This was a mixed methods process evaluation involving participants who were randomized into the intervention group of the RETREAT RCT (ACTRN12623000780651). The trial compared the “My Joint Tai Chi” intervention comprising an unsupervised online Tai Chi program plus educational information and an exercise adherence support app (“My Exercise Messages” app) with online education control for people with knee osteoarthritis [9]. It enrolled 178 Australian participants, with 89 randomized to the Tai Chi group. The enrollment period was from August 2023 to December 2024. The process evaluation included both quantitative and qualitative measures obtained via self-reported questionnaires.

For the overarching RCT, people were recruited from the Australia-wide community via online advertisements and social media. The main inclusion criteria were: knee osteoarthritis diagnosis using the National Institute for Health and Care Excellence clinical osteoarthritis criteria [10]; history of knee pain ≥ 3 months; knee pain on most days in the past month; knee pain in the past week during walking ≥ 4 on an 11-point numerical rating scale (NRS), and; home internet connection and a device with access to the internet. The RCT protocol [9] outlines the detailed eligibility criteria. For this study, only participants randomized to the Tai Chi intervention were eligible to be included.

A detailed description of the intervention (“My Joint Tai Chi”) and its development has been published [7,9]. The intervention was delivered via a website [11] with the intervention design features summarized in Multimedia Appendix 1. Briefly, “My Joint Tai Chi” was co-developed by Tai Chi instructors, people with osteoarthritis, physiotherapists, and osteoarthritis researchers. The main component of the intervention was a Tai Chi program delivered as a series of 12 different prerecorded 45-mintue videos (one video per week, with each one to be performed 3 times per week for 12 weeks) of a Tai Chi lesson led by an experienced Tai Chi instructor (JH). Each video consisted of a warm-up, a cool-down, and a modified 10-form Yang-style Tai Chi sequence, specifically adapted for individuals with hip and/or knee osteoarthritis and no/minimal prior Tai Chi experience. The program began with basic movements, gradually introducing new movements each week to progressively increase the level of difficulty. The website also included educational information about Tai Chi, living with knee osteoarthritis, treatment options, and the benefits of exercise for osteoarthritis, as well as a recommendation to download and use the “My Exercise Messages” mobile app [12] to facilitate adherence to the Tai Chi program.

Participant characteristics were extracted from baseline data collected for the RCT. Data regarding the process measures were collected via self-reported questionnaires administered 12 weeks after randomization in the trial, unless otherwise stated below. All data were collected via an online data collection platform REDCap [13,14].

The process evaluation measures collected were the following:

User engagement measures:

User experience measures:

Quantitative measures were summarized as mean (SD) or median (IQR) for continuous variables, and counts or percentages for categorical variables.

For the open-text responses about the positive and negative aspects of the program, a qualitative inductive content analysis approach was used [18]. All open-text data were exported to Microsoft Excel and subsequently analyzed in NVIVO (QSR International Pty Ltd 2015, released 2024). Two researchers (SZ and JP) independently conducted the initial analysis, beginning by reading through all responses to familiarize themselves with the dataset. Date was deemed invalid and removed if the answer was not related to either positive or negative aspects of the program. The remaining data were then coded separately for each question by identifying “units of meaning” within participant responses that reflected either positive or negative aspects of the program. Researchers avoid using preconceived categories, instead immersing themselves in the data to allow new insights and thematic categories to emerge [19]. Each researcher independently generated preliminary themes and subthemes by organizing related codes and distinguishing them from unrelated ones. These themes and subthemes were labelled and defined, with a focus on manifest content [18,20]. Throughout the process, the researchers iteratively moved between the research question, original data, coding, and development of themes to ensure coherence and accuracy. The codes, subthemes, and themes were then discussed with the broader research team, initially with RN and subsequently with KB and RH, to consider alternative categorizations of the data (such as merging or reorganizing themes) and their descriptions.

Informed consent was obtained from participants before their completion of baseline questionnaires via online forms using REDCap (Research Electronic Data Capture) [13,14]. Ethics approval was obtained (Reference number 2023-26838-47588-6) in May 2023 from the Human Research Ethics Committee of the University of Melbourne. All data collected in this study were deidentified and stored on secure university servers, accessible only to the researchers using a password.

Of the 89 participants allocated to the Tai Chi group in the RCT, 64/89 (72%) completed all process measures at 12 weeks. Table 1 presents the baseline characteristics, and Multimedia Appendix 2 presents the primary outcomes of participants who did and did not complete the process measures. No major differences were identified between these subgroups. The number of participants who self-reported their Tai Chi sessions via fortnightly questionnaires was at its lowest (56/89, 63%) during weeks eight and ten, and at its highest (67/89, 75%) during week two. A total of 74 (83%) participants completed at least one questionnaire across the 12 weeks, while 15 (17%) did not complete any questionnaire at any time point.

Adherence to the Tai Chi program is shown in Table 2. Among those who provided adherence data (n=74), the average number of days Tai Chi was performed per week over weeks 1-12 was 2.3 (SD 1.1), with 54 (73%) participants achieving “acceptable” adherence (doing Tai Chi ≥ 2 days per week). The proportion of participants classified as adherent in each fortnightly period was stable throughout the 12 weeks, ranging from 71% in weeks 10 to 80% in weeks 8 (Figure 1). The adherence rate was lower when including those who did not provide adherence data and were categorized as nonadherence (Multimedia Appendix 3). The EARS also demonstrated high adherence to the program, with a median score of 20 (IQR 13–23.5) out of 24 reported at 12 weeks.

Five responses were classified as invalid. Sixty-two participants provided at least one response to these questions, while 27 participants did not provide any response. Participants reported high satisfaction and a strong likelihood of recommending the program to others. Also, they perceived the program to be highly credible, acceptable, and usable (Table 3).

Positive and negative aspects of the Tai Chi program are presented in Table 4, alongside themes, subthemes, and exemplar quotes.

Regarding positive aspects of the program, three themes were identified. First, participants described it as an engaging and well-delivered program. They appreciated the clear instructions and demonstrations from the instructor, and felt the program was well-structured and easy to understand. They also valued the comprehensive coverage of movements and information, delivered in a flexible format that allowed them to practice, rewatch, and rewind sessions without time or location constraints. Second, participants reported improvements in their physical and mental outcomes. For example, some noticed improvements in knee pain, function, body awareness, mental focus, and alertness. Third, participating in the Tai Chi program was considered a positive experience. Most participants found it relaxing and calming, and enjoyed the opportunity to learn something new.

In terms of the negative aspects, 4 themes were identified. First, participants raised concerns about program delivery. Some found the sessions too long and unrealistic in terms of the time required, while others felt the sessions were slow, boring, and repetitive. In contrast, a few people felt the sessions went too fast and wished they could be longer. Some participants also felt there was too much talking from the instructor. The prerecorded video format meant participants were unable to communicate with an instructor and gain feedback or check whether their movements were correct, which left some participants feeling frustrated. Second, participants reported various learning challenges. These included physical difficulty in performing complex movements, particularly due to a lack of coordination, and cognitive difficulty in remembering the movements. Third, technological challenges were identified. Issues such as poor internet connection and video presentation (eg, overly close-up shots) made learning more difficult for some. Fourth, some participants expressed dissatisfaction with their personal outcomes, feeling the program neither addressed their problems nor met their expectations.

In addition, suggestions for improvements were provided. Some participants wanted more information about Tai Chi before starting the program. There was also interest in having interactive components, such as an online forum for questions or follow-up opportunities with the instructor.

This process evaluation of a 12-week online unsupervised Tai Chi intervention showed that among participants who completed the measures, most were adherent to the Tai Chi program, reported high satisfaction with the program, a strong likelihood of recommending it to others, and perceived it as credible, usable, and acceptable. Many participants described the program as engaging and well delivered, reporting a positive experience and gaining improvements in their knee condition. However, some participants expressed concerns about aspects of the program delivery and encountered learning and technological challenges, with a few dissatisfied with their outcomes. To our knowledge, this is the first study to evaluate user engagement and experience with an online unsupervised Tai Chi intervention in people with knee osteoarthritis.

Overall engagement with the Tai Chi program was good, with nearly three quarters of participants deemed as having “acceptable” adherence. Compared to 2 other online unsupervised exercise programs we previously developed for knee osteoarthritis, one a 12-week yoga program [21] and the other a 6-month strengthening program [22], the EARS score in our study (median score of 20 out of a maximum score of 24) was higher than those reported in the other programs (both around 15-16) [21,22], suggesting that participants were more adherent with the Tai Chi than with these other programs. Furthermore, in contrast to another 9-week online unsupervised physical activity intervention for knee osteoarthritis (“Joint2move”) [23], which had an adherence rate of 46% (defined as 6 out of 9 modules completed) [24] and where module completion declined from 80% to 40% over time [23], adherence to our Tai Chi intervention remained stable (71%-80%) throughout the 12-week period, indicating sustained and consistent engagement. However, our adherence rates may be inflated as participants with missing adherence data were excluded from the calculation, which can introduce bias. Our recalculated adherence rates were lower when we included these participants and categorized them as nonadherent. One possible reason for the good engagement with our Tai Chi intervention is the incorporation of an adherence support tool, the “My Exercise Messages” app. This app was adapted from our previously developed, behavioral change theory-informed, automated SMS text messaging program, which showed improved self-reported adherence to unsupervised home exercise for people with knee osteoarthritis [25]. Given that 75% of our program completers reported using the app, it is possible that it contributed to the good adherence to our Tai Chi intervention. This is consistent with findings from systematic reviews showing that the use of behavior change techniques is associated with improved exercise adherence [26,27].

The generally positive user experiences with our Tai Chi program, as captured by the quantitative measures, align with findings from a systematic review [28] and a scoping review [29], showing that older adults reported high satisfaction, usability, and acceptability toward Tai Chi delivered remotely via synchronous videoconferencing. Specifically, satisfaction with the Tai Chi intervention (median of 9 out of 10) was slightly higher than that reported in our yoga program (mean of 8 out of 10) [21], and usability of the Tai Chi intervention (median of 85 out of 100) was higher than that of our strengthening program (mean of 5.3 out of 7) [22]. Under the credibility domain, the relatively low believability score (median of 6 out of 10) when participants first commenced the program likely reflects their initial uncertainty about Tai Chi, with their subsequent engagement and generally positive experiences suggesting that believability may have increased over time had this been remeasured. However, qualitative open-text responses in our study showed that while some participants described the program as well-structured with clear instructions, others found it too slow, repetitive, and boring due to excessive instructions. These contrasting experiences are consistent with our qualitative interview exploring in-depth experiences of a subgroup of 20 people who participated in this Tai Chi program [30], as well as other studies exploring home-based Tai Chi (delivered either as an addition to in-person classes or via videoconferencing) for people with chronic diseases (eg, lower back pain, dementia, and cystic fibrosis) where users also reported variable experiences [31-33]. Given that Tai Chi and yoga share similar elements of movement-based mind-body connection, we can draw further insight from 2 qualitative studies exploring yoga instructors’ experience of adapting in-person yoga classes to synchronized online delivery via videoconferencing [34,35]. These studies found that the online classes were intentionally designed to be slower, simpler, and more structured than face-to-face classes, where the instructor increased instructions and emphasized interoception (a technique of focusing on internal bodily sensations to improve mind-body awareness) [34-36]. As our Tai Chi program was delivered via pre-recorded videos without any instructor interaction with participants, we deliberately designed it to be slower and more repetitive to accommodate for the absence of hands-on assistance and direct feedback from an instructor. However, this teaching format might be perceived differently for individuals with variable learning preferences and needs.

The learning and technological challenges encountered by some participants may have also negatively impacted their experience with the program. Similar difficulties have been reinforced in our qualitative study [30] and reported in other studies of online exercise programs for older adults, where participants struggled to learn movements through a two-dimensional interface and experienced various internet-related issues [34,37-39]. The nature of an online unsupervised learning format may further reduce participants’ engagement, given the lack of face-to-face interaction and social support from being part of a group-based class. Collectively, our findings align with previous research showing that factors such as preferred teaching styles in mind-body exercise, movement complexity, digital literacy, and supervision are important in influencing participants’ engagement and experience with online mind-body exercise programs [29,40-42].

The fact that many people with knee osteoarthritis reported positive engagement and experiences with our online unsupervised Tai Chi program, but some others had less favorable experiences, indicates that there is no one-size-fits-all approach to exercise programs. This is supported by a Cochrane review showing similar symptomatic benefits can be obtained from a range of different exercise types, dosage, and delivery modes in knee osteoarthritis [43]. Instead, offering and improving access to a variety of exercise options that align with patient preferences may be important for optimizing engagement, adherence, and outcomes. For future implementation, providing additional information about the Tai Chi exercises (such as readings or videos) prior to program commencement may help participants better prepare for online unsupervised exercise programs.

In addition, offering a suite of optional shorter, prerecorded videos may enable users to select sessions that best suit their needs, while also helping to ensure they are adequately challenged. Finally, a hybrid approach that combines in-person sessions with digital Tai Chi could provide participants with initial instruction and feedback, enhancing their confidence before transitioning to independent online practice.

To date, more than 47,747 users from over 141 countries have accessed our freely-available unsupervised online-strengthening [44] and yoga [45] programs, highlighting the global demand for unsupervised home-based resources for osteoarthritis management. Our online Tai Chi program now adds to the growing suite of online exercise resources, providing an additional accessible option to support patient uptake of osteoarthritis exercises and potentially alleviate demands on health care services.

Strengths of this study include a mixed methods process evaluation embedded within a robust RCT. By combining both quantitative and qualitative data, the study provided a more nuanced understanding of user engagement, program experiences, and considerations for future implementation. However, several limitations should be acknowledged. First, the process evaluation completion rate was moderate, with almost one-third of participants in the Tai Chi group not completing the evaluation. This may have introduced bias. For example, if noncompleters were those who either did not engage with the program or had less positive experience, the positive outcomes reported may have been overestimated, limiting the representativeness of the findings. Second, we were unable to capture accurate website continuous adherence metrics for the Tai Chi program, which would have provided a more objective measure of adherence. Instead, adherence data in this study were self-reported by participants, which may have introduced social desirability and recall biases that can affect the data accuracy [46,47]. However, a systematic review showed that self-reported diaries measuring the frequency of unsupervised interventions are the most commonly used method for assessing adherence [48]. While this approach has limitations, it remains a practical option when objective methods are not feasible. Future work could incorporate objective adherence metrics, such as those obtained from wearable technology or app-based logs to triangulate findings. Third, qualitative data were collected via an optional open text box allowing a maximum of 3 responses, which may mean our qualitative data may lack granularity. However, our qualitative interviews, published elsewhere [30], provide deeper information from a smaller subset of participants that complement the findings of the present study. Finally, participants were largely older adults, highly educated (university degree or higher), and located in major cities, and they were also required to own a device and access the internet. This may limit the generalizability of our findings to more culturally and linguistically diverse groups, socioeconomically disadvantaged populations, and people with lower digital or health literacy. Future studies could evaluate the program in broader populations to ensure its acceptability and applicability.

Among participants who undertook a 12-week online unsupervised Tai Chi intervention and completed process measures, the majority were adherent with the weekly Tai Chi program, reported high satisfaction with the program and were likely to recommend it to others. The program was perceived as credible, usable, and acceptable. However, some participants expressed concerns related to program delivery, encountered learning and technological challenges, and a few were dissatisfied with their outcomes. This free and scalable Tai Chi intervention has the potential to enhance patient access to guideline-recommended osteoarthritis exercise.