Virtual health coach acceptance in point-of-care settings among community-dwelling older adults: a mixed-method proof-of-concept study

1 Introduction

Many older adults experience age-related changes in physical and cognitive functioning, as well as health conditions that may affect their quality of life and daily activities (World Health Organization, 2024). This phase of development is linked to health issues that affect overall quality of life (Noto, 2023). Older adults face a greater risk of age-related conditions, especially sensory decline (e.g., in vision, hearing) (Correia et al., 2016) and musculoskeletal problems, which can inhibit participation in daily activities (Tomida et al., 2024). Physical health conditions in later life have been associated with psychosocial and mental health outcomes, such as depression and loneliness (Shankar, 2023). Additionally, musculoskeletal disorders are recognized as key factors influencing mental health outcomes among older adults, as they are linked to falls, frailty, sleep disturbances, and reduced mobility, depression, anxiety, and poorer cognitive functioning (Roberts et al., 2016; Welsh et al., 2020).

Preventing falls and maintaining functional independence are hence essential for promoting healthy aging and improving quality of life in older adults. Physical activities, especially those that include balance and strength training, can enhance body coordination (Do and Tolos, 2025). Among these exercises, Tai Chi and Qigong (TCQ) have been shown to enhance physical functioning, mood, weight management, and cardiovascular risk factors. They focus on meditative movement forms that produce positive outcomes for both physical and psychological health (Rogers et al., 2009). Qigong, the ancient foundation of all Tai Chi practices, can be performed standing, walking, sitting, or lying down. Specifically, TCQ connects movements in a continuous sequence, allowing the body to shift from foot to foot while maintaining a lower center of gravity, all while focusing on deep breathing and mental concentration to achieve body-mind harmony (Wong et al., 2001). It is proven that mindfulness exercises significantly improve the synthesis and release of brain-derived neurotrophic factors, which may affect cognitive function and support neurogenesis (Park et al., 2023; Yang et al., 2022).

While older adults continue to experience health challenges, technology-driven healthcare services are currently being integrated into clinical practice, showing considerable benefits in terms of access to usable, efficient, safe, client-specific, and quality care services for older individuals (Dino et al., 2022). In this context, point-of-care technologies refer to technologies designed to deliver healthcare services, support, or interventions at or near the location where care is provided. This broader concept differs from point-of-care testing, which specifically refers to diagnostic testing performed near the patient. In the present study, the focus is on point-of-care technologies, particularly mixed-reality (MR) systems that can support health-related interventions for older adults. This approach may provide more accessible and personalized alternatives to conventional healthcare delivery (Wagner et al., 2006).

Digital health technologies, including MR applications for exercise, have the potential to increase physical activity and reduce sedentary behavior among adults (Greco et al., 2023). As a POCT, MR offers significant potential for effective healthcare programs. It is the blending of the physical and digital worlds, encompassing varying levels of applications from both the real and the artificial (Speicher et al., 2019). MR enables the implementation of virtual coaches, which could promote healthcare as POCT among older adults. Previously, a study (Hammouda et al., 2025) reported the positive benefits of virtual coaches on the physically inactive and found that they have great potential to deliver highly personalized health interventions, leading to measurable changes in health behavior and improvements in body composition. However, a significant gap exists in understanding how emerging technologies such as MR and virtual coaches are being adopted by older adults.

Previous scholars emphasize the importance of examining various aspects of technology adoption through quantitative and qualitative inquiry that assesses both individuals’ lived experiences and behavioral perceptions for technology use (Guetterman et al., 2015). Particularly for older adults who may experience physical, mental, and technological challenges, such an approach is crucial in evaluating their perceived value of technology-driven tools. Therefore, this study aimed to examine older adults’ intention to adopt virtual coaches from the Mixed Reality-Oriented Virtual Coach Experience–Qigong and Tai Chi Module (MOVE-QTM) program, and to model multiple variables related to the Unified Theory of Acceptance and Use of Technology (UTAUT) and perceived point-of-care testing (POCT) outcomes. Specifically, the study sought to (1) describe the demographic profile of older adult participants, (2) develop and evaluate a parsimonious structural equation model, and (3) align quantitative results with qualitative outcomes to understand older adults’ behavioral intention to use MR–based virtual coaching technologies.

2 Related literature

The acceptance of new technologies by end users has become a key focus in information systems. It involves various models and theories that explore how individuals adopt and engage with technology, as well as the factors that influence ongoing use (Momani, 2020). The Unified Theory of Acceptance and Use of Technology (UTAUT) suggests that certain factors directly impact the intention to use or the actual use of technology. These include (1) performance expectancy (PE), (2) social influence (SI), (3) effort expectancy (EE), (4) facilitating conditions (FC), and (5) behavioral intentions (BI), all of which play important roles as direct determinants of user acceptance and behavior (Ayaz and Yanartaş, 2020). As a theoretical model, UTAUT indicates that the perceived likelihood of adopting technology depends on these constructs (Xue et al., 2024). In healthcare settings, UTAUT is widely employed to understand and forecast technology adoption by identifying factors that influence clients’ willingness to use new technologies (Philippi et al., 2021). As healthcare technologies continue to grow, it is equally important to examine broader health-related factors that shape how individuals interact with these systems, particularly the physical and mental factors that may affect acceptance and use.

The perception of positive benefits of using a virtual coach as a point-of-care technology (BE-POCT) is crucial in shaping older adults’ willingness to adopt and use such systems. Virtual coaches represent a transformative healthcare technology that provides cost-effective ways to improve human wellbeing across various areas, including physical health, nutrition, cognition, social engagement, and emotional wellbeing (Kamali et al., 2020). Also, social influence (SI), or the individual’s perception that important people around them expect or encourage them to engage in a specific behavior (Yap et al., 2022), is necessary to propel technology use. Recommendations from family members, peers, or healthcare professionals can boost older adults’ perception of the technology’s value and benefits. Since older adults often have limited familiarity with technology and may struggle to use digital systems effectively (Cabañero-Garcia et al., 2025), the effort needed to operate the virtual coach (Effort Expectancy, EE), as well as access to devices and support (Facilitating Condition, FC), also shape their view of its benefits. When the technology is seen as user-friendly, older adults are more likely to associate it with convenience. Furthermore, performance expectancy (PE) can enhance the perceived benefits of technology use. When older adults believe that a virtual coach can improve their health outcomes or help with daily tasks, they tend to have a positive outlook about the technology. Hence, the following model of variables and proposed hypotheses (Figure 1) was developed.

3 Methods

3.1 Research design

This proof-of-concept study employed an explanatory-sequential mixed-methods design to model factors and examine experiences related to behavioral intention to use a virtual exercise coach delivering TCQ via MR in point-of-care settings among community-dwelling older adults. This research design is appropriate for gathering numerical data and contextualized insights, serving as a two-phase research design in which quantitative data are initially collected and analyzed, followed by qualitative data collection and analysis, which are also used to elucidate the numerical findings (Gustavsen, 2018). It also enables the conceptual and analytical integration of results in order to provide more holistic and comprehensive perspectives (Benson, 2020). Especially in contemporary health services that include emerging or developing technologies, mixed methods research offers powerful and innovative tools for investigating complex processes and systems in healthcare by addressing causality, generalizability, and magnitude of effects, along with exploring how various phenomena occur, developing theories, and describing the nature of individuals’ experiences for quantitative and qualitative areas, respectively (Fetters et al., 2013). This study was retrospectively registered under embargo status at OSF. io.

3.2 Subject and sampling

The sample was drawn from the older adult population of Valenzuela City, Philippines. A total of 38 community-dwelling older adults were recruited from senior centers in the metropolitan area. The participants of the study were selected purposively using the following criteria: (1) aged 60–75 years old, (2) male or female, (3) willing to use the Microsoft HoloLens, and (4) willing to participate, given that they provided informed consent. Those with (1) severe sensory impairments, (2) neurological and psychiatric disorders, (3) mobility impairments, and (4) motion sickness were excluded from the study.

3.3 Virtual coach development

The developed system, the Mixed Reality-Oriented Virtual Coach Experience–Qigong and Tai Chi Module (MOVE-QTM), is an enhanced, research-based version of a former community service program (Dino, Dion, Abadir, Budhathoki, Huang, et al., 2025) that aimed to improve mobility among older adults living in a community. This updated program builds on the foundations of the initial version, introducing a broader, more structured set of multimodal exercise maneuvers specifically designed to reduce fall risk and enhance mobility among older individuals. Figure 2 showcases a sequence of 16 illustrated poses demonstrating the multimodal set of Tai Chi and Qigong exercises, incorporating movements characterized into four primary sets: Opening Qigong, Earthly Palm Qigong, Longevity Qigong, and Tai Chi Forms.

This intervention features an innovative MR-based Virtual Coach, developed from the authentic movements of a real human coach, captured via a motion capture system. The virtual coach is displayed to the user through a transparent head-mounted optical display (OST-HMD; Microsoft® HoloLens). Figure 3 shows various images related to the development of the virtual coach (leftmost) a motion-captured digital avatar performing a Tai Chi exercise routine in a virtual environment (middle) a human model in a motion-capture suit performing the movements tracked by a projection displayed on monitors, and (rightmost) a recording of facial expressions to enable realistic avatar-to-human interaction. Various software and related developer tools used in this study include Unreal Engine® (version 5.0), Motion Capture: Motive®, Live Link: Face® (version 1.4.2), Mixed Reality Toolkit® (MRTK), MetaHuman® (version 5.0), and UX Tools®. Prior to full implementation of the technology, the alpha version of the system was pilot tested with two adult participants to assess initial device usability and functionality.

3.4 Tools and measurement

The following tools were used in the study:

Demographic Profile Sheet. The demographic profile sheet is a researcher-designed questionnaire that asks participants about their name, age, address, civil status, educational attainment, socioeconomic status, technology competency, technology literacy, and general health status.

Universal Theory of Acceptance and Use of Technology (UTAUT) Questionnaire. The UTAUT tool (Venkatesh et al., 2003) consists of determining factors, particularly (1) performance expectancy, (2) social influence, (3) effort expectancy, (4) facilitating conditions, and (5) behavioral intention, serving as direct determinants of user acceptance and usage behavior widely used for measuring technology acceptance (Ayaz and Yanartaş, 2020). The present study slightly modified the UTAUT tool to include concepts about the developed virtual coach technology. The questionnaire comprises 20 questions, answered on a five-point Likert scale from 1 (strongly disagree) to 5 (strongly agree). Questions 1–4, 5–8, 9–10, 11–16, and 17–20 measure PE, EE, SI, FC, and BI, respectively.

Point of Care Testing (POCT) Questionnaire. A descriptive validity assessment was conducted to evaluate the clarity and appropriateness of the POCT questionnaire items for the target participants. This assessment provided an initial evaluation of whether the items were understandable and appropriately represented the intended constructs before the main analysis. The resulting values were interpreted based on the established criteria, with higher values indicating greater agreement regarding the clarity and appropriateness of the items. The POCT questionnaire was designed to gather information on the use, benefits, and challenges of technologies in specific healthcare settings or among specific groups of individuals, supporting the development of prompt and appropriate interventions for clinical decision-making (Oehme et al., 2022). The POCT tool developed by Lilly et al. (2020) was slightly modified to incorporate the technology developed for the present study and align with its objectives. The questionnaire consists of 14 items rated on a five-point Likert scale ranging from 1 (strongly disagree) to 5 (strongly agree). The validity and reliability of the original questionnaire were established in a previous study (Lilly et al., 2020).

Semi-Structured Interviews. A semi-structured framework was used to maintain consistency in the primary questions and flexibility in follow-up probes. This format was specifically suited for (1) collecting open-ended, detailed qualitative data, (2) exploring participants’ thoughts, feelings, and beliefs about the area of study, and (3) delving deeply into personal and, at times, sensitive issues (DeJonckheere and Vaughn, 2019). Semi-structured interviews were conducted with all participants upon completion of the MOVE-QTM sessions to elicit their experiential responses to the virtual coach technology. The interview guide was developed deductively from the four UTAUT constructs, with Behavioral Intention as the outcome domain, and each construct operationalized through a set of probing indicators informed by the study’s theoretical framework. All interviews were conducted in Filipino to ensure linguistic accessibility and naturalistic expression, with probes adjusted iteratively based on participant responses to elicit elaboration on both positive and negative experiences and areas of difficulty.

3.5 Data collection

The data collection process consisted of four phases: (1) orientation, (2) intervention, (3) survey and interview, and (4) debriefing.

Orientation Phase. Participants were oriented to the study procedures, MOVE-QTM, and the use of the Microsoft HoloLens. The researchers explained the intervention and provided instructions on how participants would interact with the virtual coach. Informed consent was obtained before proceeding to the intervention.

Intervention Phase. Participants individually underwent the MR-driven Tai Chi and Qigong exercises using the Microsoft HoloLens. The intervention was delivered on an individual basis, with participants completing the exercises separately rather than simultaneously with other participants. The intervention lasted approximately 30 min, including warm-up and cool-down exercises. Participants performed the exercises in a mixed-reality environment, where the virtual coach was overlaid onto their real-world surroundings, allowing them to simultaneously view their physical environment and the virtual coach. They followed the virtual coach’s movements and instructions throughout the exercise session. A technical team assisted with the operation of the device and program, while a health team was available to monitor participants during the session.

Survey and Interview Phase. Following the intervention, participants completed the UTAUT and POCT questionnaires individually. Semi-structured interviews were subsequently conducted in Filipino to gather their experiences and suggestions regarding MOVE-QTM.

Debriefing Phase. Participants were debriefed regarding their experience with the device and intervention, followed by a post-intervention health check.

The study was conducted in the Philippines in collaboration with the U.S.-based research team. Prior to data collection, the research team underwent training on the study protocol, administration of the MOVE-QTM intervention, participant orientation, and data collection procedures. The study instruments and interview materials were also reviewed for linguistic and contextual appropriateness for the Filipino participants. Where translation was required, the materials were translated into Filipino and reviewed to ensure that the meaning and intent of the original English versions were retained.

A health check was conducted by the health team to assess participants’ condition and ensure their safety before and after the intervention. The health team monitored participants for any discomfort or adverse symptoms during the exercise session. The data collection was conducted by six members of the local research team, comprising nurses, biologists, IT experts, and policy reviewers. Team members were responsible for participant orientation, intervention administration, survey distribution, qualitative interviews, and monitoring during the intervention, according to their assigned roles.

3.6 Data analysis

To model the perceived benefits and intention for use among community-dwelling older adults, a Structural Equation Modeling (SEM) through Partial Least Squares (PLS) was performed. As a statistical method, PLS-SEM aids in the precise examination and estimation of relationships among a set of variables (Diño and De Guzman, 2015), as well as the complex evaluation of their theoretically determined relationships (Hair and Alamer, 2022). The interpretation and analysis followed the established guidelines for SEM interpretation and reporting (Hulland, 1999) and used the PLS Graph software (Hair et al., 2021). Construct reliability of the measurement instrument was examined using composite reliability and Cronbach’s alpha (Kock, 2015), which determines the internal consistency of a set of measurement items (Hair et al., 2021; Straub and Gefen, 2004). On the other hand, convergent validity, indicating the quality of the tool, was evaluated by (a) confirming that the item loadings on each construct are statistically significant (p < 0.05) and meet or exceed a value of 0.50, and (b) calculating the Average Variance Extracted (AVE), reflecting the proportion of variance by a construct in relation to measurement error with a threshold of 0.50 or higher (Hair et al., 2021; Kock, 2015). Moreover, AVE was also utilized to assess discriminant validity, where a construct’s AVE should exceed its squared correlations with other constructs (Fornell and Larcker, 1981).

To integrate the quantitative and qualitative data strands, a joint display was constructed using the guidelines of a previously published framework (Guetterman et al., 2015). Rather than leading with statistical parameters, this format foregrounds participant voice, with the qualitative excerpt selection itself signaling sentiment valence. Participant excerpts were originally spoken in Filipino and translated into English for presentation.

3.7 Ethics

This research complies with the standards outlined in the Declaration of Helsinki and the Philippine Health Research Ethics Board (PHREB) guidelines, as specified in the 2022 National Ethical Guidelines for Research Involving Human Participants (NEGRIHP). To ensure voluntariness, participants provided informed consent and were informed that they could withdraw at any time during the study without consequences. Since the study involved mixed-reality simulations, participants’ health was carefully monitored before, during, and after the intervention. A board-certified gerontologist and a team of nurses were on standby to monitor participants’ health. Transportation, meals, and refreshments were supplied to all participants. The privacy and confidentiality of participants were strictly maintained throughout the study. Each participant was assigned a pseudonym (e.g., MR-TCQ-01) to use during their involvement. Data was stored securely in an encrypted institutional cloud storage service, and will be retained for 5 years after collection. Afterward, files will be deleted and physical copies shredded. The (anonymized) Institutional Ethics Review Committee (Reference Number: 2025-IERC-00XXX) provided clearance to the study.

4 Results

4.1 Participants’ profile data

Table 1 presents the demographic characteristics of the participants. The study included 38 respondents (N = 38), with a dominant representation of females (97.37%) and only one male participant (2.63%).

TABLE 1. Convergent validity assessment: indicator loadings and average variance extracted (AVE).

In terms of age, most participants were in the 60–64 age group (50.00%), followed by 65–69 (34.21%), 70–74 (13.16%), and 74–79 (2.63%), indicating that the sample is largely composed of younger older adults. For civil status, married (39.47%) and widowed (39.47%) participants were equally represented, followed by separated (10.53%), single (7.89%), and not indicated (2.63%). In terms of educational attainment, most participants had completed secondary education (52.63%), followed by tertiary education (36.84%), and primary education (10.53%). Regarding financial status, the majority were classified as low income (55.26%), followed by not indicated (42.11%), and a small proportion from middle income (2.63%), with no participants reported under high income.

4.2 Measurement model assessment

Table 1 presents the convergent validity assessment conducted to determine whether the observed variables adequately represent their corresponding latent constructs. Convergent validity was evaluated using indicator loadings and average variance extracted (AVE), following a criteria (Amora, 2021). The criteria include: (a) indicator loadings should be 0.50 or higher, (b) associated p-values should be below 0.05, and (c) cross-loadings should be significantly lower than the corresponding indicator loadings.

The indicator loadings for Behavioral Intention (BI) ranged from 0.75 to 0.87, Facilitating Condition (FC) from 0.77 to 0.78, Social Influence (SI) from 0.74 to 0.85, Benefits as POCT (BEPOCT) from 0.69 to 0.86, Effort Expectancy (EE) from 0.66 to 0.87, and Performance Expectancy (PE) from 0.82 to 0.91. These values suggest that all indicators exceeded the recommended threshold of 0.50. Also, all associated p-values were below 0.05, suggesting that all indicator loadings were statistically significant. Four indicators under Benefits as POCT (BEPOCT), namely, BEPOCT1, BEPOCT3, BEPOCT8, and BEPOCT12, were removed from the analysis because their cross-loadings exceeded their corresponding indicator loadings, thereby violating the criterion. To further support convergent validity, the AVEs for all constructs exceeded the recommended threshold of 0.50 (Fornell and Larcker, 1981; Kock, 2015; Kock et al., 2012). Specifically, the AVE values were 0.76 for BI, 0.60 for FC, 0.67 for SI, 0.62 for BEPOCT, 0.66 for EE, and 0.79 for PE. These findings confirm that the constructs demonstrated adequate convergent validity.

Table 2 presents the discriminant validity assessment, which determines whether each latent construct represents a unique concept without excessive overlap with other constructs. We assessed discriminant validity using two approaches: the Fornell-Larcker criterion (Fornell and Larcker, 1981) and the HTMT2 (Roemer et al., 2021). Under the Fornell-Larcker criterion, the bolded diagonal values in Table 2 (0.87, 0.78, 0.82, 0.79, 0.81, and 0.89) represent the square roots of the average variance extracted (AVE) for BI, FC, SI, BEPOCT, EE, and PE, respectively. The lower triangle presents the inter-construct correlations. Discriminant validity is supported when each diagonal value exceeds the corresponding correlations in its row and column. This condition was satisfied for all constructs in Table 2.

TABLE 2. Discriminant validity statistics: Correlations, Square Roots of Average Variance Extracted (AVEs), and Heterotrait–Monotrait ratio (HTMT2).

The HTMT2 criterion, presented in the upper triangle, provides an additional assessment of discriminant validity by comparing heterotrait correlations with monotrait correlations for each pair of distinct constructs. HTMT2 is not calculated for a construct compared with itself; therefore, the diagonal values are reserved for the square roots of AVE under the Fornell-Larcker criterion. All HTMT2 values ranged from 0.42 to 0.84, remaining below the recommended threshold of 0.85. These findings indicate that the latent constructs demonstrate adequate empirical distinctiveness. Taken together, the Fornell-Larcker and HTMT2 results support the establishment of discriminant validity across the latent constructs.

Table 3 presents the reliability coefficients of the latent variables, assessed using composite reliability (CR) and Cronbach’s alpha (α). The results show that Behavioral Intention (BI) obtained CR = 0.93 and α = 0.89; Facilitating Condition (FC) obtained CR = 0.86 and α = 0.78; Social Influence (SI) obtained CR = 0.89 and α = 0.84; Benefits as POCT (BEPOCT) obtained CR = 0.94 and α = 0.93; Effort Expectancy (EE) obtained CR = 0.88 and α = 0.82; and Performance Expectancy (PE) obtained CR = 0.94 and α = 0.91. All constructs demonstrated CRs and Cronbach’s alphas of 0.70 or higher, indicating acceptable internal consistency and reliability of the measurement model (Cronbach, 1951; Fornell and Larcker, 1981).

TABLE 3. Reliability coefficients of the latent variables.

4.3 Structural model evaluation

Table 4 presents the model fit and quality indices used to evaluate the adequacy and explanatory power of the structural model. The results meet a previously published criterion (Kock, 2015). The results demonstrated that the model achieved acceptable to ideal levels across all evaluation criteria. The Average Path Coefficient (APC) yielded β = 0.293 and p = 0.012, indicating a statistically significant relationship among the model paths, as the value is below the recommended threshold of p < 0.05. Similarly, the Average R-squared (ARS) yielded β = 0.553 (p = 0.000), while the Average Adjusted R-squared (AARS) yielded β = 0.513 (p = 0.000). Both indices were statistically significant and considered acceptable, suggesting that the model has strong explanatory power.

TABLE 4. Model fit and quality indices.

In terms of collinearity assessment, the Average Block VIF (AVIF) produced a value of β = 2.092, whereas the Average Full Collinearity VIF (AFVIF) yielded β = 2.674. Both values were below the ideal cutoff of 3.3, indicating minimal multicollinearity issues and demonstrating that the constructs were sufficiently distinct from one another. The Tenenhaus Goodness of Fit (GoF) value was β = 0.615, which exceeds the criterion for a large effect size (≥0.36). This finding indicates a strong overall model fit and substantial explanatory power of the structural model. Furthermore, the Standardized Root Mean Squared Residual (SRMR) and Standardized Mean Absolute Residual (SMAR) obtained values of β = 0.098 and β = 0.094, respectively. Both values were below the acceptable threshold of 0.10, confirming that the model residuals were within acceptable limits and further supporting the model’s adequacy. Overall, the findings indicate that the structural model demonstrated acceptable to ideal fit indices, significant explanatory strength, low collinearity, and strong overall predictive capability.

Table 5 and Figure 3 present the direct and indirect effects of the variables included in the structural model. Among the direct effects, Facilitating Conditions (FC) significantly influenced Behavioral Intention (BI) with β = 0.278, SE = 0.144, p = 0.030, and f² = 0.169, indicating a medium effect size. Similarly, Social Influence (SI) demonstrated a significant positive effect on BI (β = 0.287, SE = 0.143, p = 0.026, f² = 0.192), also reflecting a medium effect size. In addition, BEPOCT significantly affected BI with β = 0.249, SE = 0.145, p = 0.048, and f² = 0.154, indicating a medium effect.

TABLE 5. Variable effects.

Effect size (f²): Based on Cohen (1988), 0.02 = small, 0.15 = medium, and 0.35 = large.; R-squared: R ² for BI, 0.515; R ² for BE-POCT, 0.590; Q-squared: Q² for BI, 0.530; Q² for BE-POCT, 0.719.

Furthermore, SI showed a strong and significant effect on BEPOCT (β = 0.656, SE = 0.121, p = 0.000, f² = 0.533), indicating a large effect size (Cohen, 1988). This finding suggests that SI was the strongest predictor within the model. On the other hand, Effort Expectancy (EE) did not significantly influence BEPOCT (β = 0.190, SE = 0.149, p = 0.105, f² = 0.113). Likewise, Performance Expectancy (PE) showed a nonsignificant negative effect on BEPOCT (β = −0.099, SE = 0.155, p = 0.263, f² = 0.056).

For the indirect effects, none of the mediated relationships through BEPOCT were statistically significant. Specifically, the indirect effect of SI on BI through BEPOCT was nonsignificant (β = 0.163, SE = 0.107, p = 0.067, f² = 0.109). Similarly, EE → BEPOCT → BI produced β = 0.338, SE = 0.112, p = 0.338, and f² = 0.033, while PE → BEPOCT → BI yielded β = −0.025, SE = 0.113, p = 0.414, and f² = 0.014, both indicating nonsignificant mediation effects. In terms of explanatory power, the model accounted for 51.5% of the variance in BI (R ² = 0.515) and 59.0% of the variance in BEPOCT (R ² = 0.590), indicating moderate to substantial explanatory capability. Moreover, the Q² values for BI (Q² = 0.530) and BEPOCT (Q² = 0.719) suggest that the structural model has strong predictive relevance (Figure 4).

4.4 Mixed methods integration

Table 6 presents the convergent comparison of PLS-SEM path coefficients alongside corresponding qualitative sentiment distributions and representative participant excerpts, organized by UTAUT construct. Each path is classified as Converging (both strands mutually reinforce), Diverging (strands point in discordant directions), or Partial (mixed evidence across subthemes). This integrative display illuminates not only where the two data sources confirm one another, but also where their divergence yields richer theoretical insight than either strand alone could provide.

TABLE 6. Joint Display on Convergence and Divergence of Quantitative (PLS-SEM) and Qualitative (UTAUT Interview Coding) Findings (N = 38; 404 coded excerpts).

▪ Qualitative sentiment percentages reflect proportions of coded excerpts (not participants) classified by trained encoders (k = 8; 404 total coded excerpts).

▪ Quantitative effects are standardized path coefficients (beta) from PLS-SEM, analysis. Effect size (f2): 0.02 = small, 0.15 = medium, 0.35 = large (Cohen, 1988).

▪ Convergence classification: Converging = quantitative significance and ≥ 60% qualitative positive sentiment align directionally; Diverging = statistically non-significant path paired with ≥ 60% qualitative positive sentiment; Partial = mixed evidence or subtheme-level heterogeneity.

▪ Participant excerpts were originally spoken in Filipino; English translations are provided for accessibility. Original-language excerpts appear in the supplementary dataset.

▪ BE-POCT, was operationalized via the POCT, questionnaire (Lilly et al., 2020) and was not a separate category in the qualitative coding scheme, which was organized around UTAUT, constructs. Direct qualitative comparison for the BE-POCT, path is therefore not available; EE, effort expectancy; PE, performance expectancy; SI, social influence; FC, facilitating condition; BE-POCT, Benefits as Point-of-Care Technology; BI, Behavioral Intention. *p < .05. ***p < .001.

The joint display analysis revealed three distinct patterns of convergence across the quantitative and qualitative strands. Social Influence (SI) showed the clearest convergence: the largest path in the model (SI → BE-POCT: β = 0.656, p < 0.001, f² = 0.533) and a significant direct effect on Behavioral Intention (β = 0.287, p = 0.026) were directly mirrored by uniformly positive qualitative sentiment, with peer encouragement and social persuasion subthemes coded 100% Positive (n = 12). In contrast, Effort Expectancy (EE) and Performance Expectancy (PE) produced the study’s most pronounced divergence: neither significantly predicted BE-POCT (EE: β = 0.190, p = 0.105; PE: β = −0.099, p = 0.263), yet both generated the highest positive sentiment proportions in the entire coding dataset (EE: 64%, n = 61/95; PE: 70%, n = 73/104). Facilitating Conditions (FC) demonstrated partial convergence, which is a significant path to BI (β = 0.278, p = 0.030) alongside the most heterogeneous qualitative profile (41% Positive, 27% Negative, 32% Neutral; n = 138). Representative participant excerpts grounded each pattern in lived experience. For SI, participants described social co-presence as both motivating and value-shaping [e.g., “It is better with more people. It encourages us more when there are many of us” (K1)]. Although the intervention was delivered individually, this comment reflected the participant’s perception that exercising with others could provide additional motivation and encouragement, rather than describing the actual mode of intervention delivery. For EE and PE, participants reported positive experiences [e.g., “Before I really struggled. Now I can already follow along” (K18); “I feel stretched out and energized” (K12)] yet these affective appraisals did not translate into the structured POCT-related health beliefs captured quantitatively, indicating an operational decoupling between felt experience and formal belief formation. For FC, device compatibility was appreciated [e.g., “It is fine because I can wear my eyeglasses with it” (K14)] while audio difficulty and the absence of real-time corrective feedback emerged as salient barriers [e.g., “Nobody tells you to fix it. That is what I was looking for” (K20)].

5 Discussion

5.1 Participants’ demographics

The demographic profile indicates that the sample is predominantly composed of female participants, mainly younger older adults, most of whom are married or widowed. This finding aligns with previous studies suggesting that female older adults are generally more likely to engage in physical activity than their male counterparts. The predominance of participants aged 60–69 years may additionally reflect the relevance of health-promoting interventions during the earlier stages of older adulthood, when individuals may still be actively engaged in maintaining their physical health, independence, and quality of life. Older women also participate more in interpersonal and community-based social activities, which may further encourage active lifestyles (Ong et al., 2024). Since social support is considered a key determinant of physical activity (Marthammuthu et al., 2023), women may be more inclined to participate in exercise when supportive and guided environments are present. In particular, women are more likely to express motivation for exercise when guided by an instructor, less likely to prefer control over exercise type and difficulty, more interested in smartphone applications, and more willing to participate in app-guided exercise programs (Lee et al., 2022). In addition, the respondents generally have secondary-level education and are largely from low-income backgrounds, suggesting a relatively homogeneous group in terms of socioeconomic status. These socioeconomic conditions may also influence patterns of physical activity, as older adults in low-income settings often engage in leisure activities shaped by social and community contexts. The findings support promoting gender-sensitive health programs that consider the differing motivations, preferences, and social participation patterns of older adults, particularly among women.

5.2 Measurement model assessment

The results confirmed that all indicators adequately represented their respective latent constructs, demonstrating strong associations between the measurement items and their intended concepts. Among the constructs, Performance Expectancy (PE) and Behavioral Intention (BI) showed particularly strong indicator representation compared to the other variables. This pattern may reflect the practical, health-oriented nature of the virtual health coach intervention. Older adults may have perceived mixed reality-assisted Tai Chi and Qigong (TCQ) exercises as directly relevant to their mobility, wellness, and self-management goals (Chen et al., 2020; Song et al., 2026), enabling them to form clearer, more consistent evaluations of the technology’s usefulness and their willingness to adopt it (Dino et al., 2024; Schaumburg et al., 2025). The conceptual proximity of these constructs to personal health outcomes may have reduced personal ambiguity, thereby strengthening convergence validity results. These findings highlight the importance of framing digital health technologies around practical health outcomes and real-world relevance for older adults (Mace et al., 2022). The integration of the familiar, meaningful wellness activities appears to enhance perceptual clarity and support acceptance of immersive health interventions in this population (Sohrabei et al., 2025).

The findings also show that all latent constructs were empirically distinct, indicating minimal overlap among the model’s variables. The results suggest that each construct captured a unique dimension of technology acceptance and usage behavior. The commendable discriminant validity provides a robust foundation for structural equation modeling (SEM) to test the causal relationships between variables. Because these constructs are empirically distinct, researchers can confidently include them without the risk of multicollinearity distorting the path coefficients.

5.3 Structural model evaluation

The structural model showed an acceptable-to-ideal fit, indicating strong explanatory power, low multicollinearity, and adequate predictive performance, with meaningful and statistically significant relationships among all latent constructs. These findings imply that the structural model provides a reliable and theoretically meaningful framework for understanding the determinants of behavioral intention and BE-POCT adoption. The model’s strong explanatory and predictive capability further suggest its potential utility in guiding future research and informing evidence-based implementation strategies for healthcare technologies.

In terms of direct variable effects, FC, SI, and BE-POCT significantly influenced BI, with SI emerging as the strongest predictor. This may be attributed to participants’ strong influence from peers, family members, and colleagues’ opinions and expectations regarding system use. Research on healthcare technology adoption found that social influence significantly affects individuals’ behavioral intention to use healthcare systems and technologies, which may be explained by collectivist cultural values, where people are more influenced by the opinions, expectations, and interpersonal norms of peers, family members, and organizations when adopting health technologies (Metallo et al., 2022; Reyes et al., 2025). Studies reported that insufficient training and a lack of technical support act as barriers to technology adoption, while adequate support and infrastructure improve users’ willingness to adopt (Alotaibi et al., 2025).

Finally, EE and PE were found not to have a significant impact on BE-POCT, while SI emerged as the factor with the strongest impact. The non-significant findings indicate that respondents may recognize the benefits of point-of-care technology regardless of its ease of use or its expected effect on performance. Prior studies in healthcare technology adoption found that both performance expectancy and effort expectancy did not significantly determine behavioral intention to use mobile health services, despite being core predictors in the UTAUT framework, suggesting that in certain healthcare technology contexts, these constructs may not be notable drivers of user attitudes (Mensah et al., 2022). These findings support the notion that in contexts where the perceived benefits of a technology are well-established or self-evident, EE and PE may carry less weight in shaping user attitudes toward that technology (Ramírez-Correa et al., 2023).

Contrary to expectations, the indirect pathways through BE-POCT were not statistically significant, indicating that perceived benefits as point-of-care technology did not mediate the relationships between the UTAUT constructs and BI as initially expected. The findings suggest that BI may instead be formed through more direct mechanisms, particularly through social encouragement and facilitating conditions, rather than through cognitive evaluation of technological benefits. From a theoretical perspective, this divergence from the hypothesized model may reflect contexts and concepts within the Socioemotional Selectivity Theory (Carstensen, 2021), suggesting that, as individuals age, social goals shift from long-term, knowledge-based goals to emotionally meaningful interactions, trust, and supportive social relationships. Also in this study, SI and FC demonstrated more immediate effects on BI compared to mediated cognitive evaluations through perceived benefits, thus suggesting that older adults’ acceptance of the MOVE-QTM system may have been shaped more strongly by reassurance, guidance, and interpersonal, social support (Kim et al., 2024) rather than by deliberate assessment of technological benefits alone. This interpretation is also supported by a related qualitative study (Dino, Dion, Abadir, Budhathoki, Balbin, et al., 2025) involving an MR-based exercise program, in which older adult participants described the intervention as a channel for socializing, building friendships, and developing a sense of belonging with fellow participants.

5.4 Mixed methods integration

The converging SI findings carry the clearest theoretical implication of the integration. The magnitude of the SI → BE-POCT path is further strengthened by the qualitative finding that every excerpt of peer encouragement and social persuasion was uniformly positive. Participants did not simply find co-exercising more enjoyable but also articulated that social visibility changed how much they valued the activity itself. This pattern is consistent with the Socioemotional Selectivity Theory, which holds that emotionally meaningful social relationships increasingly guide decision-making in later life, and with collectivist cultural orientations documented in the Philippine older adult context (Hofer, 2020). The EE and PE divergence, by contrast, points to a more fundamental issue in applying UTAUT to affectively rich technologies with older adults. The standard model treats ease of use and perceived performance as deliberate cognitive antecedents to acceptance, and the qualitative data suggest these constructs operated as immediate affective responses in this sample, who felt ease and felt benefit that did not consolidate into formal health management beliefs. This decoupling is consistent with prior observations in healthcare technology contexts where benefits are emotionally accessible or experientially immediate and suggests that longitudinal designs are needed to examine whether repeated use eventually bridges the gap between experiential appraisal and structured POCT belief formation (Korte et al., 2020; Teebagy et al., 2022).

The partial convergence on FC points to a gap between overall adoption enabling and moment-to-moment experience quality. Despite the significant FC → BI path, specific barriers (e.g., unclear audio instructions, device slippage, and the absence of real-time corrective feedback) generated friction that the aggregate Likert score did not capture. Addressing these issues in future iterations of MOVE-QTM is likely to strengthen the FC effect and improve overall adoption. More broadly, results demonstrate the added value of integrating explanatory-sequential mixed-methods, in which the quantitative model identified which constructs predicted adoption, while the qualitative strand explained why and surfaced barriers invisible to the structural model alone (Toraman Turk et al., 2025).

Although the present study assessed technology acceptance following a one-time supervised intervention, future implementation may consider how MOVE-QTM could be accessed and used beyond the research setting, particularly among lower-income older adult population. Potential delivery through senior centers may provide opportunities for shared access, social interaction, and technical assistance. The appropriate number of participants who can engage with the intervention simultaneously and whether older adults can use the technology may also independently warrant further investigation. These implementation considerations may aid in highlighting the roles of Social Influence, Effort Expectancy, and Facilitating Conditions in shaping Behavioral Intention among future studies.

6 Conclusion

This study examined the acceptance of the MOVE-QTM mixed-reality Tai Chi and Qigong virtual coach among community-dwelling Filipino older adults, using an explanatory-sequential mixed-methods design. Quantitative PLS-SEM analysis revealed that Social Influence, Facilitating Conditions, and Benefits as Point-of-Care Technology significantly predicted Behavioral Intention, with Social Influence emerging as the dominant driver, exerting the strongest effect on BE-POCT. Effort Expectancy and Performance Expectancy, despite generating the most positive qualitative sentiment in the dataset, did not significantly predict BE-POCT, a divergence that the mixed-methods integration illuminated as a decoupling between affective experiential appraisal and formal health benefit cognition. Qualitative interview coding further revealed that peer encouragement, social co-presence, and practical enabling conditions shaped acceptance in ways that the structural model alone could not fully explain. Together, these findings suggest that for older Filipino adults, acceptance of mixed reality health technologies is shaped more by social embeddedness and interpersonal support than by individual evaluations of usability or technological performance.

These findings carry both theoretical and practical implications for the design and implementation of digital health interventions for older adults. Theoretically, the study extends the UTAUT framework by demonstrating that, in affectively rich and socially embedded healthcare technology contexts, Effort Expectancy and Performance Expectancy may operate as implicit experiential responses rather than deliberate cognitive antecedents, which is a qualification particularly relevant for aging populations whose technology engagement is increasingly mediated by social and emotional priorities. Practically, the findings underscore the importance of group-based delivery models, community-oriented implementation strategies, and targeted hardware and software refinements, particularly in audio clarity, device ergonomics, and real-time corrective feedback, to strengthen the enabling conditions that support sustained adoption. Future research should examine whether the felt benefits of the MOVE-QTM consolidate into formal POCT-related health beliefs over repeated use and test the generalizability of these findings across diverse older adult populations and healthcare settings.

7 Limitations

The predominance of female participants limited the examination of potential sex-related differences in technology acceptance and intervention experiences. Future studies may further explore male involvement in exercise-based technology interventions and factors contributing to the predominantly female participation in such programs. The exclusion of individuals with disabilities also limits the generalizability of the findings to older adults with diverse functional abilities. Future research should consider including older adults with disabilities to examine the accessibility and applicability of MOVE-QTM across different functional abilities. Additionally, the study was conducted among a culturally homogeneous Filipino population, which may limit the transferability of the findings to older adults from other cultural and geographic contexts. Future studies should examine MOVE-QTM among more culturally diverse populations to determine whether technology acceptance and intervention experiences vary across contexts. The uneven distribution of participants across age groups also limited meaningful comparisons of Behavioral Intention between younger and older adults. Future studies should therefore recruit larger and more proportionally distributed age groups to examine potential age-related differences in Behavioral Intention and technology acceptance. Finally, the intervention was delivered in a single session, limiting the assessment of sustained engagement, continued use, and changes in Behavioral Intention following repeated exposure. Longer-term studies are needed to better understand the continued acceptance and engagement with MOVE-QTM.

Generative AI statement

The author(s) declared that generative AI was not used in the creation of this manuscript.

Any alternative text (alt text) provided alongside figures in this article has been generated by Frontiers with the support of artificial intelligence and reasonable efforts have been made to ensure accuracy, including review by the authors wherever possible. If you identify any issues, please contact us.

Publisher’s note

All claims expressed in this article are solely those of the authors and do not necessarily represent those of their affiliated organizations, or those of the publisher, the editors and the reviewers. Any product that may be evaluated in this article, or claim that may be made by its manufacturer, is not guaranteed or endorsed by the publisher.

Footnotes

Edited by: Sait Alp, Trabzon University, Türkiye

Reviewed by: Elena Remillard, Georgia Institute of Technology, United States

Elizabeth Wianto, Maranatha Christian University, Indonesia