Are we There Yet? Immersive Virtual Reality to Improve Cognitive Function in Dementia and Mild Cognitive Impairment

Meelad Sayma, BMBS, BSc; Remco Tuijt, BA, MSc; Claudia Cooper, BMBS, MRCPsych, MSc; Kate Walters, BMBS, BMedSci, MRCGP, MMed, MSc, PhD


Gerontologist. 2020;60(7):e502-e512. 

In This Article


Study Selection

Our search identified a total of 2,824 citations, Figure 1 breaks down the sources of obtained articles. One hundred seventy-five articles remained that were primary studies addressing virtual reality and dementia/MCI. These articles were broken down into four groups: (i) papers addressing cognition and other dementia/MCI outcomes, (ii) papers addressing physical rehabilitation only, (iii) VR as a modality of assessment, (iv) other. Sixty-nine articles specifically addressed VR as a modality of assessing and diagnosing dementia or cognitive impairment. Seventy articles specifically addressed physical rehabilitation only, most of which looked at gait abnormalities in Parkinson's disease. Fourteen articles addressed two or more of these groups (but not cognition in dementia). Although nine articles did not address any of these topics, most of these focused on training health care professionals or family members to manage or understand dementia or MCI. At this stage, we noted that most, if not all, articles would not meet our criteria to be determined as "fully immersive." As a result, articles were included if they established that a clear attempt was made to immerse the participant in the intervention, such as isolating the participant in a darkened room, or using enlarged screens to deliver the intervention.

Figure 1.

PRISMA chart demonstrating the screening process of papers in the systematic review.

Thirteen articles remained for potential inclusion. Out of these 13 articles, four were determined to use nonimmersive virtual reality and therefore excluded. Four other articles included all older adults in their participant group, and not just people with MCI or dementia. Due to the heterogeneity of included articles, we were unable to conduct a meta-analysis of the included data, as a result descriptive analysis was conducted. Studies excluded participants with severe medical or psychological conditions or had some sort of disabling neurosensory conditions making it difficult to deliver therapies. Of the four studies excluded because the interventions were judged to be nonimmersive, one of these studies utilized augmented reality delivered via a phone app and therefore we judged not to give an immersive experience (Bormans, Roe, & De Wachter, 2016). Three of these studies solely used an iPad-based exer-game used in participants living rooms, deeming this experience un-immersive (Anderson-Hanley et al., 2018; Lee, Lee, & Song, 2015; Wall et al., 2018).

Study Quality and Risk of Bias Within Studies

Most studies utilized extremely small samples sizes (n = 1 to n = 57). However, both randomized-controlled pilot trials were methodologically sound when assessed against the CASP critical appraisal checklist (Maggio et al., 2018; Optale et al., 2010). It was more difficult to assess the noncontrolled studies, and based on the aims of the studies, inclusion of a control group would have been useful to clarify and demonstrate outcomes. The two studies that blinded study personnel were not explicit about how this was done. The appendices show CASP quality appraisal with additional appraisal for certain bias risks described in Supplementary Material.

Description of Included Articles

Only a small number of studies were included based on inclusion and exclusion criteria (Maggio et al., 2018; Manera et al., 2016; Moyle, Jones, Dwan, & Petrovich, 2018; Optale et al., 2010; White & Moussavi, 2016). As a result, we did not exclude any studies based on their quality or risk of bias. A lower threshold for immersion was also accepted, with articles using semi-IVR techniques accepted for inclusion in this review. All five articles were included for final analyses. One article originated from Australia, one from France, two from Italy and one from Canada. Two articles were randomized-controlled feasibility studies, two articles were noncontrolled pilot trials and one article described a case study of an intervention with one participant. Most included studies were published between 2016 and 2018, with only one article published in 2010. Of the studies that used a control condition, two used face-to-face therapy and one used a paper-based activity.

The mean length of time the participants received the intervention for was 58 days (range n = 1–183), with some participants only receiving one session in total and some receiving three sessions per week for over 3 months. In four out of the five studies, all participants completed the intervention, this is apart from Optale and colleagues, where five out of the 36 participants dropped out of the study before the intervention was complete (three participants died and two left the home; Optale et al., 2010). Table 3 shows a summary of the characteristics of included studies, with Supplementary Data in Appendix B giving a detailed breakdown of each article.

Participants and Baseline Cognitive Function

A total of 119 participants were included across the five included studies (range 1–57). Two studies utilized the Mini-Mental State Examination (MMSE) as a key aspect for inclusion of participants to the study and/or baseline measurement of participants. Maggio and colleagues (2018) included participants with an MMSE ranging between 11 and 26, whereas Manera and colleagues (2016) included patients with an MMSE ranging between 16 and 28, having a slightly higher threshold of cognitive function for inclusion. Optale and colleagues (2010) utilized a Verbal Story Recall (VSR) test to include participants, with participants included if they scored below 15.76. Moyle and colleagues (2018) and White and Moussavi (2016) used documented diagnoses as part of their inclusion criteria, using diagnosis of dementia and MCI, respectively, whereas White and colleagues also recorded a Montreal Cognitive Assessment (MoCA) score of 24 on their single patient. No studies reported recruitment response rates.

Mean age of participants was 76.8; 54% were female and 46% male. Moyle and colleagues (2018) and White and Moussavi (2016) did not report the level of education of their participants. Maggio reported that 25% of participants received a primary school education, 55% high school/secondary, and 20% had a university education. Optale and colleagues (2010) stated that participants were standardized for education without stating the levels of education. Manera and colleagues (2016) reported 1.8% of participants had an unknown level of education, 1.8% had no education, 33% described only primary education, 32% were educated to secondary level, and 32% to university level.

Use of IVR Interventions

Two studies used full IVR therapy and three studies used a semi-immersive approach to VR therapy. A broad range of hardware was utilized throughout the studies. Two studies utilized an HMD (Optale et al., 2010; White & Moussavi, 2016). Three studies utilized large screens or "video walls" as their visual hardware, two of these screens were considered "interactive" (Maggio et al., 2018; Manera et al., 2016; Moyle et al., 2018). Two studies utilized infrared motion sensors to track participants' movements (Maggio et al., 2018; Moyle et al., 2018). One study used a wireless mouse to track motions (Manera et al., 2016). The remaining studies used HMD motion tracking, and Optale and colleagues also used a modified office swivel chair equipped with trackers to track and feedback movement (Optale et al., 2010; White & Moussavi, 2016). It appeared all studies utilized some form of audio feedback; however, only Optale and colleagues and White and colleagues appeared to use headphones to deliver this feedback, the rest used external speakers. No studies utilized haptic feedback as part of their hardware suite. The mean immersion score, based on the criteria set out in Table 1, was 7.4 out of 10 (range 5–9). None of the studies described whether they assessed the level of presence described by their participants.

It is important to note that in one study, authors described external factors that may have affected immersion and thus participants' presence in their VE (Moyle et al., 2018). The study took place in two separate care homes and the authors note that results were very different by facility. One facility offered a quiet dark room with dimmed lighting with little background noise to conduct the intervention, whereas the other facility offered a room near a social room with a budgerigar present.

Interventions were extremely heterogeneous in their theme, task, and goals. Maggio and colleagues utilized a variety of VR-based games as the basis of their intervention. Each game targeted a specific domain of cognitive function, for example, a game focused on memory training asking participants to memorize and recall objects seen in a previous VE. No other study took a cognitive domain-specific approach. White and colleagues developed a memory game, asking participants to memorize and identify "target windows" in a virtual house; however, this memory game appears to have been utilized to improve all cognitive domains—assessing improvements using MoCA or MMSE. No other interventions were developed in the White paper to address other cognitive domains. Manera and colleagues developed a task aimed at differentiating colors and patterns, of varying difficulties. Optale and Moyle developed a VR scenario for participants to navigate through, with Optale and colleagues allowing participants to activate film clips as they navigate their path. Studies gave varying detail in explaining theoretical considerations underlying the development of their VR interventions, which mostly focused on the multisensory immersion freeing the participant from external distraction.

Maggio and colleagues stipulated that VR provides multisensory stimulation to create a realistic environment and improve motivation and the adhesion of participants to rehabilitation, while grounding each individual activity in a different cognitive domain. Manera and colleagues described that VR creates a more realistic environment, ability for faster feedback, and a greater degree of personalization than paper tasks. They went further when describing their task design was based on the principles of the classical cancelation task, employed for instance in the Attention Process Training—an intervention designed to rehabilitate attentional problems in people with brain injuries. Optale described that VR therapy "frees the patient" from external distraction and encourages selective attention, without fully outlining the proposed mechanisms or grounding of their specific tasks. White and colleagues stipulated that navigational training tasks may reduce Alzheimer's disease pathology—justifying their task design without a specific focus on the use of VR. The theoretical grounding behind the Moyle and colleagues' intervention was not clearly described.


Two studies focused on acceptability of IVR therapy: Manera and colleagues, and Moyle and colleagues. Manera and colleagues reported that participants were highly satisfied with IVR therapy and reported low levels of anxiety and fatigue. Participants who were more satisfied with the IVR condition compared with the paper-based control condition, reported feeling less secure in using the IVR condition. Moyle and colleagues reported statistically significant lower apathy after using the IVR condition and participants appeared to express more enjoyment having used the IVR condition compared with baseline participant data from a different cohort of people with dementia (1.4 vs. 2.8, p = .008; Moyle et al., 2018). Maggio and colleagues, White and colleagues, and Optale did not report data on acceptability.

Feasibility and Potential Effectiveness

Three studies recorded estimates for the effects based on cognitive functioning. Maggio and colleagues, and Optale and colleagues described statistically significant improvements in MMSE scores (+2.15 [p = .014] after 2 months of three weekly sessions of intervention and +0.74 [p = .044] after 6 months of three weekly sessions, respectively), whereas White and colleagues described no statistically significant improvements in the participant's MoCA score, offering three weekly sessions for 7 weeks. Maggio and colleagues, and Optale and colleagues also described improvements in their participants' general depression scores (GDS) along with MMSE and VSR (see summary table; GDS = −0.23 [p = .812] and −1.05 [p = .025], respectively; Maggio et al., 2018; Optale et al., 2010). The study by White and Moussavi (2016) offered no objective improvements in cognitive scoring of their single participant, and they however describe some subjective improvements observed by the participant's wife. In addition, Moyle and colleagues did not use standardized global cognitive scoring systems to assess improvements in cognitive domains; however, they did report improvements in alertness (p < .001) based on observed emotional rating scale scores, akin to the cognitive domain of attention (Lindsley, 1988). Manera and colleagues did not report changes in cognitive functioning.