This study received ethics approval from the Institute of Visual Informatics, Universiti Kebangsaan Malaysia. The approval was granted on July 2, 2024, with the reference number UKM.IVI.600-4/6/P130610.

All participants provided written informed consent before the start of any experimental procedures. The researchers clarified the study’s objectives, methods, possible risks (including cybersickness symptoms), and potential benefits. They also made it clear to participants that they could withdraw from the study at any point without facing any consequences.

All personally identifiable information and harvested data were fully anonymized to ensure respondent privacy. Data were stored securely on a password-protected server, and only central research team staff had access.

Participants did not receive any monetary or nonmonetary compensation for their participation. Their involvement in this study was entirely voluntary.

This study was designed as a pilot study using a 2×2 factorial design (display mode: VR/MR×feedback: soothing/intense) to effectively explore the feasibility of varying exergame configurations before a large trial. Participants were sequentially assigned to 1 of the 4 conditions using a rotating 1:1:1:1 schedule upon enrollment. The primary goal of this multicondition design was not to conduct formal hypothesis testing between groups, which would require a much larger sample. Instead, its purpose was to ensure that our usability and feasibility assessment covered a range of potential interaction styles. This approach allows us to identify any major negative outcomes (eg, excessive cybersickness or frustration) that might be unique to a specific condition, providing critical insights for optimizing a single, refined protocol for future, larger-scale home-based studies.

A research team initially recruited 86 community-dwelling older people who were capable of living independently, via media such as community promotions at community centers and newspaper advertising locally. Such a process was used to maintain the sample’s characteristics closely aligned with those of the target end user group for future home-based VR/MR Tai Chi exergames [13].

In this pilot phase of research, screening led to a dropout of 16 participants (16/86, 18.6% of the original sample) who could not accommodate to the virtual environment or who experienced severe dizziness. Notably, this screening process suggests that the final sample was skewed to some degree to VR tolerance, that is, comprised individuals better predisposed to accommodate virtual environment exposures. This characteristic is relevant in estimating the extent to which the findings generalize to a broader community of older individuals.

We performed an analysis of the 16 participants who were excluded due to a failure to adapt to VR. This group comprised 11 (68.8%) women and 5 (31.2%) men, a gender distribution similar to that of the retained sample. However, a systematic difference was evident in age: 43.8% (7/16) of participants excluded were aged older than 75 years, a proportion more than double that of the sample maintained (15/70, 21.4%). This finding suggests that older age could be a determining factor in intolerance to VR in this population.

The aforementioned recruitment and screening processes adhered to clearly defined criteria. The inclusion criterion was enrollment of participants aged 60 years and older [14,15]. Exclusion criteria encompassed any known adverse reactions to VR experiences and severe cardiorespiratory or musculoskeletal diseases that would significantly impede participation in the experimental procedures, severe vision loss, and other health conditions unsuitable for the use of immersive devices (eg, vertigo and epilepsy).

Self-report alone was considered insufficient; therefore, a brief experiential screening was recommended before unsupervised use.

Data regarding demographic information, health background, and daily exercise habits were collected from the 70 valid participants using a self-developed “basic information questionnaire.” The detailed characteristics are presented in Table 1.

The 70 valid participants included in the final analysis exhibited certain demographic tendencies: they were predominantly female (52/70, 74.29%) and largely comprised “younger” older adults aged 60 to 75 years (55/70, collectively 78.56%). In terms of health status, although a considerable proportion (40/70, 57.14%) self-reported chronic conditions such as cardiopulmonary diseases, they were also individuals who actively managed their health, with a high percentage (66/70, 94.29%) undergoing regular physical examinations. This characteristic of “living actively with illness” suggests that the sample group might hold a more open attitude toward health-promoting interventions, but it also necessitates that such interventions ensure safety and appropriateness.

This pilot group reported very high levels of physical activity, such that the vast majority (66/70, 94.29%) used a daily exercise regimen and participated in a broad range of activities, such as walking, square dancing, and Tai Chi. Overall, this sample is a subgroup of predominantly female, aged between 60 and 75 years, who exercise on a regular basis, approach health care management in an active manner, and have good tolerance to VR environment (as a consequence of preliminary screening that eliminated 18.6% (16/86) who reported dizziness while experiencing VR). These characteristics are relevant to explain why this sample reports such a high level of interest and positive response to the VR Tai Chi game. While these same characteristics will limit how widely the investigation can be generalized to broader older groups, such as those who exercise less or who possess poorer VR tolerance.

The VR Tai Chi exergame used in this study was designed for older individuals. Its fundamentals can be equally analyzed using the MDA (mechanics, dynamics, and aesthetics) approach [16,17].

It guides users through 16 selected Tai Chi exercises. It is interactive via a head-mounted display (HMD) and hand controllers reacting to head and hand movements. Users must synchronize their pose to 3 hotspots (head and both hands) provided in the on-screen instruction. A movement only proceeds after a successful success window, and a movement can save up to 3 hotspot holds (left hand, right hand, and HMD) before advancing. If users complete the entire sequence of 16 movements ahead of schedule, they will repeat the sequence immediately, beginning with the first movement. This generous success condition was a conscious design choice intended to provide a sense of accomplishment. It has 2 display modes, VR and MR, and can come equipped with either soothing or upbeat music and vibration feedback. Additionally, players can skip the current move or pause at will during gameplay.

Playing involves players’ efforts at mirroring exemplified postures, receiving immediate multisensory feedback, and regulating progression during a trial lasting a maximum of 5 minutes. One key dynamic reveals itself in the gamification mechanics: participants can choose to meet only the bare minimum threshold needed for progress (“single hotspot for 3 s”) to progress effortlessly or shoot for immaculate matching across all hotspots to progress to a higher “accuracy” level. This allows participants to balance challenge against victory on a preference and ability level. It is this dynamic that later results regarding user experience build upon.

The game aims to evoke multidimensional aesthetic experiences, such as a “sense of competence” from successfully completing movements, a “sense of challenge” from improving performance, feelings of “relaxation or motivation” engendered by different musical feedback, and a “sense of immersion” provided by the VR/MR environments. These subjective experiences are intended to encourage physical exercise in older users in an engaging manner, and their achievement was assessed using the Game Experience Questionnaire (GEQ) [18,19]. Design philosophy appears to be interested in desirable affective effects and perceived competence corresponding to self-determination theory principles, such that satisfying the need for competence can improve intrinsic motivation.

This study used the Meta Quest 3 (Meta Platforms, Inc) HMD for presenting both VR and MR conditions [20,21]. The choice of specific, commercially available hardware, such as the Meta Quest 3, enhances the transparency and potential replicability of the study.

Exercise took place in an area of approximately 5 m², comparable to a typical living room, to ensure home feasibility and safety [22,23].

When participants arrived at the laboratory, they first completed an informed consent process and a general information questionnaire. Next, a short adaptation training session was completed to enable participants (and especially those without any training in VR/MR) to become familiar with both the immersive environment and how to use the equipment. This adaptation step is a valuable step in acclimating older adults to new technology, such as VR, because it can be used to help reduce technology anxiety and achieve a minimum baseline level of comfort and competence in how to use the equipment before beginning the intervention [24,25].

A visual summary of the participant flow and 2×2 design (VR/MR×soothing/intense) appears in Figure 1. This approach ensured that group sizes remained balanced and that potential confounding factors, such as personal preference or prior experience, were minimized. Participants followed a fixed 16-movement sequence; each 5-minute session could end before all movements were finished due to the time limit or restart at the first movement if the sequence was completed early. After this orientation, each participant performed 2 formal 5-minute gameplay sessions under their assigned experimental condition (VR/MR×soothing/intense feedback). The better performance out of the 2 sessions was recorded for analysis. All single-session metrics (eg, posture, accuracy) were computed per session; the higher of the 2 sessions was retained for analysis. This procedure ensured that every participant experienced the complete set of movements while also minimizing the impact of first-trial unfamiliarity on the final outcome [26].

VR provided a fully virtual scene; MR overlaid guidance onto the real room (Figure 2). “Soothing” used gentle cues; “intense” used more salient multimodal cues; task rules were identical [27].

The experimental protocol was observed and supervised in a one-on-one capacity by trained researchers to be able to control and respond to any new sensory stimulus or sensation of disorientation. Subjects were informed they could withdraw from participation at any time if they experienced any type of discomfort or abnormal level of fatigability. Such precautions are particularly relevant in research among older individuals and potentially disorienting technologies such as VR. To regulate fatigue and potential dizziness between sessions, exposure was scheduled in 2 approximately 5-minute sessions, and participants could rest and pause whenever needed.

This was a single-visit, supervised laboratory study, consisting of the 2 gameplay blocks described earlier, with no home-based or longitudinal follow-up. Postsession questionnaires captured immediate responses; repeated exposures to control for novelty were not implemented.

This research took a multidimensional perspective to assess usability and feasibility, where primary data were drawn from:

We took an integrated approach, triangulating subjective experience (GEQ), physical comfort (VRSQ), and actual performance (game logs) to provide a sound foundation for a unified and rigorous assessment of the game’s usability and feasibility.

We adopted a target engineering level of 80% movement precision to be used as an internal and exploration benchmarking.

Internal consistency was evaluated on the present sample using Cronbach α and Spearman-Brown split-half coefficients. As several components (tension/annoyance, challenge, and negative affect) exhibited pronounced floor effects with near-zero variance, internal consistency coefficients for these scales were not estimable; for these, we report distributional summaries only. Reliability results for all measures are presented in Table 2. Per the GEQ guidelines, we dropped the story-related immersion item (item 3), leaving 5 items for this component; reliability values in Table 2 reflect this adapted set.

The key game performance indicators were defined as follows: score was the system-generated composite points; accuracy (%) was the proportion of frames where hotspots were within tolerance; and posture (count) represented the total number of successful hotspot holds. Per movement, the player aligns the left and right controllers and the HMD with 3 predefined hotspots; holding any hotspot for ≥3 seconds is counted as one successful hotspot hold. The 16-movement sequence loops whenever it is completed before the 5-minute limit; therefore, posture is an open-ended count (no fixed cap) within the time window. For interpretability, 48 holds correspond to one full loop (16 movements×3 hotspots).

Consistent with the a priori plan, no confirmatory between-condition hypothesis testing (main effects or interaction) was conducted. To enhance transparency, per-condition descriptive summaries (medians and IQRs for the 4 cells) are provided in Table 3.

Handled sporadic missingness via median imputation only; no multiple imputation was performed [29].

Data normality was examined using the Shapiro-Wilk test [30,31]. Most key variables were found not to be normally distributed (Table 4). Nonparametric methods were consequently unanimously selected across hypothesis testing to yield robust and valid inferences. That is, median dimension scores on the GEQ were compared against a theoretical midpoint score of 2.0 using 1-sample Wilcoxon signed-rank tests. Wilcoxon tests correspondingly compared the median against a theoretical score of 0 on the VRSQ symptom scores (rated 0‐3) to determine whether symptoms were distinctly present. With the exploratory nature of this item-level analysis, no adjustment for multiple comparisons was made on these 9 tests. Spearman ρ was used to examine associations between key factors. A level of significance was predetermined at an α of .05.

Descriptive statistics summarized the GEQ, the VRSQ, and game log indicators as median (IQR) [19,32].

As reflected in Table 5, subjective gaming experience as measured using the GEQ was highly positive. The median levels on positive aspects like “positive affect” (median 4.0), “flow” ( median 4.0), “competence” (median 3.8), and “sensory/imaginative immersion” (median 3.8) all far surpassed the theoretical midpoint of the scale (P<.001 in all). Medians on negative aspects such as “tension/annoyance” (median 0) and “negative affect” ( median 0) were below midpoint to a statistically significant degree (P<.001 in both). The median level on “challenge” was low to a statistically significant degree as well (median 1.4, P<.001).

The physiological comfort of the screened participants was assessed using the VRSQ. Overall, the incidence and severity of reported symptoms were low. Detailed item-level medians, statistical test results, incidence rates, and severity scores for reporters are presented in Table 6. The P values in the table should be interpreted as exploratory signals of which symptoms were most frequently endorsed, rather than as confirmatory hypothesis tests.

A look at the in-game objective performance data listed in Table 7 reveals wide individual variations among older adult participants. Participants demonstrated a fluid interaction rhythm with a median time spent per action of 6.4 seconds. However, large variations across major indicators were apparent due to wide IQRs seen across score and accuracy measurements (Table 7). Median movement accuracy was 68.5% (IQR 54.2%‐82.3%). Scores higher than 48 indicate that participants completed more than a single loop (or accrued numerous hotspot holds within a movement) within a 5-minute period.