Telerehabilitation versus face-to-face rehabilitation in the management of musculoskeletal conditions: a systematic review and meta-analysis

Abstract Background Musculoskeletal conditions such as spinal pain and osteoarthritis are among the leading causes of years lived with disability worldwide. With the COVID-19 pandemic forcing many healthcare providers to change the way in which care for chronic conditions is delivered, telehealth is an alternative to face-to-face consultations that can be used for both assessment and provision of therapy and support. Objectives To identify, appraise and synthesise findings from all randomised controlled trials (RCTs) that compared telehealth to face-to-face consultations for patients with any type of musculoskeletal condition. Methods Systematic review and meta-analysis. We used the GRADE approach to assess the quality of evidence related to all outcomes. We searched three electronic databases (PubMed, Embase, CENTRAL), clinical trial registries and citing-cited references of included studies. Results Five RCTs were includable: one in patients with osteoarthritis of the knee, one in patients with osteoarthritis of the knee or hip in preparation for a total joint arthroscopy and three after total knee replacement. Telehealth was conducted by video in four trials and by phone in one. A total of 402 participants were analysed across the five trials. There were no significant differences in pain outcomes (WOMAC) between telehealth and face-to-face therapy immediate post-intervention (mean difference (MD): 0.12 (95% CI −2.3 to 2.6, p = .92) or two months post-intervention (MD): 1.2, (95% CI: −2.7 to 5.1, p = .55). Similarly, outcomes related to function, quality of life and satisfaction were comparable between the two modes of delivery immediate post-intervention, with no significant differences reported. Conclusion There is limited low quality evidence that there is no significant differences between telehealth-based delivery of rehabilitation to patients with osteoarthritis or following knee surgery and face-to-face therapy for pain, function, quality of life and satisfaction. These findings should be should be interpreted with caution due to the small number of included studies and small sample size. HIGHLIGHTS Findings based on a small number of trials and very low-quality evidence suggest that there is no difference between telehealth and face-to-face consultation for rehabilitation for adults post-knee surgery. Outcomes related to pain, function and quality of life were comparable between the two modes of delivery. There is a significant gap in knowledge relating to cost outcomes, warranting studies that further evaluate cost-effectiveness of telehealth and the subsequent long-term sustainability of telehealth.


Introduction
Musculoskeletal conditions such as spinal pain and osteoarthritis are among the leading causes of years lived with disability worldwide [1]. These often chronic conditions are associated with significant impact on the individual, due to pain and reduced function, and society due to work absenteeism and reduced work productivity. Musculoskeletal conditions are also associated with substantial financial costs [1]. According to 2008/2009 data, in Australia, musculoskeletal conditions accounted for 9% of health care expenditure, making them the fourth most costly health condition, behind cardiovascular, oral health and mental disorders [1]. Treatment is the dominant component of this expenditure which includes hospital admitted patient services (e.g. joint replacements), out of hospital expenses (e.g. outpatient clinics) and pharmaceuticals (e.g. pharmaceutical benefits scheme) [2]. Guideline management for most musculoskeletal conditions begins with advice, condition-specific education and exercise [2,3]. Surgical treatment options such as joint replacement surgery for knee or hip osteoarthritis should only be considered following conservative management and when symptoms are no longer responsive to noninvasive approaches [3]. With the prevalence and treatment costs of musculoskeletal conditions increasing, there is growing recognition for the need to identify effective treatment options that enable timely and equitable access to services irrespective of location, accessibility or public health policies, such as lockdowns or quarantine, in response to .
Telehealth is a promising mode of health service delivery that may increase access to health services. Telehealth includes telephone and videoconferencing to provide healthcare at a distance [5], and provides an alternative to face-to-face consultations for both assessment and the provision of healthcare and support. The main benefits of telehealth for both patients and health care professionals are access and convenience [5]. Several systematic reviews have evaluated the effectiveness of telehealth for the assessment and management of musculoskeletal conditions [6][7][8]. These reviews suggest telehealth is effective for improving function for people with osteoarthritis and following joint replacements (hip and knee arthroplasty) [8,9]. For chronic low back pain moderate quality evidence suggests that telehealth is not superior to minimal interventions (i.e. patient education) for reducing pain, disability or function in the short or medium term [7]. Compared to control, three trials found telehealth to be superior for improving quality-of-life in people with back pain of any duration (ranging between immediate post-and 12 months post-intervention) [7]. These findings, however, have several limitations related to their recency, quality and quantity of studies included and none have evaluated the cost-effectiveness of telehealth in comparison to standard face-to-face consultations.
With the recent COVID pandemic, telehealth has been a pivotal tool facilitating medical treatment when in-person contact was restricted. Given the current rapid increase in telehealth delivery, an updated, comprehensive systematic review that synthesises estimates of treatment effectiveness from randomised controlled trials (RCTs) is warranted to inform and guide the delivery of physiotherapy healthcare services. Therefore, the aim of this systematic review was to identify evidence for the effectiveness of telerehabilitation for the management of patients with musculoskeletal conditions in terms of pain, function, quality of life, satisfaction and costs as compared to conventional face-to-face rehabilitation. We used the GRADE approach to assess the quality of evidence related to all outcomes.

Methods
This systematic review is reported following the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) statement [10] and the review protocol was developed prospectively. Any deviations from the protocol are reported in the relevant methods section.

Study designs
We included RCTs of any design (i.e. individual, cluster, factorial, crossover). All other primary study types (e.g. controlled nonrandomised studies, qualitative studies and observational studies) and reviews, were excluded.

Population
Studies were included if their populations of interest comprised individuals of any age or gender, diagnosed with any type of musculoskeletal condition. This included post-operative rehabilitation.

Interventions
Therapy interventions involving a synchronous telehealth component (i.e. involving real-time communication) via either phone or video, were included. Studies examining novel treatments for rehabilitation (i.e. experimental interventions for musculoskeletal care that were considered outside of standard care), as well as asynchronous data transfer (e.g. E-mail), virtual reality software, automated feedback, webbased modules or specialised online apps were excluded. Studies were included if they involved care provision by a physiotherapist (physical therapist) or occupational therapist. Studies which involved patients consulting a specialist (i.e. orthopaedic surgeon), were excluded. The focus of this review was on primary health care providers in the community.

Comparators
We included studies with a comparable (i.e. similar or equivalent frequency and duration of therapy) face-to-face comparator or telehealth comparator (i.e. video intervention with telephone comparator). Any comparator that involved a wait-list control or clinically inequivalent active comparator was excluded, as these interventions were considered too dissimilar to the telehealth interventions investigated.

Outcomes
The primary outcomes were pain and function scores, and secondary outcomes were quality of life, patient satisfaction and costs of therapy. Outcomes were grouped by time points, i.e. immediate postintervention, and short-term impact (two months post-intervention). See Appendix A for scales used to measure outcomes of interest.

Search strategy
Electronic databases, including PubMed (via MEDLINE), Embase (via Embase) and CENTRAL (via the Cochrane Library) were searched for potentially relevant primary studies from inception until 18 November 2020. The search string was translated for use in other databases using the Polyglot Search Translator [17]. Complete search strategies are provided in Appendix B. The searches were deliberately broad, as the present review was conducted as part of a series of systematic reviews on the effectiveness of telehealth compared to face-to-face for healthcare provision in primary care and allied care.

Other searches
In addition, a backward and forward citation search of the included studies was undertaken using the Scopus database on the 15 December 2020 to identify any further relevant studies. Clinical registries (clinicaltrials.gov and WHO ICTRP) were searched on 25 March 2021. Search strategies are provided in Appendix B.

Restriction on publication type
No restrictions by language or publication date were applied. Only published trials that were available in the full text were considered for inclusion. Studies published as abstract only (e.g. conference abstract) with no additional results information available (e.g. from a clinical trial registry record) were excluded.

Screening and data extraction
Paired authors (NK, HG, RP, MC, AMS, JC, PG) independently reviewed all titles and abstracts identified by the electronic search against the inclusion criteria. One author (JC) retrieved full-text for includable conditions and two authors (NK, RP) screened the full-texts for inclusion. Disagreements were resolved by consensus or referral to a third author, when necessary. Paired authors (NK, RP, MC) independently extracted data from all eligible trials using three data extraction forms: Table of Characteristics form, Primary and Secondary Outcomes data form and Risk of Bias form. Any discrepancies were resolved by discussion or by reference to a third author (see Box 1).

Risk of bias
Two authors (NK, MC) independently assessed the risk of bias for each included trial using the Cochrane Collaboration's Risk of Bias tool 1 [18]. We used Cochrane Risk of Bias Tool 1 rather than Tool 2, as Tool 1 allows for the assessment of biases arising from study funding and conflict of interest (under domain 7, other bias). The following domains were assessed as either 'high', 'low' or 'unclear': method of random sequence generation and allocation concealment (selection bias), blinding of participants and personnel (performance bias), blinding of outcome assessment (detection bias), incomplete outcome data (attrition bias), selective outcome reporting (reporting bias) and other bias (assessing biases from conflict of interest and funding issues). All disagreements were resolved by consensus.

Assessment of evidence using GRADE (grading of recommendations, assessment, development and evaluations) analysis
Although we had not pre-specified this in the protocol, two authors (NK, ZAM) independently assessed each outcome using the GRADE approach, to provide further understanding around the quality of evidence being evaluated for each comparison and at each time point [19]. Five domains were evaluated: risk of bias, inconsistency, indirectness, imprecision and publication bias. The consensus was sought between authors and reached for each domain's direction and strength. All RCTs in this review were initially designated as high quality. After evaluating the principal domains, a Box 1 . List [20][21][22].

Statistical analysis
Review Manager 5.4 was used to calculate the effect of interventions [23]. A meta-analysis was conducted where data were sufficient to pool (i.e. 2 or more trials reporting on the same outcome). For all outcomes (i.e. pain, function, quality of life), the results are presented as mean differences or standardised mean differences. We used a random effects model, in anticipation of considerable heterogeneity. Statistical heterogeneity was assessed using the I 2 statistic. The individual was used as the unit of analysis, where possible. However, where data on the number of individuals with primary and secondary outcomes of interest was not available, we extracted the information as it was presented (e.g. mean scores for the differences between groups). We did not contact investigators or study sponsors to provide missing data.
A funnel plot was not created, as fewer than 10 studies were included in the analysis. Subgroup analyses were conducted according to duration of follow-up.
As none of the included studies were graded at 'high' risk of bias in three or more domains, a prespecified sensitivity analysis assessing the impact of excluding trials rated at high risk of bias for three or more domains, was not conducted.

Search results
The electronic search retrieved 5423 references, supplemented with 192 references from forward and backward citations of the included studies and 94 records identified from the clinical trial registry search, resulting in 2048 records to screen after deduplication. Screening these on title and abstract excluded 1719 references, leaving 329 articles for which we obtained full-text. Screening of these full texts excluded another 316 which left five RCTs (reported in nine articles) for inclusion in this systematic review (see Figure 1). Reasons for exclusions are reported in Appendix D. We identified four potentially relevant clinical trials that are currently in progress, these are listed in Appendix E.
Characteristics of the five included trials are presented in Table 1. A total of 402 participants were analysed across the five trials. Three trials were conducted in Canada [24-27, 31, 32], and one each in Australia [29, 30] and Nigeria [28]. Four trials focused on patient rehabilitation (or prehabilitation) in preparation for knee arthroplasty as a result of significant osteoarthritis [25-32]. One trial included a mixed population of people with hip and knee osteoarthritis who were awaiting joint replacement surgery [24]. In all trials, rehabilitation comprised prescribed exercise programs delivered by physiotherapists. All five trials compared conventional therapy in clinic settings (face-to-face) versus the home or clinic-based treatment rooms for telehealth interventions; no trial compared the different modes of telehealth delivery. Telehealth was conducted by video in four trials [24-27, 29-32], and by phone in one trial [28].

Risk of bias
Overall, the included trials were mostly of low or unclear risk of bias. In particular, random sequence generation, incomplete outcome data and selective reporting were at low risk of bias for 100% of the trials. Due to the nature of the interventions (telephone or video vs. face-to-face), blinding of the patients and health care providers was not possible, resulting in a high risk of performance bias for all trials. Allocation concealment, blinding of outcome assessment and other biases were generally at low risk of bias, however, some trials were rated at unclear risk of bias, due to nonreporting ( Figure 2).
The quality of evidence for both pain outcomes (immediate post-intervention and two months postintervention) was very low (downgraded due to indirectness and imprecision).
Russell et al. (2011) explored the impact of telehealth on WOMAC pain scores, however, reported mean differences pre-and post-rehabilitation for intervention groups. Following the intervention, the telehealth group achieved outcomes comparable to those of the conventional rehabilitation group in the WOMAC pain subscale with no significant differences observed. Mean differences before and after treatment were reported as 2.97 (SD: 2.31) in telehealth vs. 2.19 (SD: 1.76) in face-to-face groups.

Function outcomes -WOMAC function
Three studies reported data on this outcome -two were able to be meta-analysed [24,25], and the remaining one summarised narratively [29] (Figure 4).
There were no significant differences in function outcomes between telehealth and face-to-face  therapy immediate post-intervention (221 participants, mean difference: À 0.24, 95% CI À 3.62 to 3.13, p ¼ .89). The quality of evidence for function outcomes was very low (downgraded due to indirectness and imprecision). At two months post-intervention, Moffet et al. reported no significant differences between telehealth and face-to-face therapy for function outcomes (198 participants, mean difference: 1.0, 95% CI À 2.61 to 4.61, p ¼ .59).
Russell et al. (2011) reported that both the telehealth and the face-to-face groups had clinically important improvements in the WOMAC function scale (within-group change, i.e. comparing baseline to post-intervention), however, the differences between the two groups were not significant. Mean differences pre-and post-intervention for each group were reported as 3.52 (SD: 2.35) for telehealth and 2.45 (SD: 1.84) for conventional rehabilitation. Moffet et al. (2015) reported an overall quality of life score using the KOOs scale. There were no significant differences between groups at the post-intervention time point: mean difference ¼ 2.60 (95% CI: À 2.67 to 7.87, p ¼ .33) or the two-month follow-up: mean difference ¼ 0.50 (95% CI: À 4.77 to 5.77, p ¼ .85). The quality of evidence for both qualities of life outcomes (immediate post-intervention and two months post-intervention) was very low (downgraded due to indirectness and imprecision).

Quality of life -physical component
Two trials reported data on quality-of-life outcomes, separated into physical and mental components and could be pooled (Doiron-Cadrin et al. 2020, Odole and Ojo 2014) (Figures 5 and 6).  Two trials provided relevant data on the quality of life -physical composite outcome (total participants, n ¼ 73). There was no difference between telehealth and face-to-face groups; standardised mean difference was: À 0.04 (95% CI: À 0.5 to 0.42, p ¼ .87). The quality of evidence for the quality of life -physical outcome was very low (downgraded due to indirectness and imprecision).

Quality of life -mental component
For the quality of the life-mental component score, we identified two trials, involving a total of 73 participants. We did not find evidence of a difference between the two treatments in this comparison, standardised mean difference was 0.14 (95% CI: À 0.32 to 0.6, p ¼ .55). The quality of evidence for the quality of life-mental outcome was very low (downgraded due to indirectness and imprecision).    intervention, there were no reported significant differences in satisfaction between groups relating specifically to the delivery of services (telehealth: n ¼ 78, mean 84.6 (SD: 11.9), p ¼ .38 vs.

Costs
One trial reported on the costs of in-home rehabilitation post-TKA via telehealth compared to conventional face-to-face therapy, using a real total cost analysis [27]. The mean total cost of total treatment from the perspective of the healthcare system (including all received and cancelled treatments) was significantly lower in the telehealth group than the face-to-face group (CAD$1,224 (SD: 241) telehealth vs. CAD$1,487 (SD: $553), MD: À 263, 95% CI: À 382 to À 143, p ¼ <.001)(Years costed: 2008-2013). Similarly, the reported mean cost per single treatment for the telehealth group was significantly lower CAD $80.99 (SD: 26.60) compared to CAD $93.08 (SD: 35.70) for the face-to-face group (p ¼ .008).

Discussion
This systematic review of five RCTs suggests that there is no difference in effectiveness between telehealth rehabilitation for knee and hip osteoarthritis compared to conventional in-person rehabilitation delivered face-to-face. Outcomes related to pain, function and quality of life were comparable between the two modes of delivery. Overall, these findings are based on a small number of trials and very low-quality evidence (downgraded for indirectness and imprecision). This suggests that the evidence available is very uncertain about the effectiveness of telehealth on pain, function, quality of life and costs and these findings should be interpreted with caution.
Our review provides an up-to-date, high-quality synthesis of the effectiveness of telehealth as a treatment option for musculoskeletal conditions. While the finding of our review generally align with those of Cottrell et al.  2017), the methods used in these previous reviews (specifically, inclusion of cohort studies) are likely to influence the estimates of treatment effect and strength of conclusions made by these reviews. By synthesising RCTs alone and through applying the GRADE criteria the findings of this review cast much greater uncertainty over the effectiveness of telehealth compared to face-to-face consultations for musculoskeletal conditions. Our review also identified and analysed two recent publications that were not considered in previously published reviews. Further high-quality trials with large sample sizes are required to improve the certainty of evidence and therefore our confidence in the effectiveness of telehealth for managing musculoskeletal conditions. Accurate reporting of both the size of the effect of an intervention on an outcome and the certainty of evidence is critically important as it is this information that is then used by clinicians, healthcare services and governments to inform practice and guide the allocation of already scarce funds and resources.
The COVID-19 pandemic has forced many healthcare providers to change the way in which care is delivered. Findings of this review suggest no difference between telehealth and face-to-face consultation for rehabilitation for adults post-knee surgery, and further evaluation may be an alternative mode of delivery for patients with musculoskeletal conditions. However, from a practical and policy perspective there is a need for increased guidance, recommendations and training for physiotherapists around the optimal delivery of telehealth interventions [33,34]. In response to this demand, professional organisations and governing bodies including the Australian Health Practitioner Regulation Agency (AHPRA) and the Australian Physiotherapy Association have developed guidelines to assist with the delivery of telehealth interventions, facilitated education and training in its use [35,36]. The pandemic and the need for alternative treatment approaches have also incited responsive changes to remuneration for healthcare providers, with telehealth consultations eligible for rebates through public and private insurance. Such measures have resulted in the rapid uptake of telehealth and demonstrate the ability of both patients and health professional to adjust to alternative models of care delivery. However, significant barriers limiting the widespread utilisation of telehealth both nationally and internationally remain. These include patients' preferences and barriers to the use of telehealth (e.g. age, level of education, level of computer literacy), clinician skills (e.g. technological self-efficacy, knowledge and skills), organisation/systems (e.g. lack of infrastructure and equipment, reimbursement) and technological factors (e.g. consumer-grade technology, internet quality) [34].
The strengths of this review lie in the robustness of the systematic review methods. It included only randomised trials, which are the least biased and most reliable forms of evidence to assess treatment effectiveness. Furthermore, our search strategy was comprehensive, including searches of multiple databases, trial registries, forward and backward citation searches to ensure a complete overview of the relevant literature. This review is also the first to include a well-defined, systematic process for rating the quality of evidence across studies, using the GRADE approach. The limitations of this review include the small number of includable trials all with relatively small sample sizes which increases the potential for selection bias. As a result, the findings of this review should be interpreted with caution. The small number of trials included is likely attributed to the very specific inclusion criteria, which limited included studies to those assessing services considered 'usual care' and those comparing the delivery of identical or similar therapies to the telehealth and face-to-face groups. This excluded several trials which utilised web-based communication, apps, asynchronous data transfer with the use of sensors and other equipment and virtual reality software. Furthermore, there was significant variation in how telehealth was delivered in included studies in terms of the types of interventions provided, sample sizes, the duration of interventions, follow-up times and reporting of results. It is also important to consider the limitations on the generalisability of our findings to the management of a broad scope of musculoskeletal conditions. There are several barriers that limit the development of high-quality studies comparing telehealth and faceto-face rehabilitation for musculoskeletal conditions. These include, among others, heterogeneity in the types of health professionals that provide therapeutic care for MSK conditions (i.e. physiotherapists, chiropractors, GPs, etc.), and the resulting variation in prescribed therapies. This leads to subsequent differences across clinical trials in both the components of interventions and in their delivery, which impacts the analysis of effects [37]. All of the studies eligible for inclusion in this review related to rehabilitation as applied to hip and knee arthroplasty and as such may not be applicable to other musculoskeletal conditions. Results should be interpreted with caution.
There is a need for further high-quality trials to determine the effectiveness, acceptability and feasibility of telehealth interventions for the management of musculoskeletal conditions. Future trials should aim to recruit larger sample sizes and follow participants over longer follow-up periods. There is also a significant gap in knowledge relating to cost outcomes, warranting studies that further evaluate costeffectiveness of this treatment approach and the subsequent long-term sustainability of telehealth. This research also has important implications in facilitating its uptake and in guiding the optimal delivery of telehealth and its integration into health services.

Conclusions
Overall, based on five RCTs, the delivery of rehabilitation via a telehealth to patients prior to or following knee surgery appears to be equivalent to therapy delivered face-to-face. These findings are based on very low-quality evidence and should be interpreted with caution. To provide a better indication of the impact of telehealth, further high-quality primary trials are warranted in the musculoskeletal domain.

Disclosure statement
No potential conflict of interest was reported by the authors.

Funding
This systematic review was commissioned by the Department of Health, Canberra, Australia, as part of a series of systematic reviews on the effectiveness of telehealth. A review protocol was created, however, due to the parameters of the contract with the funder was unable to be uploaded onto an online registry. The funder was involved in the refining of the study question (PICO). The funder was not involved in the conduct, analysis, or interpretation of the systematic review, or in the decision to submit the manuscript for publication.