The Xbox/Kinect use in poststroke rehabilitation settings: a systematic review

ABSTRACT Background: Active games based on virtual reality have been widely used in the rehabilitation of many clinical conditions. However, studies on the use of Xbox/Kinect are rare, and technology application in stroke treatment is not clear yet. Objective: To verify the outcomes (O) analyzed in randomized controlled trials (C; S) that investigated the use of Xbox/Kinect (I) in patients with stroke (P). Methods: This is a systematic literature review that meets PRISMA standards and the eligibility criteria according to the PICOS strategy. The search procedure was performed by two researchers. The research strategy was repeated in case of divergence. Effect size was calculated by Cohen's formula and Hopkins rank. The risk of individual bias was assessed using PEDro Score and Higgins Classification. Results: The main outcomes were postural balance and activities of daily living, with four studies addressing these variables. However, only one study showed the effect of Xbox/Kinect intervention on balance as large, as in two other studies evaluating manual dexterity and depression, respectively. Conclusion: The greater use of Xbox/Kinect in treating patients after stroke is in recovery of balance and motor function, and the evidence support its application. These findings enable the use of virtual reality technology through Xbox/Kinect in rehabilitation programs, focusing on postural balance and motor skills. However, conclusive results are still not possible. Therefore, caution in the use of this technology is required.

Active games based on virtual reality (exergames) have been often used in the prevention of falls of older people 1 and in the rehabilitation of patients with several pathological conditions 2 . Despite most commercial devices have not been designed for rehabilitation, they have been effective on several treatments. These activities require wide movements 3 , stimulate balance 4 and motor coordination 5 , and are double tasks 6 , with effects upon physical and cognitive skills.
Exergames have gained popularity with Nintendo Wii launching in 2006 7 . Interaction between the user and the games takes place through a wireless control and a force platform, which allow converting the movements into game commands, and most of the times the user is represented by an avatar 8 . Several investigations have been approaching the therapeutic application of exergames, with positive results in several capacities of healthy subjects and in the treatment of patients with physical disabilities, such as stroke, vestibular disorders, other balance alterations, orthopedic problems, among others 9 .
The Xbox 360 Kinect, launched in 2009, has also been used as a therapy. This Microsoft console eliminated users' direct contact with the hardware, and the movements are digitally captured through an infrared camera that enables subject's interaction with the virtual environment through their body image 10 . This console increased the odds of virtual reality in the rehabilitation. However, scarce studies use this technology in specific neurological treatments, as in patients with stroke. The most significant results were found in the improvement of physical functions, physical activity levels, and cognition 11,12 .
The post-stroke patient's disability varies depending on the region of the injury in the central nervous system. These sequelae might be motor, speech, language and/or cognitive deficits. Hemiparesis is common, which might result in the impairment of tonus, reflex and voluntary movements, postural balance and gait, hence causing damage in activities of daily living 13 . Therefore, the use of exergames is a viable strategy, because not only it is recreational, but it also stimulates cognitive functions and promotes movement through game interaction.
The use of Nintendo Wii in the rehabilitation of patients with stroke has been well documented in literature 9,[14][15][16][17][18][19] . Findings on this device and its effects show, mainly, the improvement of postural balance and motor functions. Meanwhile, rare studies have been investigating the use of Xbox/Kinect, especially in the treatment of stroke 2,12,17,20 in physical and mental rehabilitation. Consequently, this systematic review aims to analyze, according to the PICOS strategy, publications (C; S) that addressed the treatment of patients with stroke (P) using the Xbox/Kinect (I), in order to identify the main findings, assessment methods, and games (O).

Type of investigation
The present study is a systematic literature review that meets the patterns of the Preferred Reporting Items for Systematic Review and Meta-Analyses (PRISMA) 21 .

Eligibility criteria
Original studies published until October 2018 in English, Portuguese or Spanish were included. Inclusion criteria were used in the search, according to the PICOS 22 : P -subjects with stroke sequelae; I -treatment with Xbox Kinect; Ccomparison group(s); O -considering the aim of the present study, letter O from the acronym PICOS has not been a limiting eligibility factor; S -randomized clinical trials.

Strategy of paper search
Papers have been researched on PubMed, PEDro, Scopus, Cochrane, Web of Science and Grey literature databases. This procedure was conducted between June and October, 2018. No author was contacted, because all data were available in the articles. For each database, combinations of keywords and English words were used, considering that Portuguese and Spanish written works presented an abstract section. Keywords as "hemiplegia", "stroke", "cerebrovascular accident", "brain vascular accident", "paresis", "monoparesis", "hemiparesis", "Xbox" and "Kinect" were used and they should appear on the title, abstract or keywords of papers. Table 1 shows the research strategy in each database. Search procedure was performed by two browsers (TBXR; GCM), and the research strategy was repeated in case of divergence (Table 1).

Criteria for paper selection
Selection started by excluding duplicates in databases. Then, titles and abstracts of the remaining papers were analysed. Articles that did not meet the objective of the review, e.g. did not include post-stroke patients who had not performed treatment with Xbox and Kinect, without comparison groups and did not present randomized controlled trials (Figure 1), were not included.

Data collection
Selected papers were read according to the designed steps. General information of each paper was extracted and tabulated to identify details, objectives, variables, assessment tools, and main results.

Risk of individual bias
The PEDro scale 23 was applied to classify the mentioned papers according to an accurate methodology to evidence possible bias. This is an 11-item scale allocating one point to each. The first one is not considered. The items are: 1 -Eligibility criteria were specified; 2 -Subjects were randomly allocated to groups; 3 -Allocation was concealed; 4 -Groups were similar at baseline regarding the most important prognostic indicators; 5 -There was blinding of all subjects; 6 -There was blinding of all therapists who administered the therapy; 7 -There was blinding of all assessors who measured at least one key outcome; 8 -Measures of at least one key outcome were obtained from more than 85% of the subjects initially allocated to groups; 9 -All subjects for whom outcome measures were available received the treatment or control condition as allocated or, where this was not the case, data for at least one key outcome were analysed by "intention-to-treat"; 10 -The results of between-group statistical comparisons are reported for at least one key outcome; 11 -The study provides both point and variability measures for at least one key outcome. The Higgins visual scale 24 was also used to classify studies. This is a validity evaluation of the studies included in the systematic review. It is an eight-item scale that attributes studies qualitative studies (uncertain risk, low risk and high risk of bias). The items are: 1 -Random sequence generation; 2 -Allocation concealment; 3 -Blinding of participants and personnel; 4 -Blinding of outcome assessment; 5 -Incomplete outcome data; 6 -Selective reporting; 7 -Other sources of bias. The graph in Figure 2 presents the result of each item for all the studies.
This stage was merely for rating and was conducted by only one researcher expert in the scale.

Data analysis
A quantitative approach was conducted through effect size estimation, using the Microsoft Excel ® software, with Cohen's formula 25 and Hopkins classification 26 for differences between means 25 , scoring the effects as trivial (0-0.19), small (0.2-0.5), moderate (0.6-1.1), high (1.2-1.9), very high (2.0-3.9), almost perfect (>4), and perfect (infinite). Cohen's effect size is calculated according to the sample size, mean and standard deviation of both (Experimental and Control) groups 25 . Hopkins classification allows us to measure the magnitude of the observed effect 26 . Interventions conducted in the studies of this review have also been analyzed, by comparing experimental and control groups on session duration and games used in each study.

RESULTS
A total of 93 papers was collected from databases, and eight were selected for analysis ( Figure 1). Boxes 1 and 2 display extracted and tabulated data of each study.
Data of each paper were extracted and tabulated to identify details regarding the objectives, assessment tools, main findings, and conclusions ( Table 2).
Balance has been assessed in three of the eight studies using the Berg Balance Scale (BBS). Lee et al. 28 observed a significant improvement in Xbox and Control groups, both in post-training and follow-up, although there were no significant differences between groups with trivial and small effects (p=0.000/p=0.003; ES: 0.10/0.22). Park et al. 30 have also found significant improvements in both groups (p<0.05) and a difference between them (p<0.05), with moderate effect (ME: 0.65). Malik and Masood 31 have had great improvements in both groups (p=0.00), with a significant difference in the Experimental Group (p=0.001), with very large effect (ES: 2.1). Song and Park 32 used biofeedback to verify the weight distribution between lower limbs and to determine balance skill. Both groups have shown an improvement (p<0.05) with a significant difference in the group with Xbox training (p<0.05) and small effect (ES: 0.40). Stability limits have been assessed in two studies through Functional Reach Test (FRT) and biofeedback. Lee et al. 38 have shown no improvements in FRT in intra and inter-groups tests (p=0.187/p=0.442; ES: 0.18/0.55). Song and Park 32 have presented an improvement in the anterior (ES: 0.51; p<0.05) and the posterior (ES: 0.37; p>0.05) limits of stability, as well as significant differences in the Experimental Group (p<0.05). Both studies diverge on the results, despite their small effect size.
Mobility/gait has also been evaluated in three studies through Timed Up and Go (TUG), 10-meters Walking Test (10mWT), and other walking tests. Lee et al. 28 assessed simultaneously gait and a cognitive task (calculation) and have found a reduction in time performance for the group with Xbox training (p=0.009; ES: 0.03/0.00). However, both between and inter groups presented a difference on the trivial effect. Park et al. 30 , with the same tests, have shown an improvement in both groups (p<0.05), and the Experimental Group presented significant differences (p<0.05), with trivial effect on TUG (ES: 0.13) and small on 10 mWT (ES: -0.25). In Song and Park study 32 , both groups presented a time reduction in performance on TUG and 10 mWT (p<0.05), in which the Experimental Group had a significant difference (p<0.05), but with a small to moderate effect (ES respectively: -0.34; -0.79).
Functional independence and activities of daily living were assessed in four of the eight studies. Lee et al. 28 used Modified Barthel Index (MBI) (p=0.494/p=0.575; ES: 0.15/0.04) and Activities-specific balance confidence scale (ABC) (p=0.963/0.528; ES: 0.22/0.07), with no differences between them and inter/intra groups. The other authors used Functional Independence Measure (FIM). Afsar et al. 27 have found no differences in both groups after the intervention (p=0.40 and p=0.95) nor between them (p>0.677). Türkbey et al. 29 presented significant improvements in both groups (p=0.018) without differences between them (p>0.05), with a small effect size (ES: 0.32). Lee 33 observed significant differences in both groups after the intervention (p<0.05), but none between groups (p>0.05) and a small effect size (ES: 0.36).  Motor function was evaluated by four studies. Afsar et al. 27 assessed upper limbs through Fugl-Meyer Assessment (FMA), with a significant improvement in both groups (p=0.04). However, there was no significant difference between groups (p<0.57) and a and moderate effect (ES: 0.66). Park et al. 30 evaluated lower limbs through the same test, with significant improvements in both groups (p<0.05), but without differences between groups (p>0.05) and moderate effect (ES: 0.37). Sin and Lee 34 used the full version test and obtained significant improvements in both groups (p<0.05). The Experimental Group had a significant difference and moderate effect (p=0.041; ES: 1.06). Türkbey et al. 29 used Wolf Motor Function Test (WMFT) and have also obtained significant improvements in both Experimental (p=0.005) and Control (p=0.041) groups, besides a significant difference in the Experimental Group (p=0.014) and moderate effect (ES: 0.79) in the group that trained with Xbox Kinect.
Three studies assessed gross motor skill with the Box and Block Test (BBT). All the studies have shown that the Xbox group exhibited significant improvements in comparison with the Control Group. Afsar et al. 27 have noticed a significant improvement (p=0.04) in both groups, significant differences between groups (p=0.007) and prominence by the Experimental Group and large effect (ES: 1.58). Türkbey et al. 29 have found a significant improvement in both the Experimental (p=0.005) and Control (p=0.025) Groups, significant differences between groups (p=0.005) and the Experimental Group leading and a moderate effect (ES: 0.69). Sin and Lee 34 obtained a significant improvement in the Experimental (p=0.001) and Control Groups (p=0.005); significant differences between groups (p=0.043), mainly in the Experimental Group, and moderate effect (ES: 0.65). Table 3 shows all data collection.

Author (year) Score
Afsar et al. 29 7 Lee et al. 30 8 Türkbey et al. 31 8 Park et al. 32 7 Malik and Masood 33 6 Song and Park 34 6 Lee 35 7 Sin and Lee 36 6 Six studies presented a training volume higher in the Experimental Group rather than in the Control Group, most of the times the double of time 29,30,31,33,34 . Four of the eight studies used bowling 28,29,33,34 , and only two 33,34 used the same game battery, comprising 20,000 water leaks, bowling, boxing, rally ball and space pop. One study 32 was unclear about the games used, despite citing the selected packages (Kinect Sports 1 and 2, Kinect Adventure and Kinect Gunstringer).
Although authors had presented data as significant, most of the results have shown a small effect on the intervention with Xbox. The BBT conducted by Afsar et al. 27 Table 4. Results of tests conducted in all the above-mentioned studies have shown a significant improvement of balance skill, gait and patients' motor function after the intervention that varied between 12 and 40 sessions, twice to five times a week, with a duration of 30 to 90 minutes.
Methodological rigour of studies has been assessed with the PEDro scale 23 (Table 4), which shows scores between 6 and 8 points, and the Higgins visual scale 24 ( Figure 2) that introduces studies presenting unclear and low risk of bias.

DISCUSSION
This systematic review shows the details of interventions with Xbox/Kinect in patients with stroke. Postural balance 29,[31][32][33] , the ability to perform activities of daily living 28,29,30,34 , general motor function 34 , upper limb motor function 28 , lower limb motor function 31 , gross motor skill 28,30,34 , motor development 28,30 , gait 31,33 , quality of life 29 , muscle tone 34 , mobility 32 , amplitude of movement 34 , strength 33 , and depression 32 have been investigated. Despite the probable clinical application related by the authors, findings have not shown, individually, differences between the intervention with Xbox/Kinect and standard treatment with functional exercises and kinesiotherapy. This finding shows that there might be a likely resemblance between the intervention with Xbox/Kinect and standard rehabilitation, especially when analyzing the effect size between groups (small/large) in different variables. The use of virtual reality-based games provides a multisensory retro feeding that improves motor refinement [35][36][37][38] , possibly through neuroplasticity mechanisms 38,39 . The improvement of physical and cognitive functions varies and occurs at short and long terms on stroke patients 1,11 . Specifically, benefits of exergames include the increase of gait velocity, balance and mobility and motivation to treatment 20 . However, the guidelines from the American Heart Association and the American Stroke Association 13 call the attention to the use of this technology in the rehabilitation of stroke patients. It presents B level, class IIb, in gait improvement and spatial orientation, and IIa class on upper limb movement practice. B level indicates that the studies have been conducted with limited populations and simply randomized or not. Class IIa indicates favorable recommendations to the procedure, with some opposite evidence, whereas class IIb indicates that the utility or effectiveness of the procedure is less established and involves higher opposite evidence 13 .
A systematic review 17 has evidenced no advantage of the rehabilitation with exergames in relation to the standard therapy on upper limb mobility and activities of daily living. According to the authors, there is insufficient evidence to conclude on the use of virtual reality-based games on gait velocity and postural balance. These data corroborate partially with the effect size findings of the present study. Nevertheless, the studies present no discussion about the limitations on the use of Xbox/Kinect; therefore, we suggest that the use of this device in certain treatments (e.g. neurological patients) might be difficult due to the therapist' s impossibility to be closer to the patient' s body, as the digitalization of user' s body image suffers interferences with another person' s presence. Maybe this might be one of the reasons for the scarce amount of investigations on this device. In case of this interference, the sensor needs to be calibrated again as mentioned by Lee 33 .
All the studies assessed in this review presented arguments in favor of exergames in the rehabilitation of these patients. Even though there was study heterogeneity, clinical effects were observed and should be considered. A large 28 and moderate 30,34 effect size has been found on the motor function assessed with the BBT on two studies that used this device. Furthermore, one of the studies presented depression with a large opposite effect 32 .
Balance presented an effect size from trivial to large 29,31,32 , which shows a great heterogeneity of results. This variability can also be observed in gait, which has presented an effect size varying from trivial to moderate 29,31,33 . Therefore, there are no consistent data to sustain the efficacy of rehabilitation with Xbox in many variables.
The largest effect size was found on postural balance, in the study where participants performed 18 intervention sessions, three times a week, using the games: 20,000 water leaks, reflex ridge, and river rush 32 . However, the duration of the treatment session was not mentioned. The largest number of sessions was 40, among all the studies analyzed with a 30-minute duration, using Kinect Adventure, Kinect Sports 1 and 2, and Kinect Gunstringer 33 . Authors have found a large effect of improvement on depression symptoms and moderate effect on gait improvement 33 .
Limitations of the present study include the impossibility of a data meta-analysis, as studies present different variables and methods. On the other hand, the classification of volunteers through questionnaires and divergent tests hampers comparison of results. The use of different games does not allow the determination of which have influenced greater upon results, as although similar, they demand distinct skills and tasks. The small number of randomized and controlled studies published on this scope increases the disparity of papers investigated herein, which reduces the possibility of an estimation of a grouped effect.
Other important limitations of the included studies were: lack of analysis regarding the injury location; number and type of sequelae. Furthermore, only one study analyzed depression 31 , and none of them verified anxiety symptoms, which could affect rehabilitation engagement negatively. Furthermore, these studies did not analyze comorbidities. Procedures of electromyography analyses were also not clear, which hindered the replication of these methods.
The use of Xbox/Kinect for the rehabilitation of stroke patients is a recent topic. The first controlled and randomized studies were published in 2013. Moreover, the studies here selected presented small samples, and the majority comprised less than 30 volunteers. The greater use of Xbox Kinect with more significant results in the treatment of stroke patients was in the recovery of motor function and postural balance. Nevertheless, conclusive findings on these and other variables were not possible yet, which increases the necessity for caution with this device in the rehabilitation. Further investigation with larger samples is recommended.