Effectiveness of Virtual Reality Systems to Improve the Activities of Daily Life in Older People

(1) This review aims to evaluate the effectiveness of treatments with virtual reality systems (VRSs) on the functional autonomy of older adults versus conventional treatment. (3) Methods: Systematic review and meta-analysis. An electronic data search was carried out, following the PRISMA statement, up to February 2020. We combined results from clinical trials using VRSs for the improvement of basic and instrumental activities of daily living. The guidelines of the Cochrane Handbook for Systematic Reviews of Interventions were followed for calculations and risk of bias. The Grading of Recommendations Assessment, Development and Evaluation (GRADE) was used to assess the quality of evidence. (4) Results: The final analysis included 23 studies with a population of 1595 participants. A moderate, but clinically significant, effect was found for basic activities of daily living (BADLs), (Standard Medium Deviation, SMD 0.61; 95% CI: −0.15–1.37; P < 0.001). A small effect was found for instrumental ADLs (Instrumental Activities of daily living, IADLs) (SMD −0.34; 95% CI: −0.82–0.15; P < 0.001). Functional ambulation was the BADL which improved the most (SMD −0.63; 95% CI: −0.86, −0.40; P < 0.001). (5) Conclusion: The use of VRSs is an innovative and feasible technique to support and improve the functional autonomy of community-dwelling older adults. Due to the very low quality of the evidence for our main outcomes, the effects of a VRS on the BADLs and IADLs are uncertain. Clinical trials of a higher methodological quality are necessary to increase the level of knowledge of its actual effectiveness.


Introduction
The morphological and functional changes associated with aging can have a negative impact on the performance of activities of daily living (ADLs). It has been estimated that 20-25% of community-dwelling people over 75 may experience limitations in their capacity to perform ADLs [1,2]. This dependency is a predictor of frailty and institutionalization [3,4]. In a recent study, the American Occupational Therapy Association (AOTA) supported the need to build a friendly environment and to implement prevention programs which include physical, social and leisure activities for older persons with a high and stable intrinsic capacity [3].
In the last decade, there has been a growing interest in the use of virtual reality systems (VRSs) to implement programs for supporting the functional ability of older adults through physical exercise [5][6][7], cognitive [8,9] and social interventions [10]. VRSs allow for the re-creation and control of virtual everyday environments, and for the planning of safe, ecological and motivating treatment activities.
Unlike the real world, the virtual world can be adapted and adjusted to the capacities and needs of each person, and thus it provides great flexibility of experiences and virtual tasks, which include the sequences of functional movements necessary to achieve autonomy in real-life activity. The analysis and control of all the activity's virtual elements by the occupational therapist enable virtual achievements which could not be attained in real life, due to either disability or to environment restrictions. These achievements increase motivation, commitment and adherence to the rehabilitation process.
The possibility of generating exercises that can be repeated over time, selecting the frequency and intensity, enables compliance with the principles of motor learning and neuroplasticity, as well as the generalization of this learning [11]. A new performance context is therefore created, with which the person identifies and gets involved. Thus, they manage to carry out actions which are not always possible in their original setting and which can help to improve their occupational performance [12]. These characteristics make virtual reality-based technology an appropriate rehabilitation technique for different people.
Various reviews have been published in recent years about the effectiveness of VRSs for older adults. These have shown positive effects on physical functioning [6,13,14] and cognitive functioning [15,16]. However, no previous review has analyzed the effects of VRSs on autonomy in the performance of activities of daily living in older adults. In addition, many of the previous reviews have focused on older populations with neurological disorders, excluding older adults without disabilities. Thus, the preventive potential of VRSs is still not known.
The aim of this systematic review and meta-analysis is to critically evaluate the effectiveness of VRSs for improving the functional autonomy of older populations and for preventing disability. These findings can be useful for future research and intervention projects aiming at preventing disability and promoting the occupational performance of older adults.

Materials and Methods
The Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) statement was employed to carry out this review [17]. The Grading of Recommendations Assessment, Development and Evaluation was also used [18].

Search
Six databases were searched electronically (Web of Science, OT Seeker, Guideline National Clearing House, databases of the Spanish National Research Council (CSIC), Scopus and Cochrane), up to February 2020. No date, study design, language or age limits were set in order to increase the sensitivity of the search. The electronic search was completed with a hand search of the bibliographical references of the studies included. The search terms were developed by a methodology consultant from the Institute of Health Sciences (ICS, Spain). See Appendix A for key search terms.
The inclusion criteria were the following: (1) studies with older adults over 60 years of age who performed activities of daily living independently, (2) studies which used VRSs in order to improve the performance of ADLs [19], (3) studies whose outcome measures included the assessment of ADLs, (4) studies considered as peer-reviewed scientific literature (Level I, Level II or Level III) [20] and (5) studies with a simple majority in the 25 items of the CONSORT checklist [21].
The studies excluded were those with participants under the age of 60 with neurological or osteoarticular disorders which severely affected functional independence, studies which did not use VRSs directly and studies which did not report on changes in occupational performance.

Review, Selection and Data Extraction
Two independent reviewers selected the studies. Discrepancies were resolved by a third reviewer. The data extraction protocol was based on the PRISMA items [17] and the Cochrane Handbook for Systematic Reviews of Interventions. The form included information regarding methodological characteristics, participants, healthcare resource, intervention, results on functional ability and conclusions. Study authors were contacted to retrieve unreported data.

Summary Measures and Statistical Analysis
Various outcome measures were used. Objective measures were included which evaluated ADLs in an overall way (Barthel Index [22], Functional Independence Measure (FIM) [23], Late Life Function and Disability Instrument (LLF&DI) [24] and the Instrumental Activities of Daily Living Scale [25]). Quantitative assessment tools were also included which assessed specific ADLs, such as ambulation or transfers (Functional Gait Assessment (FGA) [26], Timed Up and Go (TUG) [27], Four Step Square Test (FSST) [28], 6-min walk test (6MWT) [29], 8-foot up and go [30] and Five Sit to Stand [31]). The effect magnitude was measured using Hedges's adjusted g standardized mean difference (SMD), with its confidence interval at 95% (95% CI). The total effect size, weighted by the sample size of the studies, was calculated using the inverse variance method and a random effects model. Its 95% CI and its statistical significance were calculated using the Z-test. The effect size was interpreted using Cohen's criteria for pooled estimates [32]; SMD > 0.20, small; SMD > 0.50, medium; and SMD > 0.8, large effect.

Assessment of Level of Evidence of the Set of Studies
The GRADE system [18] was used, considering eight factors to reduce or increase the level of evidence. The factors for downgrading the level were (1) risk of bias; (2) inconsistency; (3) indirect evidence; (4) imprecision by the calculation of the optimal information size (OIS); (5) publication bias. The factors considered for upgrading the level of evidence were (1) large effect size (SMD >0.8); (2) dose-response effect; (3) control for confounding factors in the individual studies.

Risk of Bias Assessment in Individual Studies
The risk of bias of each article was assessed independently by two reviewers using the items included in Review Manager (RevMan), version 5.3. (The Nordic Cochrane Centre, The Cochrane Collaboration: Copenhagen, Denmark, 2014). Based on five "risk of bias" items, we determined that studies at: • a low risk of bias were those in which all items were assigned a low risk of bias; • an unclear risk of bias were studies in which one or more items were found to be at an unclear risk of bias; and • a high risk of bias were studies in which one or more items were found to be at a high risk of bias.

Publication Bias
Publication bias was assessed using a funnel plot created with RevMan, complemented with a DOI plot created with METAXL. Egger's method, Begg's test with Epidat 3.1 and the Luis Furuya-Kanamori (LFK) index were used. An LFK index ≤ 1 was considered as no asymmetry, >1 ≤2 as minor asymmetry and >2 as major asymmetry.

Results
In total, 1083 articles were identified. After removing duplicates, 985 records were screened, with 52 manuscripts selected for complete reading. The full text review included 23 studies applying VRSs to improve the functional ability of independently living older adults over 60 years of age. Figure 1 shows the flow chart of this selection. The total number of participants was 1595, with an age range of 60-96 years.

Characteristics of Included Studies
The included studies were published within the last nine years (2011-2020). That indicates that the use of VRSs is a very recent research area. The aim of these studies was to analyze the effect of VRSs on two areas of occupation and IAVDs; functional ambulation, the capacity to make transfers, cognitive function, physical condition and quality of life. Data about design, sample size, characteristics of the population, type of intervention and assessed outcomes are shown in Appendix B.
Five studies evaluated the overall impact of VRSs on BADLs [33][34][35][36][37] and three studies examined IADLs [38][39][40]. We also identified six studies which assessed the effectiveness of VRSs on the capacity to make transfers and twenty-one studies which reported on their effectiveness in improving functional ambulation.
The studies were randomized and quasi-randomized clinical trials. Eighteen studies were Level I (randomized controlled trials) and five studies were Level II (two groups, nonrandomized studies). The control groups received interventions consisting of physical exercise sessions (six studies) and health education (three studies). The remaining fourteen studies did not design any intervention for the control group. Only five studies carried out a follow-up after the intervention [38,39,[41][42][43].

Intervention with VRS
The use of VRSs was the main intervention technique. Seven studies used VRSs specifically designed for rehabilitation [34,39,42,[44][45][46][47]. The rest of the studies used virtual reality active video games (Nintendo ® Wii, XBox ® , Sony ® PlayStation and Xavi Sport ® ). No study used head-mounted displays. Appendix C shows the devices and exergames used in each article. The duration of the interventions was from 1 to 24 weeks, with sessions between 20 and 50 min.

Characteristics of Included Studies
The included studies were published within the last nine years (2011-2020). That indicates that the use of VRSs is a very recent research area. The aim of these studies was to analyze the effect of VRSs on two areas of occupation and IAVDs; functional ambulation, the capacity to make transfers, cognitive function, physical condition and quality of life. Data about design, sample size, characteristics of the population, type of intervention and assessed outcomes are shown in Appendix B.
Five studies evaluated the overall impact of VRSs on BADLs [33][34][35][36][37] and three studies examined IADLs [38][39][40]. We also identified six studies which assessed the effectiveness of VRSs on the capacity to make transfers and twenty-one studies which reported on their effectiveness in improving functional ambulation.
The studies were randomized and quasi-randomized clinical trials. Eighteen studies were Level I (randomized controlled trials) and five studies were Level II (two groups, nonrandomized studies). The control groups received interventions consisting of physical exercise sessions (six studies) and health education (three studies). The remaining fourteen studies did not design any intervention for the control group. Only five studies carried out a follow-up after the intervention [38,39,[41][42][43].

Intervention with VRS
The use of VRSs was the main intervention technique. Seven studies used VRSs specifically designed for rehabilitation [34,39,42,[44][45][46][47]. The rest of the studies used virtual reality active video games (Nintendo ® Wii, XBox ® , Sony ® PlayStation and Xavi Sport ® ). No study used head-mounted displays. Appendix C shows the devices and exergames used in each article. The duration of the interventions was from 1 to 24 weeks, with sessions between 20 and 50 min.

Effectiveness of VRSs on Activities of Daily Living
Eight studies provided sufficient data to report on the effectiveness of VRSs in improving the performance of activities of daily living. As Figure 2a shows, the effect size of VRSs in improving the performance of BADLs is moderate but significant; the total SMD was 0.61 (95% CI: −0.15; 1.37). Four of the studies which assessed effectiveness on BADLs obtained a large effect (d > 0.8) [33][34][35]37] and two studies found a medium effect [33,38]. However, the effectiveness of VRSs was smaller for the three studies which assessed IADLs [38][39][40]. The total SMD was −0.34 (95% CI: −0.82; 0.15) and the effect found was small (Figure 2a). This indicates that the interventions with VRSs had a small non-significant effect (p = 0.3) for improving the performance of IADs compared to the control group.

Effectiveness of VRSs on Functional Ambulation and Transfers
Thirty-two outcomes from twenty-two studies provided sufficient data to analyze the effectiveness of VRS in the improvement of functional ambulation in older adults. The total SMD was −0.63 (95% CI: −0.86, −0.40, p = 0.001). This indicates that the interventions with VRSs had moderate, but significant effects ( Figure 2a). The group of participants which received treatment sessions with a VRS increased their level of performance in ambulation compared to the control group that received traditional physical exercise.
Six outcomes of transfer ability from six studies were analyzed. The effect size of the treatment programs with VRSs on the ability to make transfers from sitting to standing and vice versa was small, with an SMD of −0.23 (95% CI: −0.71, −0.25; P < 0.001). Only two studies obtained a large effect [24,35], whereas the rest obtained a low effectiveness. Figure 2b shows the SMD for each study, as well as the total SMD. Figure 3 shows the risk of bias assessment for each study. The risk of bias was moderate for the studies which addressed the effectiveness of VRSs in the performance of ADLs. Only one study did not do random sequence generation, and seven out of eight studies lacked clarification about the allocation concealment. Five studies (62%) did not blind outcome assessors and the incomplete outcome data bias was low.

Risk of Bias of the Individual Studies
The risk of bias for functional mobility and transfers was high ( Figure 3). In functional ambulation, four studies had a high risk of bias for random sequence generation [35,[48][49][50], none of the studies could blind participants, only six blinded outcome assessors [37,38,41,47,51,52] and only one had a low risk of bias for allocation concealment. On the other hand, loss to follow-up was significant in only one study (Keogh et al., 2014

Effectiveness of VRSs on Activities of Daily Living
Eight studies provided sufficient data to report on the effectiveness of VRSs in improving the performance of activities of daily living. As Figure 2a shows, the effect size of VRSs in improving the performance of BADLs is moderate but significant; the total SMD was 0.61 (95% CI: −0.15; 1.37). Four of the studies which assessed effectiveness on BADLs obtained a large effect (d > 0.8) [33][34][35]37] and two studies found a medium effect [33,38]. However, the effectiveness of VRSs was smaller for the three studies which assessed IADLs [38][39][40]. The total SMD was −0.34 (95% CI: −0.82; 0.15) and the effect found was small ( Figure 2a). This indicates that the interventions with VRSs had a small nonsignificant effect (p = 0.3) for improving the performance of IADs compared to the control group.

Effectiveness of VRSs on Functional Ambulation and Transfers
Thirty-two outcomes from twenty-two studies provided sufficient data to analyze the effectiveness of VRS in the improvement of functional ambulation in older adults. The total SMD was −0.63 (95% CI: −0.86, −0.40, p = 0.001). This indicates that the interventions with VRSs had moderate, but significant effects ( Figure 2a). The group of participants which received treatment sessions with a VRS increased their level of performance in ambulation compared to the control group that received traditional physical exercise.
Six outcomes of transfer ability from six studies were analyzed. The effect size of the treatment programs with VRSs on the ability to make transfers from sitting to standing and vice versa was small, with an SMD of −0.23 (95% CI: −0.71, −0.25; P < 0.001). Only two studies obtained a large effect [24,35], whereas the rest obtained a low effectiveness. Figure 2b shows the SMD for each study, as well as the total SMD. Figure 3 shows the risk of bias assessment for each study. The risk of bias was moderate for the studies which addressed the effectiveness of VRSs in the performance of ADLs. Only one study did not do random sequence generation, and seven out of eight studies lacked clarification about the allocation concealment. Five studies (62%) did not blind outcome assessors and the incomplete outcome data bias was low.

Risk of Bias of the Individual Studies
The risk of bias for functional mobility and transfers was high ( Figure 3). In functional ambulation, four studies had a high risk of bias for random sequence generation [35,[48][49][50], none of the studies could blind participants, only six blinded outcome assessors [37,38,41,47,51,52] and only one had a low risk of bias for allocation concealment. On the other hand, loss to follow-up was significant in only one study (Keogh et al., 2014).

Heterogeneity
The studies which assessed the overall performance of ADLs (I 2 = 79%) and transfers (I 2 = 77%; chi-squared p = 0.0007) showed a high heterogeneity. In the studies which analyzed functional ambulation, heterogeneity was moderate, with I 2 = 55% (chi-squared p = 0.004). The forest plots (Figure 2a,b) show a significant variability between the studies which favored the experimental intervention.

Publication Bias
The funnel plot and the DOI plot ( Figure 4) show signs of publication bias. Both plots present asymmetry, mainly in those studies with fewer participants but a greater impact on functional performance. The LFK index showed minor asymmetry both for ambulation (LFK = 3.78) and for transfers (LFK = 0.36). The statistical significance of Begg's test for ambulation was p = 0.0235 and that of Egger's test was p = 0.0057. The statistical significance of Begg's test for transfers was p = 0.13 and that of Egger's test was p = 0.11.

Heterogeneity
The studies which assessed the overall performance of ADLs (I 2 = 79%) and transfers (I 2 = 77%; chi-squared p = 0.0007) showed a high heterogeneity. In the studies which analyzed functional ambulation, heterogeneity was moderate, with I 2 = 55% (chi-squared p = 0.004). The forest plots (Figure 2a,b) show a significant variability between the studies which favored the experimental intervention.

Publication Bias
The funnel plot and the DOI plot ( Figure 4) show signs of publication bias. Both plots present asymmetry, mainly in those studies with fewer participants but a greater impact on functional performance. The LFK index showed minor asymmetry both for ambulation (LFK = 3.78) and for transfers (LFK = 0.36). The statistical significance of Begg's test for ambulation was p = 0.0235 and that of Egger's test was p = 0.0057. The statistical significance of Begg's test for transfers was p = 0.13 and that of Egger's test was p = 0.11.

GRADE Quality of Evidence
The quality of evidence was low to very low for studies which used VRSs to optimize ADL performance, and moderate to low for studies using VRSs to improve functional mobility. Appendix D specifies the reasons for decreasing the level in each of the study groups.

Discussion
This systematic review and meta-analysis examined the effectiveness and the quality of evidence of VRSs to support the functional autonomy of non-disabled older adults. Although there are still few studies that have analyzed the effectiveness of intervention programs based on virtual reality on older persons' occupational performance, the results are encouraging. Seventy-three percent of studies showed significant improvement in the functional ability of this population group. However, the GRADE grading of recommendation was low to very low due to the risk of bias, inconsistency and imprecision of the analyzed data. Therefore, these studies should be treated with caution.
Additionally, clear signs of publication bias were found in studies which assessed functional mobility (LFK: −3.3). A detailed analysis showed that the studies with fewer participants had better results. This indicates the possibility that there might be unpublished studies which did not show significant effects of VRSs.
In most studies, the intervention with VRS was applied through commercial exergames or active video games; only seven studies used VRSs specifically designed for rehabilitation. This may be due to the accessibility and low cost of active game consoles. Moreover, commercial exergames offer a

GRADE Quality of Evidence
The quality of evidence was low to very low for studies which used VRSs to optimize ADL performance, and moderate to low for studies using VRSs to improve functional mobility. Appendix D specifies the reasons for decreasing the level in each of the study groups.

Discussion
This systematic review and meta-analysis examined the effectiveness and the quality of evidence of VRSs to support the functional autonomy of non-disabled older adults. Although there are still few studies that have analyzed the effectiveness of intervention programs based on virtual reality on older persons' occupational performance, the results are encouraging. Seventy-three percent of studies showed significant improvement in the functional ability of this population group. However, the GRADE grading of recommendation was low to very low due to the risk of bias, inconsistency and imprecision of the analyzed data. Therefore, these studies should be treated with caution.
Additionally, clear signs of publication bias were found in studies which assessed functional mobility (LFK: −3.3). A detailed analysis showed that the studies with fewer participants had better results. This indicates the possibility that there might be unpublished studies which did not show significant effects of VRSs.
In most studies, the intervention with VRS was applied through commercial exergames or active video games; only seven studies used VRSs specifically designed for rehabilitation. This may be due to the accessibility and low cost of active game consoles. Moreover, commercial exergames offer a virtual re-creation of environments and challenges that are highly immersive, appealing and enjoyable for the user, thus increasing commitment to the task. In line with other reviews [53,54], our results suggest that the use of commercial virtual reality consoles, such as the Nintendo Wii ® , PlayStation ® and Xbox Kinect ® , can be beneficial and feasible in improving the functional autonomy of older persons.
We found that the use of VRSs is more effective and has greater evidence in the field of occupation corresponding to BADLs and ambulation than in the performance of IADLs. The reason for these results may be that IADLs are more complex activities, both in their treatment and their assessment, because they depend largely on the social and environmental context [55]. These factors are not addressed in the intervention programs based on VRSs. In this regard, recent reviews [56,57] conclude that improvement in physical performance does not always translate into improvement in IADLs. Nevertheless, we found a direct link between an increase in physical functional test scores (TUG, 6MWT) and the performance of basic activities. We assume, from these results, that intervention with VRSs could help older persons to live independently and safely in their own home and could facilitate productive aging.
This review and meta-analysis has several strengths. Its methodology did not set a time or language limit on the publications and made use of a wide range of databases in order to increase the sensitivity of the search. It also included articles of high methodological quality: 78% of the studies were Level I and the remaining 22% were Level II articles.
However, the articles included had a number of important methodological limitations: the sample size, the risk of bias and the post-intervention follow-up. Only two studies had more than 100 participants, and sixteen studies had fewer than 50 participants. Regarding the risk of bias, none of the studies could blind participants and only seven studies blinded assessors. The lack of post-intervention follow-up made it impossible to know the long-term and preventive effect of VRSs on the functional ability of older persons.

Conclusions
As a conclusion, our findings suggest that the use of VRSs is an innovative and feasible technique to support and improve the functional autonomy of community-dwelling older adults. However, clinical trials of a higher methodological quality are necessary to increase the level of knowledge of its actual effectiveness.

Conflicts of Interest:
The authors declare no conflict of interest. Table A1. Key search terms.

Category Key Search Terms
Population Aging, elderly, geriatr *, gerontology, late life, older adults, oldest, seniors, old age.
Intervention: VRT ("Wii *" OR "gaming technology" OR "active videogame" OR nintendo OR "video game*" OR exergam* OR Wii OR kinect OR Xbox OR "Play Station" OR "Xavi Sports") OR ("virtual reality" OR virtual-reality tecnhology OR "Virtual navigation" OR "Virtual environment" OR "Simulated reality" OR "virtual rehabilitation" OR vr-based OR "virtually simulated" OR "use-computer interface")   Downgrading for inconsistency was considered because I 2 = 79%.

Inconsistency results
No serious inconsistency Downgrading for inconsistency was not considered because I 2 = 58%

Indirectness evidence No serious indirectness
In none of the studies were there substantial differences between the population, intervention or outcomes measured in the studies and those established in the systematic review.

Indirectness evidence No serious indirectness
In none of the studies were there substantial differences between the population, intervention or outcomes measured in the studies and those established in the systematic review.

Imprecision
Moderated ↓1 level The OIS was calculated, the resulting OIS being 276 participants with MDS = 0.3. The number of patients included in the meta-analysis is 212, moderately lower to the OIS. Downgrading for imprecision was considered.

Imprecision
No serious imprecision The OIS was calculated, the resulting OIS being 276 participants with MDS = 0.3. The number of patients included in the meta-analysis is 1097, far superior to OIS. Downgrading for imprecision was not considered. Downgrading for publication bias was considered because the DOI plot shows marked asymmetry with LFK = 3.3 (major asymmetry).

Magnitude of effect
Moderated ↓1 level SMD = |0.31| Upgrading for large magnitude of effect was not considered.

Magnitude of effect
Not serious SMD = |0.58| Upgrading for large magnitude of effect was not considered.

Dose-response gradient
Not considered. Dose-response gradient Not considered.

No plausible confounders
Not considered because studies in this meta-analysis are not observational.

No plausible confounders
Not considered because studies in this meta-analysis are not observational.

Quality rating
Very low quality: we are moderately confident in the effect estimate; the true effect is likely to be close to the estimate of the effect, but there is a possibility that it is substantially different (GRADE Working Group grades of evidence)

Quality rating
Moderate-low quality: we are moderately confident in the effect estimate; the true effect is likely to be close to the estimate of the effect, but there is a possibility that it is substantially different (GRADE Working Group grades of evidence)