This study conducted a systematic review to synthesize diverse research, focusing on how XR-based exercise interventions for older adults can address exercise adherence through service design strategies. This review followed the transparent, structured process of the PRISMA (Preferred Reporting Items for Systematic Reviews and Meta-Analyses) guidelines (Multimedia Appendix 1) [48] to ensure reproducibility and methodological rigor.

Studies were included if they met all of the following criteria: (1) they enrolled participants with a mean age of older than 50 years—this standard is based on the World Health Organization’s definition of older people in resource-limited settings, where life expectancy is generally shorter [49]; (2) they focused on XR-based interventions involving physical activity, exercise, or rehabilitation; (3) they reported the use of at least one service design method (eg, personas, cocreation, user journey mapping, user-centered design, or participatory design) during the development or implementation of the intervention; (4) they reported outcome data related to exercise adherence, including at least one of the following dimensions: frequency, duration, intensity, or accuracy of exercise performance; and (5) they used or proposed quantitative, qualitative, or mixed method evaluation strategies, with evidence of feasibility, usability, or preliminary outcomes.

Studies were excluded if they met any of the following conditions: (1) they focused solely on cognitive training or entertainment without an exercise component; (2) they did not include older adults as the primary target population; or (3) they were editorials, commentaries, opinion pieces, or lacked sufficient methodological detail or relevance to the research question.

A structured search was conducted across 6 academic databases: PubMed, Scopus, Web of Science, CINAHL, PsycINFO, and the ACM Digital Library. The search strategy combined Boolean operators and thematic clusters covering: XR technologies (“virtual reality” OR “augmented reality” OR “mixed reality” OR “extended reality” OR “immersive technology”), service design methods (“service design” OR “user journey” OR “journey mapping” OR “experience map” OR “co-design” OR “participatory design” OR “user-centered design” OR “human-centered design”), and the target population and activity (“older adult” OR “elderly” OR “geriatric” AND “exercise” OR “physical activity” OR “rehabilitation”). Searches were limited to peer-reviewed papers published in English between January 2020 and July 2025. Full-text availability was required for eligibility assessment and data extraction; records for which the full text could not be accessed were excluded at the screening stage. Gray literature sources and trial registries were not systematically searched, as this review focused on peer-reviewed studies that provided sufficient methodological detail on service design activities and XR intervention implementation. The full database-specific search strategies are detailed in Multimedia Appendix 2.

All retrieved records were imported into a Zotero library (version 7.0.24; Corporation for Digital Scholarship), and duplicates were removed. Two independent reviewers (JN and SW) screened the titles, abstracts, and full texts of the remaining papers against the predefined inclusion and exclusion criteria. Discrepancies were resolved through discussion; if consensus could not be reached, a third reviewer (YY) was consulted.

Data extraction was performed using a standardized Microsoft Excel 2021 spreadsheet to ensure consistency across studies. Two reviewers independently extracted data on study characteristics, service design strategies, UX, and adherence outcomes. Discrepancies were resolved through discussion until consensus was reached. To classify the application stages of service design strategies, this study adopted the TiSDD framework [40]. Table 1 summarizes the definitions and typical activities of each stage, which served as the reference for coding service design applications in the included studies.

Data synthesis was organized around 3 research questions and divided into 4 sections, including study characteristics, service design outcomes, UX outcomes, and adherence outcomes. This thematic framework enabled a narrative synthesis that examined how service design tools were used, how they supported continuity and UX, and what types of adherence outcomes were reported.

To assess the quality of the studies, critical appraisal was conducted using the Mixed Methods Appraisal Tool (MMAT; version 2018; McGill University) [50]. The MMAT is a specially designed tool that can be used to assess the quality of different types of studies in the same review, including qualitative, quantitative, and mixed methods studies. The quality of all studies was assessed by the first reviewer (JN) and checked for accuracy by 2 other authors (YY and SW). In line with MMAT guidance, the tool was not used to score or exclude studies, but to support the interpretation of the findings [50].

The total number of identified records was 973. After duplicates were removed, 949 records remained, and these were screened for title and abstract relevance. The number of excluded studies after title and abstract screening was 921. The full texts of the remaining 28 papers were read and analyzed, of which 19 were excluded due to factors such as the absence of exercise-related data, duplication, non–older adult populations, lack of service design elements, or lack of XR-based interventions. As a result, 9 studies were included in the final review (Figure 1).

This review included 9 studies published between 2020 and 2025, conducted across 9 countries: Italy, Portugal, the Netherlands, the United Kingdom, Germany, Spain, Canada, Cyprus, and China. Of the 9 studies, 67% (6/9) were mixed methods designs [51-56], 11% (1/9) was a mixed methods study with an embedded randomized controlled trial [57], 11% (1/9) was a quasi-randomized controlled trial [58], and 11% (1/9) was a nonrandomized quantitative study [59]. A total of 242 participants were involved across all studies, with sample sizes ranging from 6 to 75. The mean age of participants ranged from 52.8 to 86.8 years. A total of 89% (8/9) included both male and female participants [51-54,56-59], while 11% (1/9) did not report gender information [55].

All 100% (9/9) studies used VR as the XR modality [51-59]. A total of 67% (6/9) of studies reported specific XR headset models [51-53,55,56,58], while 33% (3/9) did not specify the hardware used [54,57,59]. A total of 78% (7/9) of studies described the XR interaction modality [51,52,54-58], whereas 22% (2/9) did not report interaction details [53,59]. A total of 100% (9/9) of studies reported some form of XR supervision and support during intervention delivery [51-59]. A total of 33% (3/9) of studies reported information on XR cybersickness or tolerability [51,54,55], while 67% (6/9) did not report tolerability-related data [52,53,56-59].

Exercise types included 44% (4/9) upper limb rehabilitation interventions [51,52,55,56], 22% (2/9) gait or balance training [54,58], 11% (1/9) lower limb training [57], 11% (1/9) adapted physical activity [59], and 11% (1/9) multidomain physical-cognitive rehabilitation [53].

Intervention settings included 44% (4/9) hospital-based studies [51-53,57], 11% (1/9) university or research laboratory [58], 11% (1/9) biomedical research institute [54], 11% (1/9) long-term care residence [56], and 11% (1/9) home-based environment [55], 11% (1/9) study used a mixed-setting intervention across nursing homes, hospitals, and home environments [59] (Table 2).

Based on the MMAT assessment, the overall methodological quality of the included studies was acceptable.

Most studies demonstrated adequate quality across their respective methodological domains. The qualitative components generally met MMAT criteria, showing appropriate study design, data collection, and coherence between data, analysis, and interpretation.

Among quantitative studies, one randomized controlled trial met all MMAT criteria. Several nonrandomized studies showed limitations related to confounding, as confounders were not clearly accounted for in the study design or analysis.

Most mixed methods studies provided a clear rationale for using a mixed methods design and demonstrated integration between qualitative and quantitative components. However, some studies showed limited methodological transparency, reflected by “can’t tell” ratings on selected MMAT criteria.

No studies were excluded on the basis of quality appraisal. Detailed MMAT ratings for each study are provided in Multimedia Appendix 3 [51-59].

The included studies evaluated UX using 5 methods. Five of the 9 (56%) of studies used standardized usability questionnaires, such as System Usability Scale, The Single Ease Question, Simulator Sickness Questionnaire, Questionnaire for User Interaction Satisfaction, and NATU Quest (Assistive Technology Usability Questionnaire for people with Neurological diseases), to assess system usability and task difficulty [51,52,54,55,59]. Five of the 9 (56%) studies evaluated emotional and immersive experience through tools like Flow Short Scale, Physical Activity Enjoyment Scale, Exergame Enjoyment Questionnaire, Physical Activity Affect Scale, Visual Analogue Scale, Observed Emotion Rating Scale, and Slater-Usoh-Steed Questionnaire [51,52,55,56,58]. Two of the 9 (22%) studies captured motivation and engagement using the intrinsic motivation inventory and a 4-point engagement scale [56,57]. Two of the 9 (22%) of studies examined self-efficacy and safety perceptions using tools such as ABC-6/8, the Bandura scale, and custom safety items [55,58]. Six of the 9 (67%) studies used qualitative feedback methods, including interviews and open-ended forms [52-57].

Participants generally reported a positive UX. Five of the 9 (56%) studies reported that interventions were enjoyable and immersive, with participants frequently highlighting high levels of engagement and emotional benefits [51,52,54,56,58]. Three of the 9 (33%) of studies indicated perceived usefulness and value for rehabilitation, with self‑report measures reflecting strong motivation and recognition of exercise benefits [55,57,59]. Two of the 9 (22%) studies included feedback from clinical staff who found the interventions both feasible and encouraging [53,57]. They highlighted outcomes such as increased patient confidence, improved mood, and reduced staff workload. However, 33% (3/9) of studies also pointed to negative aspects, including fatigue, discomfort from wearing head-mounted displays, and a decrease in enjoyment during cognitively demanding or extended sessions [51,54,57] (Table 4).

This review analyzed exercise adherence outcomes reported across the included studies. Given the heterogeneity in adherence definitions and measurement approaches, no conversion or standardization of adherence metrics was performed. Instead, adherence outcomes were synthesized by grouping indicators according to their underlying behavioral constructs, allowing comparison within similar categories while highlighting dimensions that remain underevaluated.

Attendance-based adherence indicators were the most consistently reported. Five of the 9 (56%) studies reported attendance-related indicators, including completion and dropout rates, with completion rates ranging from 80% to 100% and dropout rates varying between 2% and 13% [53,55-58]. In this review, references to “high adherence” were primarily derived from attendance-related indicators, as these were the most frequently and consistently reported measures across studies.

Performance-based adherence indicators were reported less consistently and captured different aspects of task execution and progression. Four of the 9 (44%) of studies monitored exercise accuracy, including task precision, movement quality, and success rates through in-game metrics or trajectory tracking systems [51,52,54,56]. Furthermore, 33% (3/9) of studies discussed performance progression across sessions, including observed improvements or declines in task success rates and response times [52-54].

Self-reported adherence indicators were the least frequently reported. One of the 9 (11%) studies assessed self-reported perceptions of adherence, with 65.8% (49/75) of participants indicating that the system supported their physical activity routines [59].

Across the 4 core adherence components synthesized in this review, including frequency, duration, intensity, and accuracy [60]. Seven of the 9 (78%) of studies clearly defined exercise frequency [51,52,54-58]. Furthermore, 89% (8/9) of studies reported the duration of exercise, including both session length and program length [51-58]. Three of the 9 (33%) studies addressed exercise intensity [55,56,59]. Four of the 9 (44%) of studies captured the accuracy of exercise performance [51,52,54,56]. None of the studies included long-term follow-up to assess sustained exercise adherence.

No studies were excluded from the review due to unclear adherence reporting; however, the variability and incompleteness of adherence measures limited cross-study comparability and constrained conclusions regarding long-term adherence (Table 5).

This review synthesized evidence from 9 XR‑based exercise interventions targeting older adults and revealed a field that was promising yet constrained by structural and methodological limitations. Most studies grounded their service design strategies in participatory, co‑design, and user‑centered approaches [51-57,59], frequently combined with iterative prototyping [51,54-59]. However, service design processes were often incomplete, with most studies concentrating on early research and prototyping stages, showing limited engagement in ideation [51] and progression to implementation. Only 3 studies progressed to the Implementation stage [53,54,57]. Although multiple stakeholders were involved and participatory principles were widely adopted, their contributions were largely confined to early-stage design decisions and usability refinement, rather than extended to shaping long-term service integration or delivery models. This uneven coverage across service stages constrained the translation of interventions into sustained service offerings. When interpreted through behavior change frameworks such as Capability, Opportunity, Motivation–Behavior [61] and behavior change techniques [62], the limited continuity of service design across later stages helps explain why sustained exercise behavior was rarely examined. While early-stage co-design and prototyping supported initial engagement and motivation, fewer interventions addressed opportunity-related factors, such as integration into everyday contexts, reinforcement strategies, and long-term support mechanisms, which are essential for maintaining behavior over time.

UX outcomes were generally positive. Participants reported pleasant experiences, high levels of immersion, and increased motivation in questionnaires, usability evaluations, and interviews. However, some studies also noted negative adverse effects, such as fatigue, headset discomfort, and decreased enjoyment during challenging or long tasks [51,54,57]. Overall, most evaluations relied on subjective reports, with limited use of objective measures, such as physiological or behavioral data. This reliance on self-reported short-term experience indicators limits the extent to which positive UX can be interpreted as evidence of sustained engagement. While enjoyment and immersion appear effective in supporting initial participation, the lack of longitudinal and objective measures constrains understanding of how these experiences translate into continued exercise behavior over time.

Most studies reported high levels of exercise adherence, though inconsistently across dimensions, despite the use of a 4D adherence framework [60] in our synthesis; few studies comprehensively reported across all dimensions. Frequency and duration were the most reported measures of exercise adherence; however, few studies compared them with established exercise guidelines. Most studies conducted minimal monitoring of exercise intensity, and only 1 applied validated physiological measurement. While some VR systems included metrics related to movement accuracy, the studies did not integrate these metrics into the formal outcome framework for exercise adherence. This fragmentation of adherence dimensions impeded cross-study comparison and masked the contribution of specific design features to behavioral engagement. Most studies assessed exercise adherence using non‑standardized, short‑term measures. Attendance and task completion were the most commonly used surrogate measures of adherence; some studies also reported adherence through in-game performance metrics, performance progression over time, or self-reported perceptions of adherence, but none adopted a standardized long-term assessment framework. As a result, the consistently high adherence rates reported across studies should be interpreted cautiously, as they primarily reflect short-term participation rather than sustained exercise continuity. The absence of standardized, multidimensional, and longitudinal adherence assessments limits insight into whether engagement persisted beyond the intervention period.

Notably, 8 studies conducted interventions in hospital, laboratory, or long‑term care settings [51-54,56-59], and 1 study conducted its intervention at home [55], with no studies implementing community‑based deployment, highlighting a critical gap for real‑world scalability. The traditional health management model mainly relied on hospital or home-centered interventions, and the increasing demand for population aging and chronic disease management posed greater challenges to this model [63-65]. In contrast, community‑based health management was considered a sustainable strategy that could provide convenient, long‑term services and promote healthy behaviors among older adults [66,67]. Community and group‑based programs provide social interaction, emotional support, and a sense of belonging, which reduce isolation and strengthen psychological stability [16,22]. Community XR interventions also faced some challenges in resource sharing and management. Community-based XR exercise often requires multiple participants to share equipment, space, and network resources. However, most community spaces lacked the specialized infrastructure necessary to support immersive exercise programs, including adequate physical space, stable power supply, network connectivity, and safety facilities [68]. This created unresolved issues such as shared use time arrangements, equipment maintenance, and technical assistance training for community staff [68,69]. For community XR projects to be truly implemented, they also need the support of policies and management mechanisms [70]. For example, by optimizing community resource allocation and conducting technical personnel training, practical problems such as equipment sharing and insufficient infrastructure can be addressed [71].

The dominance of controlled environments in the reviewed studies, therefore, limits the interpretation of continuity and scalability outcomes. While such settings support feasibility testing and supervised delivery, they do not reflect the operational conditions under which older adults engage in exercise in everyday community contexts, making it difficult to assess whether XR-based interventions can be sustained beyond experimental or clinical settings.

The findings of this study aligned with broader existing evidence. With the application of technology-assisted interventions in the field of rehabilitation and exercise for older adults, participatory methods and user-centered design principles were gradually integrated into the entire process of intervention development [72,73]. This trend also appeared in VR rehabilitation and digital health projects, where researchers generally emphasized the widespread application of participatory design and rapid prototyping design [35]. However, our synthesis identified a critical difference: early creative ideation and comprehensive multi‑stage service design remained underused. Most studies adopted fragmented design processes without forming a closed loop between service design and adherence monitoring, which limited the ability to anticipate long‑term engagement barriers and optimize real‑world implementation.

Several studies confirmed that immersive VR interventions provided positive feedback to older adults, including heightened feelings of pleasure, increased willingness to participate in activities, and improved emotional well-being [74,75]. However, some studies reported side effects, such as fatigue, mild physical discomfort, or sensitivity to the headset after the intervention [76,77].

High short-term adherence appeared consistent with prior findings on gamified VR engagement [26,78]. However, this adherence pattern likely resulted from 2 limitations: first, the intervention duration was insufficient, and second, the adherence assessment did not form a standardized indicator, but rather participants achieved a real and sustained change in behavior. In interdisciplinary research practice, standardized compliance assessment played a crucial role. It bridged definitional gaps across disciplines, enabled unified evaluation criteria, and offered standardized measurement tools for longitudinal tracking [60,79]. Furthermore, due to the lack of follow-up data, researchers could not determine whether such adherence persisted beyond the short term [80].

However, a critical underexplored dimension in prior work and our synthesis concerned the setting of implementation. Existing research suggested that implementing XR intervention in community settings could support social interaction, emotional regulation, and behavioral improvement [81-83]. However, most studies focused on hospitals, laboratories, and care facilities. This narrow setting focus weakened the ecological validity of research findings and hindered a comprehensive evaluation of the long-term sustainability of interventions. Integrating XR into community programs could have improved accessibility, peer support, and emotional connection; however, this remained an overlooked area in current research.

This review had some limitations. First, the number of studies included in the analysis was relatively small, and the sample size of most individual studies was also small. More importantly, the studies showed pronounced heterogeneity in design, baseline characteristics of participants, and intervention implementation processes. These differences not only reduced the statistical power of the overall analysis but also limited the generalizability of the research conclusions. In addition, restricting inclusion to studies with accessible full texts may have excluded relevant work based on access rather than content, introducing potential selection bias; gray literature and trial registries were not systematically searched, which may have resulted in the omission of unpublished or ongoing studies. Second, none of the included studies established longitudinal follow-up links, which made it impossible to determine whether the short-term participation adherence translated into long-term behavioral change or functional improvement. Third, regarding intervention settings, most studies focused on hospitals, care facilities, or laboratory environments. In contrast, empirical evidence from community or home settings remained relatively limited, which also weakened the applicability of the findings to real-life scenarios. Finally, although this review adopted an XR-oriented scope, all included studies used VR interventions, reflecting the current concentration of evidence rather than an intentional focus on VR, and exploration of AR or MR modes remained relatively limited. Studies applying AR or MR technologies were identified in the broader literature and have demonstrated positive physical or cognitive outcomes among older adults [84,85]; however, these approaches have not yet been systematically examined within service design-oriented frameworks. The limitations of this technology exploration made it difficult to reveal the comparative advantages of different XR methods in terms of accessibility, usability, and situational adaptability, which affected the comprehensive understanding of the overall application value of XR technology.

To address the shortcomings of current research, future studies should adopt standardized, multidimensional adherence assessment metrics [60]. By establishing unified assessment dimensions and criteria, these metrics could resolve inconsistencies caused by variations in measurement and provide a stable foundation for long-term follow-up studies. Exploration of community-based interventions should be strengthened, with designs that closely mirror real-world exercise environments to address contextual barriers to adaptation effectively. Furthermore, service design tools such as user journey maps should be incorporated into the intervention process, leveraging their visualization capabilities to identify potential dropout risk points at each stage of the intervention, providing a concrete reference for iterative optimization of intervention plans. In parallel, multimodal tracking systems should be developed that integrate physiological data collected by wearable devices, participant behavioral logs, and in-game performance metrics. These systems should generate adaptive feedback to provide participants with more personalized support for interventions. Finally, future research should expand its technological scope by exploring AR and MR modalities and evaluating their potential for more inclusive and ecologically valid exercise interventions.

This review systematically evaluated how XR-based exercise interventions for older adults incorporated service design strategies to support exercise adherence. Although participatory design and gamified VR experiences have been widely used in related research, most studies have remained fragmented and short-term. There were clear deficiencies in both the systematic integration of multistage service design processes and the effective incorporation of standardized compliance frameworks; a comprehensive research practice system had not yet been established. Research findings indicated generally high levels of exercise adherence; however, conclusions varied across studies, and few focused on the intensity of adherence, long-term maintenance, or the ecological validity of the findings. Further analysis from the perspectives of intervention setting and technology application revealed that most existing interventions focused on hospital or laboratory settings, whereas the exploration of community settings, as well as the development of other XR modalities such as AR and MR, was relatively limited. To address this gap, future XR intervention research should integrate structured service design tools throughout the intervention lifecycle, develop standardized adherence metrics, establish multidimensional adherence tracking mechanisms, and extend deployment to real-life settings. By promoting synergy among behavioral engagement, service continuity, and situational adaptability, the effectiveness of XR intervention in helping older adults build sustainable exercise habits, while also providing an empirical reference for the innovative development of inclusive digital health services.