The Effectiveness of Telerehabilitation on Balance and Functional Mobility in Patients with Stroke: A Systematic Review and Meta-Analysis

Objective: The aim of this systematic review and meta-analysis was to investigate the effectiveness of telerehabilitation on improving balance and functional mobility in stroke survivors. Methods: Comprehensive searching was conducted from inception to May 2022. The inclusion criteria were studies evaluating the effectiveness of telerehabilitation in stroke survivors. Data regarding participants, intervention, outcome measures, and main results were extracted. PEDro scale and the Grading of Recommendations Assessment Development and Evaluation (GRADE) were used to assess the methodological quality and quality of evidence, respectively. Data Analysis: A total of fourteen articles) 594 patients) were included. A meta-analysis using a random-effect model was performed on thirteen studies )530 patients). Standardized mean difference (SMD) with 95% confidence interval (CI) was calculated for balance and functional mobility. Results: PEDro scale revealed ten good-quality studies, three fair-quality studies, and one poor-quality study. According to the available evidence, telerehabilitation has a small effect size in improving both balance (SMD 0.33 [95% CI 0.03 to 0.63]; P =0.03; low quality of evidence) and functional mobility (SMD 0.27 [95% CI 0.02 to 0.52]; P =0.03; low quality of evidence). Conclusion: Telerehabilitation may improve balance and functional mobility in stroke survivors. However, it is evident that more high-quality research is required due to the existence of low to very low-quality evidence with limited confidence in the effect estimate. Registration: PROSPERO registration number (CRD42022306410).

patients remotely by using communication technology (Jiang et al., 2018). Various technologies can be used to communicate between the patient and the rehabilitation specialist such as smartphone apps, web-based videoconferencing, etc. (Rogante et al., 2010).
Telerehabilitation services can be utilized to supplement and improve the quality of already available rehabilitation services. Stroke survivors have raised concerns about the absence of long-term assistance and continued unfulfilled rehabilitation needs after discharge from the hospital (Ullberg et al., 2016).
Telerehabilitation has the potential to reduce rehabilitation costs for both therapists and patients, as well as provide an opportunity to obtain rehabilitation services for patients who live in rural and distant areas and those with severely restricted mobility (Peretti et al., 2017). In addition, telerehabilitation can provide therapists with an alternative, innovative, and valuable method of providing rehabilitation to help stroke survivors recover, as well as to adjust and follow their progress remotely (Lloréns et al., 2015).
Telerehabilitation is a method of providing rehabilitation programs, so the mechanisms that lead to recovery should be similar to those seen in traditional rehabilitation programs that are widely recognized to lower the risk of institutional care and long-term impairment and raise independence in daily life activities (Kalra & Langhorne, 2007;Pollock et al., 2014). Improvements in function following treatment programs ending have been related to physiological recovery and brain neural plasticity (Kwakkel et al., 2004).
Given the expansion of this field's research and the potential of telerehabilitation to increase accessibility and quality of rehabilitation services while lowering costs, a review was necessary to evaluate this approach. Additionally, it is important since health providers are increasingly providing telerehabilitation services to patients. Consequently, the purpose of this systematic review and meta-analysis was to investigate the effectiveness of telerehabilitation in improving balance and functional mobility in stroke survivors.

Study Design
This study involved systematic review and meta-analysis conducted in accordance with the Preferred Reporting Items for Systematic Reviews and Meta-analyses (PRISMA) guidelines (Page et al., 2021). The review's protocol was registered in the International Prospective Register of Systematic Reviews (PROSPERO) with a registration number (CRD42022306410).

Eligibility Criteria
Studies were eligible for inclusion if all the following PICOS criteria were met: (P) Stroke survivors (aged > 18 years, no restriction based on the type of stroke, sex, or race). An Intervention group (I) received telerehabilitation program (no limitation based on type of technology used) in conjunction with other therapy or alone. A control group (C) did not receive any type of telerehabilitation. The outcomes of interest (O) were balance as a primary outcome, and functional mobility as secondary outcome. The study design (S) was randomized controlled trials, pilot trials, or clinical trials with a control group that were published in English with available full text. Studies that did not meet the previous criteria were excluded. The number of articles and the cause for exclusion are presented in Figure 1.

Study Selection Process
All records identified through searching were imported into reference management software (EndNote 20). The duplicates were removed, and an initial review of titles and abstracts was done by two independent reviewers. If the abstract did not include sufficient information, full text was screened. After that, for all studies that initially matched the inclusion criteria, a fulltext careful review for final decision was performed. Lastly, reference lists and citation searching of all included articles were performed to find further eligible studies. Any discrepancies between the two reviewers (NA and AA) were resolved by a third reviewer (MA) in a consensus meeting.

Data Extraction, Synthesis and Analysis
In this review, both qualitative and quantitative synthesis were conducted in accordance with PRISMA guidelines (Page et al., 2021), and outcomes of interest were balance (primary outcome) and functional mobility (secondary outcomes).

Data Extraction
A custom data extraction form was used to extract the data from each included study. Data extraction was independently performed by two reviewers (NA and AA) and reviewed by a third reviewer (AS). The following data were extracted: authors' names, publication year, participants' characteristics (sample size, sex, age), outcome measures used, telerehabilitation and control interventions (type, frequency, duration of sessions and program), follow-up, and main results. These data are represented in Table 2.
In order to conduct meta-analysis, the sample size, mean, and standard deviation of the telerehabilitation and control groups were extracted. If any study results were not reported in form of mean and standard deviation, the study's authors were contacted to obtain those data.  Berg Balance Scale (BBS) Barthel Index (BI) -Mobility Domains TRG: A WSN telerehabilitation system was used to conduct the balance training program (including static/dynamic sitting balance, touch screen manipulation, etc.) instructed by a remote therapist via videoconferencing. This system also involves vital signs monitor. CG: Conventional balance training program (including static/dynamic sitting balance, ball manipulation, etc.) instructed by a therapist via personal face-to-face contact.

No
Both groups showed significant improvement on BBS score (P < 0.001) and non-significant change in BI-Mobility (p= 0.088). In addition, no significant betweengroup differences were observed in all outcomes (P  0.05). Three 50-min sessions/week for 4 weeks (Lloréns et al., 2015) Sample size= 30 (M=17, F=13) Age: * TRG= 55.47±9.63 CG= 55.6±7.29 Berg Balance Scale (BBS) Performance-Oriented Mobility Assessment -Balance (POMA-B) Brunel Balance Assessment (BBA) TRG: Virtual reality-based telerehabilitation training program. The virtual environment involved a central circle and various items placed around it. Participants were asked to step on these items with the nearest foot while maintaining the other foot within the circle. There are nine defined levels of difficulty based on configuring the location of appearance, etc. CG: In-clinic virtual reality-based training program. Same procedures as TRG. # Both groups received conventional physical therapy in the clinic 4 weeks A significant effect was detected in both groups in all outcomes at postintervention (P<0.01), these improvements were preserved as non-significant differences were detected from post-intervention to the follow-up. However, no significant between-group TRG: Home-based tele-supervising rehabilitation including physical exercises (balance exercises, walking exercises, etc.) and electromyography-triggered neuromuscular stimulation (ETNS). Therapists supervised the participants by live videoconferencing. CG: Physical exercises and ETNS program carried out face-toface in the outpatient rehabilitation department.

weeks
Both groups demonstrated a significant effect within groups in improving BBS over time (P<0.001), but no significant between-group differences was observed (p > 0.05). 1-hour physical exercises and 20-min ETNS, twice in a working day for 12 weeks and a total of 60 sessions (Vloothuis et  Meta-analysis Review Manager (version 5.4) software was used to conduct the meta-analysis and generate forest plots. Overall effect size with 95% confidence interval was calculated using random-effects model. Standard mean difference (SMD) was used as effect measure and its clinical significance was interpreted according to Cohen's effect size classifications: small (0.2), medium (0.5), and large (0.8) (Cohen, 2013). In addition, 2 statistic was used to assess variability (heterogeneity) across trials, and results were rated according to the following classifications: low ( 2 ≤ 25%), medium ( 2 26 − 50%), and high ( 2 ≥ 75%) (Huedo-Medina et al., 2006).
Moreover, subgroup analyses were also conducted to assess the effectiveness of telerehabilitation alone or combined with conventional rehabilitation versus various control interventions.

Sensitivity Analysis
Two sensitivity analyses were conducted to evaluate the robustness of the results: (1) omitting outlier studies that had a confidence interval does not overlap with the confidence interval of the pooled effect (Viechtbauer & Cheung, 2010), (2) excluding studies with a poor methodological quality.

Publication Bias
SPSS (version 28) was used to assess the publication bias when the meta-analysis included 10 or more studies (Sterne et al., 2011). Evaluation was done by visual inspection of funnel plot and Egger's test which is used to test funnel plot asymmetry (Egger et al., 1997). In presence of publication bias, the trim and fill method was used to adjust the overall effect size based on missing studies (Duval & Tweedie, 2000).

Risk of Bias Assessment
PEDro scale was used to assess the included studies' methodological quality. It was developed by the Physiotherapy Evidence Database (PEDro) (Maher et al., 2003), and it is a reliable scale that is commonly used in systematic reviews (da Costa et al., 2013). There are 11 items on this scale with a total score of 10 as criteria one is not included in the overall score (Maher et al., 2003). The following criteria were used to rate the included studies: excellent quality (9-10), good quality (6-8), fair quality (4-5), or poor quality (≤3) (Foley et al., 2003). Two independent reviewers (NA and AS) assessed the methodological quality, and any disagreements between the two reviewers were resolved by a third reviewer (AE).

Quality of Evidence Assessment
The assessment of quality of evidence was performed by two reviewers (MA and AF). The Grading of Recommendations Assessment, Development, and Evaluation (GRADE) system was used to assess certainty of evidence and make a summary of findings table. There are four levels for rating evidence: 'High quality', 'Moderate quality', 'Low quality' or 'Very low quality'. Each of these levels has a description that focuses on the certainty of the results, how much the effect estimate is closely related to the true effect and the influence of evidence degree for recommending future studies. GRADE evaluation was determined through consideration of five domains (risk for bias, inconsistency, indirectness, imprecision, and publication bias) (Atkins et al., 2004;Balshem et al., 2011;Guyatt et al., 2011).

Study Characteristics
Fourteen studies with a total of 594 participants were included in this review, 278 participants were in telerehabilitation groups and 316 in control groups. Sample size ranged from 10 to 95 and included both genders. All included studies were published during the period from 2012 to 2022.
Balance was evaluated in all included studies by using different outcome measures, such as Berg Balance Scale (BBS), Performance-Oriented Mobility Assessment -Balance (POMA-B), Brunel Balance Assessment (BBA), Mini Balance Evaluation Systems Test (Mini-BEST), Lean-and-Release Assessment, and Step Test. While mobility and gait were assessed by using 10 Meter Walk Test (10MWT), Rivermead Mobility Index (RMI), 6-Minute Walking Test (6MWT), Functional Ambulation Category (FAC), Mobility Domains of Barthel Index (BI), Stroke Impact Scale (SIS), and Nottingham Extended ADL Scale (NEADL). Moreover, Timed Up & Go Test (TUG) and Postural Assessment Scale for Stroke Patients (PASS) were used to measure both balance and mobility. Monitoring and follow-up periods ranged from two weeks to 12 months in six studies (Chen et al., 2017;Cikajlo et al., 2012;Lloréns et al., 2015;Saywell et al., 2021;van den Berg et al., 2016;Vloothuis et al., 2019), while other studies did not follow up the patients.

Risk of Bias in Studies
According to the PEDro scale, included studies' methodological quality values varied from 3 to 8 out of 10. Ten studies were rated as good quality (Burgos et al., 2020;Chen et al., 2017;Junata et al., 2021;Lin et al., 2014;Lloréns et al., 2015;Salgueiro et al., 2022a;Saywell et al., 2021;van den Berg et al., 2016;Vloothuis et al., 2019;Wu et al., 2020), three studies were rated as fair quality (Chen et al., 2021;Krpic et al., 2013;Salgueiro et al., 2022b), and one study was rated as poor quality (Cikajlo et al., 2012). All included studies did not perform subjects' and therapists' blinding. In addition to these criteria, the least met criteria were allocation concealment (six studies), assessors' blinding (nine studies), and intention-to-treat analysis (nine studies). Thus, the highest risk of bias for these studies was selection, performance, detection, and attrition bias. Details of the methodological quality appraising of the included studies are shown in Table 3.
Overall, subgroup analysis showed non-significant favor toward telerehabilitation with presence of small to medium effect sizes. Furthermore, the subgroup differences test revealed a non-significant subgroup effect (P = 0.67) (Figure 2). Sensitivity analysis was performed by excluding Wu et al., (2020) which had a larger effect size than other studies in the meta-analysis and was considered an outlier study. Even after removing this outlier study, telerehabilitation was still associated with a small effect size in improving balance, however the result was no longer significant (SMD 0.15 [95% CI -0.03 to 0.34]; P =0.1). Moreover, excluding Wu et al., (2020) had a significant effect on reducing the heterogeneity across studies ( 2 = 0%) (Appendix A).

Effects of Telerehabilitation on Functional Mobility
Five studies with 262 participants were eligible to be included in this analysis (Lin et al., 2014;Salgueiro et al., 2022a;Saywell et al., 2021;van den Berg et al., 2016;Vloothuis et al., 2019). The overall effect was favorable for telerehabilitation in terms of improving functional mobility in stroke survivors with a significant difference and small effect size (SMD 0.27 [95% CI 0.02 to 0.52]; P =0.03). Moreover, variability across studies evaluation showed presence of low heterogeneity ( 2 = 4%) (Figure 3). Subgroup analysis showed a significant medium effect size of telerehabilitation when compared to conventional rehabilitation (SMD 0.6; P=0.01), and a non-significant small effect size of telerehabilitation plus conventional rehabilitation when compared to conventional rehabilitation (SMD 0.18; P=0.27), while balance training showed a non-significant small effect when compared to telerehabilitation (SMD -0.12; P=0.77). However, subgroup differences test showed non-significant subgroup effect (P = 0.19) (Figure 3).
Since no studies of poor methodological quality were included in this meta-analysis and there were no outlier studies, a sensitivity analysis was not conducted.

Publication Bias
Visual inspection of the funnel plot showed slight asymmetry, an indication that publication bias may be present ( Figure  4). However, a non-significant Egger's test indicated absence of publication bias (p=0.9). Additionally, the overall effect size was not modified by the 'trim and fill' method, demonstrating that there was no need for adjustment based on missing studies.

Figure 4
Funnel Plot for the Meta-analysis of the Effect of Telerehabilitation on Balance

Quality of Evidence
According to the GRADE system, the quality of evidence was rated from 'Low' quality to 'Very Low' quality due to the presence of a serious or very serious risk of bias, inconsistency, or imprecision. Table 4 includes further details on the GRADE evaluation. Reasons to downgrade the current evidence: a Crucial limitation for more than one criterion; b Crucial limitation for one criterion; c High heterogeneity 2 > 75%; d Small sample size.
⨁⨁◯◯: Low quality (limited confidence in the estimated effect); ⨁◯◯◯: Very Low quality (very limited confidence in the estimated effect).

Discussion
The purpose of this systematic review and meta-analysis was to investigate the effectiveness of telerehabilitation in improving balance and functional mobility in stroke survivors. The results showed that telerehabilitation was associated with a significant but small improvement in balance (SMD 0.33; low-quality evidence) and functional mobility (SMD 0.27; low-quality evidence) immediately post-intervention in stroke survivors. The methodological quality of the included studies was rated as good quality in ten studies, fair quality in three studies, and poor quality in one study.
Improving balance is one of the most important rehabilitation aspects that should be targeted during rehabilitation programs for stroke survivors to improve their safe mobility and overall function (Wu et al., 2020). This may require a long-term course of rehabilitation that can be delivered via telerehabilitation, which is an effective way to allow easier access to physiotherapy for patients with serious disabilities (Peretti et al., 2017). In addition, previous studies reported that telerehabilitation could decrease the costs of treating and rehabilitating patients (Nizeyimana et al., 2022;Perry et al., 2011). Unfortunately, a number of obstacles prevent telerehabilitation from being widely used, such as administrative license, medicolegal ambiguity, financial stability, and absence of technology infrastructure particularly in low-income countries (Akbik et al., 2017;Sarfo et al., 2017). To get through these obstacles, telerehabilitation systems need to be developed and studied more in the future.
The findings of this review suggest that telerehabilitation could help stroke survivors improve their balance and functional mobility. However, there is a significant variation amongst the included studies in terms of the interventions utilized, the comparison interventions, and the outcomes measured. In addition, different information and communication technologies were used; for example, some research relied on telephone calls, while others included videoconferencing, mobile apps, and virtual reality systems. Furthermore, the length of rehabilitation programs and the frequency of follow-up or interactions with medical staff varied from one study to the next. There is currently insufficient evidence in the literature to determine which model or telerehabilitation tool is best for these individuals, and future comparative studies are recommended.
Subgroup analyses were carried out to assess the impact of various interventions used in the included studies. However, neither of the analyses proved that the effects of the various subgroups varied. The possible explanation was the presence of a non-significant small effect of telerehabilitation alone or combined with conventional rehabilitation on balance and functional mobility when compared with various control interventions in most of the included studies.
Funnel plot asymmetry and the presence of medium statistical heterogeneity among trials in the balance meta-analysis could be due to the existence of outlier study and methodological quality issues (Egger et al., 1997). According to the sensitivity analyses, the overall significant effect of telerehabilitation on balance was affected by removing the outlier and poor methodological quality studies. However, none of these analyses changed the direction of the effect, with the greatest decrease in heterogeneity and effect size when the outlier study was excluded.
Due to the lack of blinding in the subjects, assessors, and therapists, as well as no allocation concealment or intention to treat analysis, there were some methodological issues, including selection, performance, detection, and attrition bias. Moreover, the GRADE system was used to classify the level of evidence, and it revealed low to very low-quality evidence, which limited the confidence in the effect estimate. Additionally, the limited number of trials, the presence of small sample sizes in many included studies, discrepancies in rehabilitation protocols, and the frequency and duration of the interventions across trials, made it is difficult to draw clinical recommendations based on the current evidence.
The current review included fourteen studies investigating the efficacy of telerehabilitation on balance and functional mobility. All of the studies were published within the last 10 years, indicating that this method of rehabilitation is still relatively new. Some ongoing studies have been found, suggesting that more research will be forthcoming. Researchers in this area should perform large RCTs with high methodological quality that could support the available evidence and provide more information, as well as report the data in accordance with a clear standardized guidelines such as CONSORT guidelines (Schulz et al., 2010). This review describes different telerehabilitation application methods and outlines both advantages and disadvantages of this approach, which could be a starting point for improving telerehabilitation's techniques and equipment. It should be noted that these trials should not necessarily show that telerehabilitation produces superior outcomes, but rather proof of comparable outcomes, so it will provide evidence to support the use of this new and alternative method of delivering rehabilitative services that is more affordable and accessible.
This review has some limitations that would limit the generalizability of the findings. Searching included only papers that were published in English. Relevant studies in other languages were neglected, which may have impacted the results and conclusion. Moreover, no analysis or conclusions regarding long-term effects could be made, as several of the reviewed papers did not provide follow-up data. Furthermore, standardized mean difference, which is less clinically significant than a mean difference, was used to compute the pooled effects. Lastly, the accuracy of the reported finding may be impacted due to presence of the outlier study.

Conclusion
According to the available evidence, telerehabilitation may improve balance and functional mobility in stroke survivors. However, it is evident that more high-quality research is required due to the existence of low to very low-quality evidence with limited confidence in the effect estimate.