Background

Knee or hip osteoarthritis are dominant sources of disability, affecting approximately 776 million people globally [1]. These conditions are leading contributors to the rapid increase in orthopaedic surgeries worldwide over the last decades, with most of the increase occurring in total knee (TKR) and hip replacement (THR) [2]. Given the large and increasing financial burden of these procedures, potential efficiencies in the model of care for arthroplasty patients are a matter of considerable policy interest [3]. Rehabilitation services form a core component of the care pathway for THA and TKA patients, as a means of facilitating the recovery of functional independence after surgery. Due to the increased life expectancy and the limited resources devoted to public health, the demand for effective and sustainable rehabilitation services seems mandatory in order to cope with the needs of the aging population [4].

Recently, innovative technologies have brought affordability and convenience to the healthcare consumers, such as eHealth, telemedicine, wearables, virtual reality (VR) and online educational tools [5]. A growing body of literature supports the use of telerehabilitation in improving patient satisfaction and health outcomes for a diverse range of clinical conditions, such as neurological diseases [6, 7], stroke [8], cancer [9], cardiac and pulmonary rehabilitation [10]. Compared to face-to-face rehabilitation, services delivered remotely via telephone or internet are more affordable and accessible, particularly for people living in rural areas [11]. In addition, telerehabilitation systems integrated with biosensors, accelerometers and educational software provide individualised support for people to monitor the progress of their physical rehabilitation at home, whilst allowing the therapist to intervene timely and effectively [12]. Several studies have shown that game-based or VR-assisted rehabilitation provides a motivating environment for achieving different therapeutic goals [13]. Importantly, these innovative technologies empower consumers to take an active role in decision-making and disease management, resulting in improvements of overall health awareness, adherence to treatment and satisfaction [14].

Despite the increasing popularity of available innovative health products in the market, there is insufficient evidence of their effectiveness or safety in musculoskeletal (MSK) rehabilitation. A few systematic reviews of telerehabilitation have been conducted but only yielded a handful of trials [15,16,17]. However, along with the rapid progress in the technologies and the growing service demand, the number of publications in this topic also increased since then, thus, it is necessary to update the evidence at a timely manner. In addition, other blooming technologies, such as game therapy and virtual biofeedback have not been well investigated. Thus, this review aimed to update the current evidence and evaluate the effectiveness and safety of technology-based rehabilitation in comparison with usual care in people undergoing TKR and THR.

Methods

A protocol for this review was registered a priori in PROSPERO (CRD42017078924) and preliminary results were presented in a conference [18]. This systematic review with meta-analyses reported according to the Preferred Reporting Items for Systematic Reviews and Meta-Analysis (PRISMA) statement [19]. All the screening, data extraction and quality assessment were performed by two authors (XW, GV) independently and any disagreement was resolved by consensus with a third reviewer (MLF).

Literature search

Six electronic databases were searched without language or time restrictions for relevant studies: MEDLINE, EMBASE, Cochrane Library, CINAHL, SPORTDiscus, Physiotherapy Evidence Database (PEDro); from inception to November 7th, 2018. The search strategy was developed by a research librarian and contained both controlled vocabulary and free text terms (Additional file 1: Appendix 1). The initial search strategies included lumbar spinal surgeries, as lumbar spinal surgeries are also highly prevalent in orthopaedic surgeries. However, there is only one study in lumbar spinal surgeries has been identified, so we only reported results for TKR and THR in this paper.

Study selection

The population of interest was people undergoing rehabilitation after elective TKR and THR. Eligible studies were randomised controlled trials (RCTs) that investigated the effectiveness of any technology-based intervention, in isolation or in combination with other interventions, compared with usual care and no treatment. Technology-based interventions were defined as any type of health-related services such as education, monitoring or treatment delivering via telecommunication technologies, internet, software or VR devices. The primary outcomes were pain and function. The secondary outcomes were quality of life, adherence, user experience and safety.

Data extraction

Trial details, including patients’ clinical profile, intervention, outcomes, attrition and adverse events (AEs), were recorded on a dedicated trial description form. Outcome data included mean score, mean difference (MD) between groups, odds ratios (ORs), risk ratios (RRs), standard deviations (SDs) and standard errors (SEs). Outcome data were extracted for short-term (immediate effect post-intervention to ≤3 months follow up), medium-term (3 to 6 months follow up) and long-term (≥ 6 months follow up) assessments. When more than one follow-ups were performed within each category, data from the shortest period of follow up were extracted.

Study methodological quality

The PEDro scale [20] was used to determine the methodological quality of each study. This 10-point scale is a valid assessment tool for the internal and external validity of randomised clinical trials, with acceptable reliability: intraclass correlation coefficient (ICCs) for inter-rater reliability of 0.56 for the total score; and 0.68 for consensus ratings [21, 22]. When available, quality scores were extracted from the PEDro database (www.pedro.org.au). Studies with a score of 7 or greater were considered “high quality” [23].

Quality of evidence

The Grading of Recommendations, Assessment, Development and Evaluation (GRADE) approach was used to appraise the quality of evidence for making clinical practice recommendations [24]. The quality of evidence was initially considered as high and downgraded based on five criteria: high risk of bias (e.g. > 25% of participants for studies with a PEDro score of ≤6), inconsistency of results (I2 > 50%), indirectness (comparison of different populations and interventions), imprecision (e.g. sample size < 400, 95% CI overlaps no effect) and publication bias (visual inspection of funnel plots and Egger’s regression test) [24].

Data synthesis and analysis

For the meta-analyses, whenever possible, outcomes were converted to a standard scale. For all variables with the same outcome, MDs or standardised MDs (SMD) with a 95% CI were calculated. Trials deemed clinically homogeneous were grouped according to 1) outcome measure, 2) follow-up duration and 3) surgery type. Between-trial heterogeneity was evaluated by visual inspection of the forest plots [25] and the I2 statistic (I2 < 50%: low to moderate; I2 ≥ 50%: substantial; I2 > 75% considerable heterogeneity) [26]. Random effect models were used to pool study results with considerable heterogeneity (i.e. I2 > 75%) [26]. Meta-analyses were performed using Review Manager, Version 5.3.

Results

Results of the search

In total, 21 RCTs (from 20 publications, N = 2971, mean age = 65.2 years old) were included after the screening of 8603 relevant studies retrieved from various databases. Figure 1 shows the PRISMA flowchart for the screening. The characteristics of included participants, interventions, outcomes and main findings are detailed in Table 1.

Fig. 1
figure 1

PRISMA flowchart

Full size image
Table 1 Characteristics of the included studies according to surgery and intervention types
Full size table

The average methodological quality of included studies was 5.8 (range: 2 to 8) on the PEDro scale (Table 1). A total of 7 studies (N = 1494, mean age = 65.8 years old) [27,28,29,30,31,32,33] were considered of high methodological quality (PEDro score ≥ 7). The most common methodological limitation was lack of blinding of the assessor observed in 10 of the 21 included trials (N = 1364); or therapist (16 trials, N = 1817).

Details of included studies

Type of technologies

A total of 11 RCTs (N = 1596) investigated telerehabilitation via telephone counselling/coaching (6 trials, N = 1070) or video-conferencing (5 trials, N = 526). Nine RCTs (N = 1120, 69.7% of all participants, mean age = 67.6 years old) included people having post-TKR rehabilitation [27, 30,31,32,33,34,35,36,37] and 2 RCTs (N = 234, mean age = 69.2 years old) included people undergoing post-THR rehabilitation [29, 38]. There is one study in TKR that used an additional accelerometer and gyroscopes to track patient’s body movement as part of the videoconference system [37].

Game-based therapy using video games, VR or biofeedback technologies was investigated in 5 trials (N = 232, mean age = 64 years old) of post-TKR rehabilitation (Table 1) [28, 39,40,41,42]. In 2 studies, participants used the Wii balance board for weight-bearing and balance exercise training [28, 40]. In another study, participants were equipped with two Wii game consoles on their legs to perform knee flexion or extension exercises [39]. One trial developed a 3-D avatar in an automatic virtual environment while using a robot-assisted walking device that simulated a normal walking process in a partial weight support condition [41]. In another recent study, participants were asked to row a boat using interactive VR with robotic-assisted passive knee range of motion (ROM) exercises [42].

There were 5 eligible studies (N = 1143) using web-based therapies, including educational software and interactive online platform, for participants following TKR (N = 594, mean age = 65.4 years) or THR (N = 549, mean age = 62.2 years). Three studies provide multimedia online training platform used by therapists for 149 TKR and 149 THR participants, respectively [43]. Two studies use asynchronous educational software designed for handheld devices for 29 TKR participants [44].

Efficacy outcomes

Pain

Our pooled analysis of 5 studies (N = 504) [27, 32, 37, 42, 44] showed that technology-assisted rehabilitation significantly improved pain measured on an 0–10-point visual analogue scale (VAS), compared to usual care, for people undergoing TKR (MD: -0.25; 95% CI: − 0.48, − 0.02) at 3 months follow up. Particularly, the subgroup analysis of telerehabilitation showed a statistically significant pain improvement (MD: -0.19; 95% CI: − 0.36, − 0.03) comparing with controls. However, both the effect sizes were too small to be of clinical significance (Fig. 2). There was no heterogeneity between the trials in telerehabilitation subgroup (P = 0.44; I2 = 0%). The quality of evidence is “moderate” due to serious risk of bias (Table 2). Due to the insufficient studies in each meta-analysis (< 10 studies), publication bias was not assessed.

Fig. 2
figure 2

Pooled effect of trials that investigated the effects of digital rehabilitation versus usual care on the visual analogue scale for pain: scale from 0 to 10, with higher scores indicating higher pain severity. Squares represent each individual study. Diamonds represent the pooled effect. Weight (%) represents the influence of each study on the overall meta-analysis. CI, confidence interval; TKR, total knee replacement; I2, heterogeneity of studies

Full size image
Table 2 Summary of the quality of evidence and strength of recommendation according to Grading of Recommendations Assessment, Development and Evaluation (GRADE) criteria
Full size table

Function

Time up and Go test (TUGT)

Our analyses pooling 2 studies (N = 207) [32, 37] showed that telerehabilitation significantly improved function, assessed via the TUGT (measured by second; less time spend indicates better function) [45] over a short term (2 weeks to 3 months), compared with usual rehabilitation for people following TKR (MD: -7.03; 95% CI: − 11.18, − 2.88). There was a substantial heterogeneity (P = 0.11; I2 = 60%). No difference was observed for those undergoing THR (MD: -0.70; 95% CI: − 1.47, 0.07) (Fig. 3). The quality of evidence was considered as “very low” because of the serious risk of bias, inconsistency and imprecision (Table 2).

Fig. 3
figure 3

Pooled effect of trials that investigated the effects of digital rehabilitation versus usual care on timed up and go test: assessed in second, with a higher number indicating worse functional ability. Squares represent each individual study. Diamonds represent the pooled effect. Weight (%) represents the influence of each study on the overall meta-analysis. CI, confidence interval; TKR, total knee replacement; THR, total hip replacement; I2, heterogeneity of studies

Full size image

6 minute walking test (6MWT)

There were two RCTs (N = 258) [31, 41] assessed mobility via 6MWT (measured by metre; longer distance indicates better mobility) [46] showing technology-assisted rehabilitation is not significantly superior to usual care in people who underwent TKR (MD: 29.36; 95% CI: − 6.99, 65.71) at the short-term (2 to 3 months) (Fig. 4). A high heterogeneity was detected (P < 0.01; I2 = 88%). The quality of evidence was downgraded to “very low” due to serious risk of bias, inconsistent results between 2 studies and indirectness of interventions (i.e. tele-rehabilitation and robotic-assisted VR were analysed together) (Table 2).

Fig. 4
figure 4

Pooled effect of trials that investigated the effects of digital rehabilitation versus usual care on six-minute walk test: assessed in metre, with a higher number indicating better mobility. Squares represent each individual study. Diamonds represent the pooled effect. Weight (%) represents the influence of each study on the overall meta-analysis. CI, confidence interval; TKR, total knee replacement; I2, heterogeneity of studies

Full size image

Western Ontario and McMaster Universities Osteoarthritis Index (WOMAC)

Four trials reported change in WOMAC on a 5-point Likert scale (standardised to 0–4 for each subscale) [47] (N = 746) [27, 30, 32, 33], 3 trials reported pain (N = 560) [30, 32, 33] and stiffness (N = 371) subscales [27, 32, 33]. There is low-to-moderate quality, downgraded for serious inconsistency and imprecision (data not shown), that telerehabilitation is not superior to usual care in improving WOMAC pain (MD: -0.09; 95% CI: − 0.22, 0.04; I2 = 15%; moderate evidence), function (MD: -0.05; 95% CI: − 0.16, 0.06; I2 = 34%; moderate evidence) or stiffness (MD: -0.07; 95% CI: − 0.32, 0.17; I2 = 67%; very low evidence) at the 3 months follow-up (Additional file 2: Fig. S1, S2 and S3).

Secondary outcomes

Quality of life

Six trials (TKR: N = 520; THR: N = 161) investigated the effect of telerehabilitation on quality of life (Table 1) [27, 33, 34, 36, 38, 44]. Meta-analysis was not feasible due to differences in completeness of reported data and inconsistent measurements. Two trials in people who underwent rehabilitation after TKR reported that telerehabilitation showed significant improvements on short form (SF)-36 mental component score (P < 0.01) [27] and physical function subscale (P = 0.031) [33], respectively. One study of THR showed physical function (P = 0.03), general health (P = 0.023) and mental health (P = 0.05) subscales of SF-36 were all significantly higher in the telerehabilitation group compared with the control group after 3 months, but all became non-significance at 9-month follow-up [38].

Adherence and user experience

Three RCTs of 472 people undergoing TKR investigated their compliance through an exercise diary [27, 31, 32]. One study showed the average time of daily home exercise in the telerehabilitation group (54.12 ± 5.71 mins) was significantly higher than the control group (48.95 ± 7.21 mins) [27]. Two studies showed no between-group differences in the number of exercise sessions finished daily [31, 32].

Four trials (N = 757) reported user experience and showed similar levels of satisfaction with both the intervention and the control [40, 43, 44, 48]. One trial of an educational software demonstrated positive user experiences, such as good clarity of instruction, ease of taking or sharing a video and ease of seeing their progress [44]. Another study of training software also received positive feedback from participants and therapists [43]. When participants were asked what they liked most about the application, no travelling to the hospital was cited by 57% and ease of access by 21% [44].

Safety

Moderate quality evidence from 3 RCTs (N = 667) showed the total number of serious adverse events (SAEs) were higher in the intervention group comparing to usual care (38 vs. 27) [29,30,31] (Table 2). However, there were no SAEs related to the intervention, while 2 events in the usual care group: one fell and one had wound bleeding during the first knee flexion exercise [31]. Of all the patients who had hospital admissions related knee issues, one in the usual care group had a leg blister below the TKR site, 3 in the usual care and 4 in the telerehabilitation group received manipulation under anaesthesia [30, 31]; one participant in the telerehabilitation group had thrombophlebitis [31]. One THR patient in the intervention group had a fever [29].

Discussion

Our review found that moderate-quality of evidence showed technology-assisted rehabilitation, in particular, telerehabilitation, had a statistically significant improvement in pain; and low-quality of evidence for the improvement in functional mobility in people undergoing TKR. The effects were however small and of arguable clinical significance. For THR, there is very limited low-quality evidence shows no significant effects. Pre-planned sub-group meta-analyses on study design (i.e. technology-based rehabilitation alone or in addition to usual care) were not performed due to insufficient studies. Most of the trials only had short-term follow-ups, therefore, the long-term effectiveness of technology-assisted rehabilitation was not ascertained.

Compared to previous studies in the field, our review has identified more than twice the number of the trials and most of the new studies added in our meta-analyses had higher methodological quality. For instance, the most recent systematic review only included 8 RCTs of post-TKR rehabilitation and 3 RCTs of post-THR rehabilitation and only provided a qualitative evaluation of those studies [15]. It concluded that the evidence was strong based on a PEDro score ≥5, which seems to be overestimated [49].

From the few studies that investigated user experience, there is a trend towards a positive impact of telerehabilitation, particularly, adherence to physical activities and compliance to rehabilitation programs [27, 31, 32]. Although the majority of the study population were older adults, their use of technologies, such as smartphone was quite high (59–49%) [50]. Similarly, in older adults with no prior experience with game consoles, most of them were highly motivated and expressed enjoyment in using the Wii Fit [39] and 86% of them were willing to continue the game therapy at home [40]. Some barriers were also demonstrated, such as poor internet connection at the participant’s home, delayed technology installation [32] and poor visual quality of the video-conference [32]. Additionally, older people may experience technological adoption barriers, such as concerns about the cost and battery life of the devices, as well as lack of familiarity with the technology [51]. These highlighted the need for cost-effective and power-efficient devices, elderly user-friendly design, sufficient training and ongoing customer support.

Importantly, the innovative devices or digital technologies should not be viewed as a distinct modality of care, but rather used as an aid/adjunct to bridge gaps or accelerate efficiency in existing healthcare delivery systems [52]. A study showed that telerehabilitation in addition to usual care was more favourable than usual care alone, whilst treatment delivered solely via telerehabilitation was equivalent to face-to-face intervention for functional improvement in people with MSK conditions [16]. In addition, validity studies reported a good agreement between face-to-face and telehealth assessment of MSK disorders of the knee (exact agreement of primary pathoanatomical diagnoses was 67%) [53]. Given the fact that technology could improve the healthcare accessibility and treatment adherence, despite its clinical effectiveness was similar comparing to conventional intervention, it still has a very promising role in circumstances when access and adherence are challenging.

Apart from some practical issues of licensure, there are potential challenges when implementing digital technologies in clinical practice. Firstly, the safety of the technology-assisted rehabilitation needs to be better understood. In our review, only a handful of studies reported AEs, although they all showed no increased harm. For game-based therapy, trials in the current review did not report any AEs, but it is reported that dynamic movements followed by different games can increase falls risks or other MSK injuries [54]. Safeguards should be taken pre-emptively when emergencies need to be solved virtually [55]. Healthcare providers embarking on careers in innovative technologies should be aware of current legal regulations to minimise risk [55]. Cost can also be a barrier when certain technology was first developed, thus, high-quality cost-effectiveness analyses are needed to demonstrate the long-term economic benefits.

There are several limitations to our review. Many studies did not perform a priori sample size calculations, which can increase the risk of underpowered (false-negative) results. Secondly, the trials used varied outcome measures which limited the pooling of results. Consensus on a set of suitable outcome measures needs to be reached for future trials. Furthermore, there is insufficient long-term follow up for ensuring the prolonged effects or safety. Lastly, a common risk of bias of the studies is a lack of blinding. As blinding of participants and therapists is not possible for most pragmatic trials, including those of technology-based rehabilitation interventions, future research should pay attention to the methodological aspects to minimise the biases.

Conclusion

There is moderate- to low-quality of evidence that current technology-enabled rehabilitation, in particular, telerehabilitation, showed most improvements in pain and function for people following TKR, comparing to usual rehabilitation. However, the effect size was too small to be clinically significant. Further high-quality studies are needed to demonstrate the long-term efficacy and safety of innovative health technologies, especially for post-THR rehabilitation.