely limit functional independence. Patients frequently present with respiratory fatigue, reduced VO2 max, and musculoskeletal issues such as muscle atrophy and generalized weakness [12, 13]. These physical deficits, often coupled with persistent dyspnea, necessitate long-term, structured rehabilitative support to restore quality of life.

The healthcare industry has effectively addressed the global health crisis by prioritizing telemedicine (TM), specifically telerehabilitation (TR), to circumvent the limitations of physical distancing and in-person care [14]. Supported by recent advancements in wearable technology, virtual reality (VR), and e-health platforms, these digital interventions have proven to be cost-effective, safe, and highly accessible alternatives to traditional rehabilitation, particularly within cardiac and pulmonary specialties [15, 16]. Research indicates that integrating TR into both acute and long-term recovery, especially for post-COVID-19 symptoms, not only yields significant physical and financial benefits for patients but also maintains high levels of clinical satisfaction and safety [9].

Therefore, this study aims to update the previous version of the review by Yang et al. (2024) [17]. Given that the underlying mechanisms of post-COVID, such as viral persistence or autoimmune reactions, are not yet fully understood and new clinical evidence has emerged especially after 2024 in form of new randomized controlled trials (RCT’s), an updated evaluation is necessary. Therefore, this study aims to evaluate the effectiveness of TR in improving physical and psychological outcomes.

Methods

Study design

A systematic review and meta‐analysis was conducted in accordance with the Preferred Reporting Items for Systematic Reviews and Meta‐Analysis (PRISMA) guidelines [18]. The prescribed procedures for this study have been formally documented and registered in the PROSPERO (International Prospective Register of Systematic Reviews) database, with the protocol number designated as CRD420251082578. No protocol amendments have been made afterwards. This manuscript serves as an updated systematic review of the published version of the prior meta-analysis conducted by Yang et al. (2024) [17]. An ethical approval statement was not required as all the information was extracted from previously published Trials.

Data sources and search strategy

We extensively searched databases including PubMed, Cochrane, and Scopus for relevant literature. While the previous version of this review (Yang et al., 2024) [17] pooled data from inception through April 2024, our updated search focused on newly published RCT’s from April 2024 till July 2025, which ultimately changed the pooled estimates. Additional trials were searched through ClinicalTrials.gov and Medrxiv.org. The search terms used were ‘COVID- 19’, ‘SARS- CoV- 2’, ‘2019- nCoV’, ‘Telerehabilitation’, ‘Telecommunications’, ‘Telemedicine’ and ‘Random’ (to capture randomized controlled trials) [Supplementary Material]. Where accessible, each database's restricted vocabulary was used in searches along with free-text words, which were tailored to each data library individually. Boolean operators ‘AND’ were used to combine different concept groups, while ‘OR’ was used to include synonyms within the same concept group. The articles published in English were considered. Furthermore, articles from the reference citations were further evaluated to ensure more qualifying results.

Inclusion criteria

(1) Individuals meeting the WHO definition of post-COVID-19 condition were uniformly included without regard to age, gender, or race. (2) Interventions should include telerehabilitation, which is defined as "delivery of therapeutic rehabilitation at a distance or offsite using telecommunication technologies”, which was further categorized by their delivery method [19]. Synchronous methods (e.g., Calvo-Paniagua et al. 2024 [20]; Demir et al. 2025 [21]) involved real-time interaction, while asynchronous interventions (e.g., Carpallo-Porcar et al. 2024 [22]; Esparza et al. 2024 [23]) utilized delayed communication. A larger portion of the literature (e.g., Altay et al. 2025 [24]; Vian et al. 2025 [25]) employed mixed (hybrid) models combining both approaches, with Yasaci et al. (2025) [15] providing a direct comparison between synchronous and asynchronous delivery. (3) Placebo (e.g., educational videos without active exercise), standard care (e.g., routine outpatient follow-up or home exercise leaflets), or waitlist were the control interventions. Regarding comorbidities or medication use during the telerehabilitation treatment, we did not impose any limitations. (4) Outcomes: at least one measure of clinical symptoms, physical function, psychological function, quality of life, or safety (including adverse events) must be included in trials. (5) Study design should be RCT. (6) English-language journal articles were included.

Exclusion criteria

(1) Studies in which telerehabilitation's results are indistinguishable from those of other treatments. (2) Duplicate publications of the same study; in such cases, only the most comprehensive report was included. (3) The original study did not contain any data that could be extracted. (4) Case reports, protocols, reviews, editorials, and conference abstracts were not included.

Study selection and data extraction outcome measures

Two reviewers (Parkash.S. and Sheikh.M.) separately examined the titles and abstracts of the retrieved records in accordance with the aforementioned criteria, excluding trials that were clearly irrelevant before retrieving the entire text of the remaining trials. This updated review process involved the addition of 12 new RCT’s to the existing database established by Yang et al. (2024) [17] which previously included 16 RCT’s. The final included trials were then independently assessed by the two reviewers, who extracted the following information: (1) first author, publication year, sample calculation, and registration number; (2) participant characteristics, including sample size, age, gender, and comorbidities; (3) the intervention and control protocol; (4) interested outcomes; and (5) important components of the risk of bias assessment. Disagreements were settled by conversation or by seeking advice from a third reviewer (Javed.M.N.).

Outcome measures

Clinical symptoms (e.g., Borg score (BS) and Multidimensional Dyspnea-12 questionnaire (MD-12)) and physical function (e.g., 30STST, 6MWD, and pulmonary function) were the main outcomes of interest. Psychological function [such as hospital anxiety and depression scales (HADS)] and quality of life [such as Saint George Respiratory Questionnaire (SGRQ), 12-item short-form health survey (SF-12), and EuroQol-5 Dimension (EQ-5D)] were secondary outcomes of interest. Outcomes were analyzed across standardized populations specifically meeting the WHO definition of post-COVID conditions to minimize clinical heterogeneity. When the telerehabilitation program was finished, the time point for the primary and secondary outcomes was the shortest time point available. Safety outcomes, primarily serious and non-serious adverse events, were not frequently reported in the included trials. However, treatment adherence was reported among the included studies and hence evaluated in this study.

Statistical analysis and quality assessment

All statistical analysis was conducted using RevMan (version 5.4.1; Copenhagen: The Nordic Cochrane Centre, The Cochrane Collaboration, 2020). OpenMeta [Analyst] (version 5.26.14) was used for meta-regression analyses [26]. Dichotomous outcomes were reported as Mantel–Haenszel risk ratios (RR), while continuous data was presented as mean differences (MD) with 95% confidence intervals (CI), using the inverse variance method. For continuous outcomes, we pooled data using the MD with 95% CI. In instances where studies reported change-from-baseline scores instead of final post-intervention values, we pooled them together as per Cochrane guidelines, assuming that the randomization process minimized baseline imbalances. To assess the potential for bias introduced by this approach, we conducted a sensitivity analysis for key outcomes such as 6MWD or Borg score to compare the results of pooled final values versus change scores. A random effects model was applied throughout due to expected variability among studies. Statistical significance was set at a p value of less than 0.05, and findings are illustrated using forest plots. Heterogeneity was evaluated using Higgins’ I2 statistic, with values of 25–50% considered mild, 50–75% moderate, and over 75% as substantial. Further, to assess the robustness of the findings and assess heterogeneity, a leave-one-out sensitivity analysis was conducted [27]. This involved sequentially excluding each study from the meta-analysis to observe its impact on heterogeneity (I2) and the pooled effect estimate.

The risk of bias (ROB) was assessed independently by two authors using the Revised Cochrane ROB tool for randomized trials (ROB 2) [28], which evaluates five domains: bias arising from the randomization process, bias due to deviations from intended interventions, bias due to missing outcome data, bias in measurement of the outcome, and bias in selection of the reported result. Subgroup analyses was also planned and stratified by intervention type, treatment duration, frequency, session length and supervision status. Additionally, the quality of evidence of each outcome was assessed using GRADEpro [29]. Disagreements were resolved through discussion among the research team to ensure consensus. Funnel plot analysis was performed, as the number of studies per outcome precluded a reliable assessment of publication bias.

Results

Search results and baseline characteristics

As outlined in the PRISMA flow diagram (Fig. 1), a systematic search was carried out across different databases including PubMed, Cochrane, and Scopus yielding a total of 1,793 results. After the removal of 856 duplicates, 937 studies were screened based on title and abstract. This initial screening led to the exclusion of 652 records. Of the 285 full-text articles assessed for eligibility, 273 studies were excluded. Consequently, 12 new studies were identified for inclusion. With 16 studies identified in the previous version of this review, a total of 28 RCT’s were added in the systematic review, comprising nearly 1,400 participants across experimental and control groups. Individual study sample sizes ranged from 16 to 109 participants, with participants’ mean ages ranging from early 20 s to mid-60 s years, reflecting a broad adult population with post-COVID-19 conditions. Studies were conducted in countries like Turkey, Spain, Brazil, China, India, Saudi Arabia, the United Kingdom, Germany, Netherlands, Sweden, Ecuador, and the United States.

Fig. 1
Fig. 1
Full size image

Preferred reporting items for systematic review and meta-analysis. (PRISMA) flow diagram of study screening and selection

A duration of 4 to 17 weeks for TR intervention was there and included multimodal components such as breathing exercises, aerobic and resistance training, respiratory muscle training, education, and tele-monitoring, delivered through synchronous or asynchronous digital telecommunicating platforms. Comparison groups were most commonly given usual care, home exercise programs, or no intervention. Key baseline demographic and clinical characteristics, including age, sex distribution, time since COVID-19 infection, disease severity, comorbidities, intervention characteristics, and outcome measures are summarized in Table 1.

Table 1 Baseline characteristics of included studies
Full size table

Clinical symptoms

The impact of telerehabilitation on clinical symptoms was assessed using the Borg score, the Multidimensional Dyspnea-12 (MD-12) questionnaire, the Visual Analog Scale (VAS) for Fatigue, and the Modified British Medical Research Council (mMRC) Dyspnea score.

Borg Dyspnea score

For Borg Dyspnea Score, four studies including 80 participants in each group reported absolute values, indicating no statistically significant difference between the telerehabilitation and control groups (MD = − 3.27, 95% CI: − 7.24 to 0.70; p = 0.11; I2 = 99%). For change values, three studies including 64 participants in each group showed statistically significant reductions in Borg Dyspnea scores favoring telerehabilitation (MD = − 4.35, 95% CI: − 7.79 to − 0.91; p = 0.01; I2 = 98%). For overall effects, pooled analysis represented a major improvement (MD = − 3.74, 95% CI: − 6.16 to − 1.31; p = 0.003; I2 = 99%) with no significant subgroup differences (p = 0.69) (Fig. 2a).

Fig. 2
Fig. 2
Full size image

Forest plot showing the efficacy of telerehabilitation on (a) Borg score (b) Multidimensional Dyspnea-12 questionnaire (c) Visual analog scale score of Fatigue (d) Modified British medical research council Dyspnea score

Multidimensional dyspnea-12 (MD-12) questionnaire

For the MD-12 questionnaire, which provides a multidimensional assessment of dyspnea, two studies including a total of 354 participants in both groups reported absolute and change values. The pooled analysis showed a statistically significant reduction in dyspnea scores favoring the telerehabilitation group compared to control groups (MD = −5.12, 95% CI: − 9.94 to − 0.31; p = 0.04; I2 = 94%). This suggests that the intervention may lead to a major improvement in the sensory and affective dimensions of breathing difficulty (Fig. 2b).

Visual analog scale for fatigue (VASF)

Fatigue levels were evaluated using the Visual Analog Scale across four studies reporting absolute values and across two studies reporting change values. The pooled analysis indicated a statistically significant decrease in self-reported fatigue among participants receiving telerehabilitation (MD = −1.83, 95% CI: − 2.69 to − 0.98; p < 0.0001; I2 = 33%) (Fig. 2c).

Modified British medical research council (mMRC) Dyspnea Score

The mMRC scale was the most frequently reported clinical symptom measure, utilized in five studies, reporting only absolute values. The pooled results demonstrated a statistically significant improvement in functional dyspnea levels for the telerehabilitation group (MD = −0.92, 95% CI: − 1.77 to − 0.06; p = 0.04; I2 = 98%) (Fig. 2d). High heterogeneity (I2 = 98%) was observed, likely reflecting the diverse patient populations ranging from mild post-COVID cases to those discharged from the ICU.

Physical function

The efficacy of telerehabilitation on physical function was evaluated through functional exercise capacity, health-related quality of life, and pulmonary function parameters.

30-s sit-to-stand test (30 s-STS)

For absolute values, eight studies (189 experimental, 172 control participants) reported 30 s-STS outcome. Telerehabilitation significantly improved performance compared to control group (MD = 1.91, 95% CI: 0.89 to 2.94; p = 0.0003; I2 = 72%). Similarly for change values, three studies (92 experimental, 88 control) reported change scores, showing significant improvement (MD = 2.39, 95% CI: 0.79 to 3.99; p = 0.003; I2 = 79%) (Fig. 3a). Furthermore, combined analysis of absolute and change values showed a significant overall improvement (MD = 2.07, 95% CI: 1.24 to 2.90; p < 0.00001; I2 = 76%), with no subgroup differences (p = 0.62).

Fig. 3
Fig. 3
Full size image

Forest plot showing the efficacy of telerehabilitation on (a) 30-s sit-to-stand test (b) 6-min walking distance (c) Physical function score of SF-36 (d) FEV1 (e) FVC (f) FEV1/FVC (%)

6-min walking distance (6MWD)

For absolute values, thirteen studies (276 experimental, 299 control participants) reported 6MWD outcome. Telerehabilitation significantly improved function compared to the control group (MD = 82.48, 95% CI: 61.03 to 103.94; p = 0.00001; I2 = 52%). Similarly for change values, three studies (92 experimental, 100 control) reported change scores, showing significant improvement (MD = 101.80, 95% CI: 50.95 to 152.65; p < 0.0003; I2 = 73%) (Fig. 3b). Pooled analysis of both values showed a significant overall improvement (MD = 85.83, 95% CI: 67.14 to 104.52; p < 0.00001; I2 = 54%), with no subgroup differences (p = 0.49).

Physical function of SF- 36

Only absolute value was reported by five studies (183 experimental, 197 control participants) for physical function of SF-36 (Fig. 3c). Telerehabilitation significantly improved physical function compared to the control group (MD = 6.12, 95% CI: 2.85 to 9.38; p = 0.0002; I2 = 83%).

Forced expiratory volume in one second (FEV1)

For pulmonary function FEV1, seven studies reported absolute values and found no significant difference between both intervention and control groups (MD = 0.16, 95% CI: − 0.01 to 0.33; p = 0.06; I2 = 45%). However, for the change values, three studies (93 experimental, 108 control) reported significant improvements in FEV1 favoring telerehabilitation without any heterogeneity (MD = 0.08, 95% CI: 0.02 to 0.15; p = 0.008; I2 = 0%) (Fig. 3d). For overall effect, pooled analysis showed a significant overall improvement (MD = 0.11, 95% CI: 0.03 to 0.18; p = 0.005; I2 = 26%).

Forced vital capacity (FVC)

Seven studies reported absolute values for FVC and found no significant difference between both groups (MD = 0.18, 95% CI: − 0.01 to 0.37; p = 0.06; I2 = 35%). Similarly, for the change values, no significant difference was observed as reported by only two studies (MD = −0.04, 95% CI: −0.11 to 0.04; p = 0.34; I2 = 0%) (Fig. 3e). Pooled analysis of both values also showed no significant overall improvement (MD = 0.09, 95% CI: −0.04 to 0.22; p = 0.19; I2 = 51%).

FEV1/FVC (%)

The FEV1/FVC (%) ratio analysis for absolute values, derived from five studies, showed non-significant differences between groups (MD = 0.83, 95% CI: − 1.04 to 2.71; p = 0.38; I2 = 13%), suggesting minimal impact on obstructive or restrictive patterns at this follow-up stage (Fig. 3f).

Psychological function

Mixed findings were indicated using the data from studies assessing Hospital Anxiety and Depression Scale (HADS) with some studies showing reductions in anxiety and depression scores, while others reported no significant difference. Four studies reported absolute values of HADS-Anxiety score, while two studies reported change values. Overall for pooled analysis of HADS-Anxiety score, no significant difference was observed between both groups (MD = −0.66, 95% CI: − 1.83 to 0.51; p = 0.27; I2 = 65%) (Fig. 4a). Similarly, four studies reported absolute values of HADS-Depression score, while only one study reported change value. A pooled analysis of HADS-Depression score showed no significant difference between both intervention and control groups (MD = −1.24, 95% CI: − 2.77 to 0.29; p = 0.11; I2 = 78%) (Fig. 4b).

Fig. 4
Fig. 4
Full size image

Forest plot showing the efficacy of telerehabilitation on (a) HADS-Anxiety (b) HADS-Depression

Health-Related Quality of Life (QoL)

Health-Related Quality of Life (QoL) was reported through SGRQ, SF-12, SF-36 and EQ-5D across various studies. A pooled quantitative synthesis was not conducted due to variation in measurements and reporting formats. Individual studies generally suggested improvements in overall health status and QoL in participants receiving telerehabilitation.

Safety and treatment adherence

Few adverse events such as dizziness, mild lipothymia, and symptom exacerbation, were reported by only a small number of studies. However, they were non-serious and infrequent. Some of the trials reported no serious adverse events while most of the studies did not report the occurrence of any adverse event. For this reason, no pooled analysis could be performed. In terms of the overall treatment completion rate (treatment adherence), all 28 studies reported their completion rate for the telerehabilitation treatment. Only five studies reported a completion rate of 100%, while most of the studies reported it around 80%. The overall intervention completion rate was 79.87% (1778 cases), of which 915 cases (76.06%) in the experimental group and 863 cases (84.36%) in the control group, and the difference in the overall treatment completion rate between the two groups was not statistically significant (RR = 0.98, 95% CI: 0.94 to 1.02; p = 0.26; I2 = 60%) (Fig. 5). Among the studies that provided reasons for attrition, dropouts were primarily attributed to dizziness specifically caused by prolonged screen time, unrelated hospitalizations, and various non-serious adverse events. However, majority of the studies did not provide any specific reason for participant withdrawal.

Fig. 5
Fig. 5
Full size image

Forest plot showing the overall treatment completion rate (Treatment adherence)

While safety was not a primary outcome in most included studies, the available data is encouraging. The majority of trials that monitored for complications reported zero serious adverse events, suggesting that the physiological demands of TR were generally well-tolerated. However, given that many studies did not explicitly report on adverse events, these findings should be interpreted with caution.

Quality assessment

The risk of bias analysis of all included studies is demonstrated in supplementary figure (S1). Fifteen out of the 28 included studies were rated low risk across all RoB 2 domains. Six of the studies had some concerns especially in domain 5. While, seven studies were rated high risk in key domains, primarily due to lack of blinding and incomplete outcome reporting. Thus, around 54% of trials were low risk overall, 21% with some concerns and around 25% with significant bias in key domains, warranting caution in interpreting pooled results.

Certainty of evidence

The certainty of evidence for key outcomes was evaluated using the Grading of Recommendations Assessment, Development and Evaluation (GRADE) approach and the results are reported in supplementary figure (S2). Evidence from randomized controlled trials was initially rated as high certainty and subsequently downgraded based on risk of bias, inconsistency, indirectness, imprecision, and publication bias.

For clinical symptoms (Borg Dyspnea Score), the certainty of evidence was rated as low to very low, primarily due to substantial heterogeneity (I2 > 90%), imprecision, and methodological limitations across studies. For physical function outcomes, the certainty of evidence was moderate for the 30s-STS, supported by consistent direction of effect across studies. Evidence for FEV₁ was rated as low to moderate, with downgrading due to inconsistency and imprecision in absolute values, although change scores showed more consistent effects. Similarly for psychological function (HADS) and health-related quality of life (Qol), the certainty of evidence was rated as low, owing to heterogeneity in outcome measures, variability in reporting, and the inability to pool data quantitatively. The certainty of evidence for safety and treatment adherence outcomes was considered moderate, as adverse events were infrequently reported. Detailed reasons for downgrading the certainty of evidence are provided in the grade evidence profile provided in supplementary material (S5).

Summary of findings table for telerehabilitation vs. control in post-COVID-19

Summary of Findings Table | Telerehabilitation vs. Control for Post-COVID-19

Patient or population: Post-COVID-19 Setting: Outpatient/Remote Intervention: Telerehabilitation Comparison: Control (usual care/in-person rehabilitation)

Outcomes

Anticipated Absolute Effects* (95% CI)

Relative Effect (95% CI)

No. of Participants (Studies)

Certainty of Evidence (GRADE)

Comments

 

Risk with Control

Risk with Telerehabilitation

    

mMRC

Mean: 0

(standardized)

MD 0.92 lower

(1.77 lower to 0.06 lower)

—

234

(5 RCTs)

⨁⨁⨁◯

Moderatea

Telerehabilitation probably results in little to no difference in mMRC

Borg Score

Mean: 0

(standardized)

MD 3.74 lower

(6.16 lower to 1.31 lower)

—

288

(4 RCTs)

⨁⨁⨁◯

Moderatea

Telerehabilitation probably reduces Borg Score slightly

MD-12

(Dyspnea)

Mean: 0

(standardized)

MD 5.12 lower

(9.94 lower to 0.31 lower)

—

354

(2 RCTs)

⨁◯◯◯

Very Lowb,c,d

The evidence is very uncertain about the effect of Telerehabilitation on MD-12

VASF

(Fatigue)

Mean: 0

(standardized)

MD 1.83 lower

(2.69 lower to 0.98 lower)

—

288

(4 RCTs)

⨁⨁◯◯

Lowe

Telerehabilitation may reduce VASF

30 s-STS

(Functional Capacity)

Mean: 0

(standardized)

MD 2.07 higher

(1.24 higher to 2.9 higher)

—

541

(8 RCTs)

⨁⨁⨁◯

Moderatef

Telerehabilitation probably increases 30 s-STS slightly

6MWD

(Exercise Capacity)

Mean: 0

(standardized)

MD 85.83 higher

(67.14 higher to 104.52 higher)

—

767

(13 RCTs)

⨁⨁⨁⨁

High

Telerehabilitation results in an increase in 6MWD

SF-36

(Quality of Life)

Mean: 0

(standardized)

MD 6.12 higher

(2.85 higher to 9.38 higher)

—

380

(5 RCTs)

⨁⨁⨁⨁

High

Telerehabilitation results in a slight increase in SF-36

FEV1

(Lung Function)

Mean: 0

(standardized)

MD 0.11 higher

(0.03 higher to 0.18 higher)

—

702

(9 RCTs)

⨁⨁⨁◯

Moderatea

Telerehabilitation probably results in little to no difference in FEV1

FVC

(Lung Function)

Mean: 0

(standardized)

MD 0.09 higher

(0.04 lower to 0.22 higher)

—

640

(8 RCTs)

⨁⨁⨁◯

Moderatea

Telerehabilitation probably results in little to no difference in FVC

FEV1/FVC (%)

(Lung Function)

Mean: 0

(standardized)

MD 0.83 higher

(1.04 lower to 2.71 higher)

—

401

(5 RCTs)

⨁⨁◯◯

Lowg

Telerehabilitation may result in little to no difference in FEV1/FVC(%)

HADS-Anxiety

Mean: 0

(standardized)

MD 0.66 lower

(1.83 lower to 0.51 higher)

—

470

(6 RCTs)

⨁⨁⨁◯

Moderatef

Telerehabilitation probably reduces HADS-Anxiety slightly

HADS-Depression

Mean: 0

(standardized)

MD 1.24 lower

(2.77 lower to 0.29 higher)

—

332

(5 RCTs)

⨁⨁⨁⨁

High

Telerehabilitation reduces HADS-Depression slightly

Treatment Adherence

(Completion Rate)

844 per 1,000

827 per 1,000

(793 to 860)

RR 0.98 (0.94 to 1.02)

2226

(28 RCTs)

⨁⨁◯◯

Lowh,i

Telerehabilitation may result in little to no difference in overall treatment completion rate

  1. GRADE Working Group Grades of Evidence: ⨁⨁⨁⨁ High: Very confident the true effect lies close to the estimate. ⨁⨁⨁◯ Moderate: Moderately confident; the true effect is likely close to the estimate. ⨁⨁◯◯ Low: Limited confidence; the true effect may be substantially different. ⨁◯◯◯ Very Low: Very little confidence; the true effect is likely substantially different from the estimate
  2. CI Confidence interval, MD Mean difference, RR Risk ratio, RCTs Randomized Controlled Trials, mMRC Modified Medical Research Council, 6MWD 6-Minute Walk Distance, 30 s-STS 30-Second Sit-to-Stand, VASF Visual Analogue Scale Fatigue, SF-36 Short Form Health Survey, FEV1 Forced Expiratory Volume in 1 Second, FVC Forced Vital Capacity, HADS Hospital Anxiety and Depression Scale
  3. *The risk in the intervention group (and its 95% CI) is based on the assumed risk in the comparison group and the relative effect of the intervention (and its 95% CI)
  4. aThese subjective scales often have lower total sample sizes across studies, leading to imprecision
  5. bSimilar to treatment adherence, the primary studies often lack blinding. For subjective outcomes like functional scores at 12 months, this significantly increases the risk of detection bias
  6. cThe I2 for this outcome is high (> 50%). The results show significant heterogeneity, which may be due to differences in the types of telerehabilitation provided
  7. dThe pooled result crosses the line of no effect, or the 95% CI is broad enough to include both a clinically important benefit and a clinically important harm
  8. eFatigue scales in telerehabilitation trials often suffer from a lack of blinding (performance bias), as participants know they are receiving the "new" intervention
  9. fDowngraded one level due to moderate-to-high heterogeneity
  10. gResults for ratio measurements in post-COVID populations often show negligible differences between groups with wide confidence intervals that cross the "no effect" line significantly
  11. hReviewing the study-level data, there are concerns regarding performance bias (lack of blinding of participants and personnel), which is common in telerehabilitation trials but affects certainty
  12. iThe total number of events and the overall sample size are relatively small. The 95% CI is wide, indicating that the true effect could range from a minimal to a substantial difference

Sensitivity analysis

Leave-one-out analysis was conducted to evaluate the heterogeneity and robustness of the pooled effect estimates by excluding individual studies from the meta-analyses and to evaluate the impact of pooling absolute post-intervention values with change-from-baseline scores. A very high heterogeneity was observed in Borg, MD-12 and mMRC scores (I2 > 98%). Therefore, we performed a leave-one-out sensitivity analysis. The exclusion of individual trials did not significantly reduce the I2 value, indicating that heterogeneity likely stems from clinical differences in study populations and the varying intensity of telerehabilitation protocols. The direction and magnitude of the overall effects remained largely unchanged following study removal, indicating that no single study disproportionately influenced the pooled results. Minimal variation was shown through heterogeneity estimates across sensitivity analyses. The findings support the stability and reliability of the primary meta-analysis results. The results of the analysis are displayed in Fig. 6.

Fig. 6
Fig. 6
Full size image

Sensitivity analysis; a mMRC, b Borg Score, c 6MWD, d VASF, e 30s-STS, f SF-36, g FEV1, h FVC, i FEV1/FVC(%), j HADS-Anxiety, k HADS-Depression

Publication bias

Due to a very large number of included studies, publication bias was assessed by visual inspection of funnel plots for various outcomes (Fig. 7). Based on the funnel plots for 6MWD and treatment adherence, the distribution of studies generally indicates a low to moderate risk of publication bias across the assessed outcomes. In the outcome 6MWD, studies were distributed on both sides of the mean effect line, though there is a slight visual asymmetry with a lack of smaller, lower-precision studies (higher standard error) on the left side of the plots. Similarly, treatment adherence showed studies clustered tightly at the apex, suggesting consistent reporting. While the concentration of studies at the top of the funnel plots indicated that the meta-analysis is primarily informed by higher-precision evidence, the relative absence of small studies with negative or null results suggested that some publication bias cannot be entirely ruled out.

Fig. 7
Fig. 7
Full size image

Funnel plot showing publication bias in; a 6MWD and b Treatment adherence

Subgroup analysis

Subgroup analyses were conducted for the 30s-STS and 6MWD outcomes to explore potential sources of heterogeneity (Supplementary Table S2). For the 30s-STS, significant improvements were observed across most subgroups, with the strongest effects found in interventions involving a combination of aerobic, resistance, and balance training (MD = 1.48; P < 0.00001) and programs lasting 1–8 weeks. Notably, unsupervised interventions yielded a larger mean difference (MD = 3.56; P < 0.00001) compared to supervised ones, though the latter also showed significant benefits (Supplementary Figure S3).

Regarding the 6MWD, all subgroup categories for intervention type, frequency, and supervision status demonstrated significant improvements (P < 0.0001), with combined exercise protocols (A + B + C) showing consistent efficacy. While programs of 1–8 weeks and those with a frequency of ≤ 3 times per week (MD = 68.87) or > 3 times per week (MD = 102.11) were all effective, interventions extending beyond 8 weeks did not reach statistical significance for this outcome (P = 0.24) (Supplementary Figure S4). Across both outcomes, session lengths exceeding 30 min consistently provided significant functional gains.

Discussion

In this updated systematic review and meta-analysis of TR in patients with post COVID 19, we aimed to evaluate both physical and psychological outcomes given the well documented burden of functional impairment and mental health issues in this population. Our findings demonstrate that TR significantly improves physical function, as evidenced by increases in 30s-STS performance and 6MWD compared with control conditions. Pooled analyses, on the other hand, revealed no appreciable advantages of TR on psychological outcomes (HADS-Depression and HADS-Anxiety). These findings complement earlier research showing that exercise programs administered remotely can improve functional ability in COVID-19 or post-COVID populations [49], but they also highlight the possibility that psychological benefits may necessitate more focused mental health care beyond exercise-based TR.

Our review includes 12 randomized controlled trials published between 2024 and early 2026, including recent high-quality trials by Yasacı et al. [15] and Altay et al. [24]. This provides updated statistical power for both physical and psychological endpoints. Additionally, this is the first post-COVID TR meta-analysis to present a direct comparison of absolute versus change score outcomes and to apply GRADE systematically, delineating where confidence in the evidence is strongest (physical function) and where it remains limited (psychological and dyspnea outcomes). Furthermore, unlike several previous reviews, we strictly adhered to the GRADE approach to evaluate the certainty of evidence for each outcome, offering clinicians a more reliable framework for evidence-based decision-making in post-COVID-19 care. As a result, this study offers the most recent and updated comprehensive synthesis of TR effectiveness in post-COVID populations, with better precision for physical outcomes and a tighter definition of evidence certainty.

Interpretation of main findings

Our meta-analysis demonstrated consistent improvements in functional exercise capacity (e.g., 30 s STS: MD = 2.07; 6MWD: MD = 85.83) in favor of TR. This also supports the idea that well-structured remote rehabilitation, including measures of aerobic, resistant, as well as breathing exercises, could potentially mitigate the effects of deconditioning following a COVID 19 infection. Previous reviews have also documented remarkable improvements in functional capacity measures through TR processes and procedures post COVID-19 infection [17].

These improvements are not only statistically significant, but also therapeutically relevant. The observed increase in 6MWD (85.83 m) surpasses the commonly cited minimal clinically important difference (MCID) of 54 m established for patients with pulmonary diseases [49]. Additionally, the gain in 30 s-STS performance (2.07 repetitions) surpasses the MCID of two repetitions reported for this test after pulmonary rehabilitation [50]. TR provides functional benefits to post-COVID-19 patients, leading to noticeable improvements in daily physical capacity.

Several hypotheses may explain the observed pattern of results. The consistent improvement in physical function is likely attributable to reversal of deconditioning through structured exercise, a well-established effect across patient populations. Most TR programs in the included studies incorporated multi component exercise regimens (aerobic, resistance, breathing), which are known to enhance muscle strength, endurance, and neuromuscular coordination even when delivered remotely. In contrast, the absence of significant psychological benefit may reflect that the majority of interventions were exercise focused and lacked cognitive behavioral components or dedicated psychosocial support. This aligns with evidence that improvements in anxiety and depression typically require explicit psychological strategies rather than indirect effects of physical activity alone. Regarding pulmonary function, the limited changes in spirometric volumes (except FEV₁ change scores) could be explained by the prolonged and heterogeneous recovery of respiratory mechanics after SARS CoV 2 infection. Peripheral muscle function and symptom perception may improve more rapidly than central airway mechanics, and the relatively short intervention durations (4–17 weeks) may be insufficient to alter spirometric parameters in a population with variable baseline lung involvement.

Measures of dyspnea (e.g., Borg, MD 12, mMRC) showed improvement in some but not all pooled analyses, with overall effects favoring TR. However, high heterogeneity in these outcomes (I2 > 90% for several dyspnea measures) and the subjectivity of self-reported scales temper confidence in these findings. These results broadly mirror earlier findings where TR demonstrated reductions in perceived exertion and dyspnea but varied by outcome measure and scale used [51].

The extremely high heterogeneity (up to 99% for some dyspnea outcomes) warrants particular attention. While we used a random effects model to account for between study variance, we also performed subgroup analyses to explore the potential sources of heterogeneity. Sufficient data was reported across the included trials specifically for the outcome 30 s-STS and 6MWD. Subgroups were stratified by intervention type (respiratory training [A], aerobic training [B] and resistance training [C]), treatment duration (1–4 weeks, 5–8 weeks and 9–12 weeks), frequency (≤ 3 times a week or > 3 times a week), session length (≤ 30 min per session, > 30 min per session or unclear) and supervision status (supervised or unsupervised). These factors are likely to influence dyspnea response to telerehabilitation. Consequently, the pooled estimates for dyspnea should be interpreted with considerable caution, as they represent an average across highly diverse populations and intervention protocols. Future trials should also report outcomes stratified by these clinically relevant subgroups to enable more informative syntheses.

In our meta-analysis, pulmonary function measures showed limited improvements overall, with significant effects observed only in change scores for FEV₁, whereas absolute values and other spirometric indices (e.g., FVC, FEV₁/FVC) did not differ significantly between the TR and control groups. This pattern may reflect the complex and prolonged recovery of respiratory mechanics after SARS CoV 2 infection, where conventional spirometric volumes may be less responsive to short term remote interventions compared with measures of functional capacity, or where spontaneous physiological recovery confounds intervention effects. Natural recovery of lung function following COVID 19 occurs predominantly during the first year after infection, with improvements of 3.1% for FEV₁ and 4.4% for FVC (as percentage of predicted values) between 3 and 12 months [16]. The modest changes observed in our pooled analysis may be partially attributable to this underlying recovery trajectory rather than the specific effect of TR. This highlights the need for controlled designs that adequately account for spontaneous physiological improvement when evaluating pulmonary outcomes. Consistent with this, several systematic reviews and meta-analyses of TR and pulmonary rehabilitation in post COVID 19 populations have reported modest or non-significant changes in traditional pulmonary function tests, even when improvements in dyspnea, respiratory muscle strength, and functional capacity are evident [9]. Overall, these findings suggest that TR primarily improves functional and symptomatic outcomes rather than directly altering spirometric parameters.

For example, pooled analyses from broader TR literature indicate non-significant effects on FEV, yet significant reductions in dyspnea and gains in respiratory and lower limb muscle strength, exercise capacity, and related performance outcomes [52]. The high variability in pulmonary outcomes underlines the challenge presented by the use of spirometry as an exclusive tool in capturing the post-COVID-19 multidimensional respiratory sequelae and might suggest that the effects of TR encompass an area of respiratory health not entirely covered by routine volumes, possibly through neuromuscular coordination gains, breathing pattern efficiency, or muscle endurance.

As a considerable difference from the significant improvements found in terms of physical functionality, pooled data related to psychological benefits found in our meta-analysis (HADS-Anxiety, HADS-Depression) were unable to identify significant benefits of TR when compared with other approaches. This finding aligns with broader evidence suggesting that the effects of TR on mental health symptoms are potentially limited when interventions are not supported with specific psychological measures, including cognitive behavior therapy or specific psychosocial support interventions. For example, the largest recent meta-analysis reported non-significant pooled effects for HADS anxiety and HADS depression outcomes, with low overall quality of evidence [17]. The minimal important difference (MID) for HADS subscales is approximately 1.5 points (corresponding to a change of about 20% from baseline) in populations with chronic respiratory disease [53]. Since our pooled estimates did not reach statistical significance, they are also below this clinically meaningful threshold, reinforcing the conclusion that exercise only TR is insufficient to improve psychological outcomes in this population.

Differences in study populations, TR modalities, timing of intervention, and outcome measurement tools likely contribute to inconsistent findings across trials and reviews, as documented in the literature [54]. Likewise, although assessments of health related quality of life were frequent in the included RCT’s, heterogeneity in instruments (e.g., SF 36, EQ 5D, and SGRQ) and reporting formats precluded quantitative synthesis in our review. Individual trials generally suggested trends toward QoL improvement with TR, but such effects were not consistent enough to support definitive conclusions. This pattern is consistent with systematic syntheses that reveal mixed results for QoL outcomes in TR trials, where some studies report meaningful gains in mental or overall QoL while others do not [55].

Comparisons with TR research in other chronic conditions provide further context. In chronic obstructive pulmonary disease (COPD), TR has been shown to improve exercise capacity and quality of life, with effects comparable to Centre-based pulmonary rehabilitation; however, psychological outcomes tend to be less pronounced unless psychosocial support is integrated [54]. Similarly, in heart failure and post stroke populations, TR improves physical function and adherence, but mental health benefits are inconsistent and often require dedicated interventions [56, 57]. The present findings in post COVID 19 patients thus mirror this broader pattern: TR effectively addresses physical deconditioning, but achieving meaningful psychological gains likely demands a more comprehensive, multi component approach that specifically targets anxiety, depression, and illness related distress. This similarity across conditions suggests that the modality of delivery (remote vs. in person) may be less important than the content of the rehabilitation program when it comes to mental health outcomes.

Safety and treatment adherence

TR was demonstrated to be an effective and safe modality, characterized by minimal adverse effects and completion rates comparable to traditional in-person controls (RR = 0.98). This finding aligns with the broader literature regarding rehabilitation for COVID-19 survivors, where adherence rates are typically high and serious adverse events remain rare. For example, previous systematic reviews have reported completion rates as high as 88% with negligible reactions attributed to TR protocols [19]. However, a more granular look at the raw data reveals a lower adherence rate in the TR group (76.06%) compared to the control group (84.36%). While this 8% disparity did not reach statistical significance, it warrants clinical attention as it may reflect the unique barriers inherent to remote interventions, such as digital fatigue, the absence of the social accountability found in face-to-face sessions, or technical hurdles that disproportionately impact older post-COVID populations. Unlike conventional care, TR demands a higher level of digital literacy and self-management, which may lead to higher attrition if not paired with robust technical support.

Furthermore, while our analysis identified no significant safety concerns, it is critical to acknowledge that safety was rarely a primary endpoint in the reviewed RCT’s. The infrequent reporting of minor adverse events and the limited number of studies specifically powered to detect safety signals suggest that the current consensus on TR safety is predicated on the absence of reported harm rather than dedicated, proactive safety testing. This perceived safety must be scrutinized due to inconsistent reporting standards across the literature; while serious adverse events were not reported, the absence of evidence should not be conflated with evidence of absence. Many trials focused exclusively on functional gains, potentially overlooking sub-clinical yet significant events such as digital eye strain, musculoskeletal strain from unsupervised exercise, or psychological distress stemming from isolation. Consequently, there is a clear need for standardized safety reporting protocols in future TR research to transition from anecdotal observations toward a truly evidence-based safety framework.

Clinical and research implications

Our results support TR as a viable option for improving physical function in adult COVID-19 survivors, especially when face-to-face rehabilitation is an option that is limited by geographic, physical, and public health factors. These results are of relevance to health systems seeking scalable solutions to long COVID-related functional impairments. Nevertheless, TR may need to be integrated with targeted respiratory and psychological interventions to address multisystem sequelae comprehensively.

Translating research data into real-world practice involves significant stakeholder engagement, including patients, clinicians, and health system administrators, to ensure programs match local contexts and are sustainable [58]. Personalized and complete remote rehabilitation models that include individualized exercise prescriptions, psychological support, and technology adaptations are an important field for development [59]. The TELEWEAR CR protocol emphasizes the use of wearable sensors to improve personalization and safety, especially for patients with complicated cardiovascular or respiratory comorbidities [60].

Despite robust effects for physical endpoints, diversity with regards to the nature of interventions, their duration of application, as well as the outcome measures considered points to the need for standardized approaches to optimize their use. Future research should examine dose response relationships (e.g., frequency and intensity of TR sessions), include active psychological support where relevant, and evaluate longer term outcomes beyond typical 4–17 week programs. There remains a need for more head to head comparisons with face to face rehabilitation, as some evidence suggests that in person programs may provide more comprehensive benefits in QoL and certain functional domains, though findings are not yet definitive [61].

Strengths and limitations

The strengths of our updated review are highlighted by a methodology based on a state-of-the-art, comprehensive model involving 28 RCTs with diverse geography, varying risks of bias, and certainty estimations utilizing the GRADE approach. The dual analysis of both absolute and change values provides a nuanced understanding of the TR effect, while subgroup analysis stratified by intervention duration, supervision type, and session length further explores potential sources of heterogeneity. Nevertheless, certain limitations warrant consideration. First, considerable heterogeneity existed among some results, especially regarding symptom relief and measures of dyspnea, which reduces precision and limits the ability to make uniform recommendations across all telerehabilitation formats. Second, the diversity of TR modalities and reporting formats makes direct comparison difficult; consequently, the pooled effect estimates should be interpreted as reflecting an average across diverse interventions rather than a single, well‑defined treatment. Further, some results had low to moderate certainty of evidence, particularly for psychological and quality-of-life outcomes. Findings for these outcomes require cautious interpretation and should be considered hypothesis‑generating rather than conclusive.

Regarding the risk of bias, the ROB 2 assessment showed that 15 of the 28 studies (54%) were low risk across all domains, while six studies (21%) had some concerns, mostly regarding the selection of the reported result, and the remaining seven (25%) were rated high risk due to a lack of blinding and incomplete outcome reporting. For outcomes relying on self-reports, such as dyspnea or quality of life, the lack of blinding may have inflated effect estimates, whereas objective outcomes like the 6MWD are less susceptible to these biases. While over half the evidence is methodologically strong, the presence of studies with important limitations necessitates caution when interpreting pooled findings. Furthermore, the lack of standardized reporting regarding participant dropout and inconsistent monitoring of non-serious adverse events limits our ability to provide a definitive feasibility profile. Specifically, reasons for dropout were not systematically collected; while some studies attributed withdrawals to dizziness from screen time, unrelated hospitalizations, or non-serious adverse events, the majority provided no specific reasons. This lack of transparency obscures patient- or intervention-related factors affecting adherence.

Differential adherence between groups was also not explored in depth. Although overall completion rates were reported (experimental: 76.06%, control: 84.36%; overall 79.87%), most trials did not report daily or weekly adherence separately, nor did they examine whether adherence varied by baseline characteristics such as age or technology literacy. Therefore, the comparable overall completion rates (RR = 0.98) should be interpreted with caution. Future trials should adopt standardized safety reporting frameworks, such as the CONSORT extension for harms, to strengthen the evidence base for TR safety and feasibility. When considered within the broader evidence on long‑term safety, remote rehabilitation continues to appear safe. Systematic reviews in cardiac populations, such as that by Anderson et al. [62], indicate that remote programs can be delivered safely over extended periods with adverse event rates comparable to Centre‑based care. This is particularly relevant in the post‑COVID era, where cardiovascular sequelae are increasingly recognized [41]. The integration of wearable sensors and real‑time monitoring, exemplified by the TELEWEAR‑CR protocol [60], may further enhance safety by enabling early detection of physiological responses and supporting personalized, data‑driven exercise prescription in alignment with emerging remote rehabilitation models [59].

Conclusion

In conclusion, TR is an effective modality for enhancing physical capacity and alleviating fatigue in post-COVID-19 patients. However, benefits related to symptoms, pulmonary function, psychological outcomes, and quality of life are inconsistent, likely due to heterogeneity in both interventions and measurement tools. While current evidence indicates it is well-tolerated with high completion rates, more rigorous reporting of adverse events in future trials is necessary to confirm its long-term safety profile across diverse clinical settings. Continued high-quality research is, therefore, needed to refine protocols and elucidate the full spectrum of benefits of telerehabilitation in post-COVID populations.