This study adopted a convergent parallel mixed methods design, in which we collected, analyzed, and converged both quantitative and qualitative data. This formative evaluation served a dual purpose: to assess the usability of the app-based intervention and its integrated features among individuals with SCI, and to capture user input to inform further refinements and customizations implemented in conjunction with the app developers specific to the study needs.

Participants with SCI were recruited via an official study website [26] or locally from a rehabilitation hospital’s clinical registry between March and July 2024. Eligible participants were English-proficient adults aged 18 years or older who had been diagnosed with SCI at least 1 year prior. Participants had either an incomplete injury at the C3 level or below or complete paraplegia and were not currently meeting the recommended SCI exercise guidelines [27]. Exclusion criteria included severe cognitive or visual impairments, inability or unwillingness to use a smartphone, or inability to download the study app.

After the initial screening, eligible participants met with the research team to complete the informed consent process, provide demographic and medical history information, and undergo a physical evaluation with the study’s physical therapist. During this session, participants received hands-on training on the study app and were prescribed a customized exercise program by a licensed physical therapist, tailored to their physical capacity, injury level, and personal goals. Before app setup, each participant completed a brief clinical assessment (including muscle strength testing, range-of-motion evaluation, and spasticity assessment) and a goal-setting session with the physical therapist to identify personal exercise goals. These findings informed the selection of appropriate exercises from a library of video-guided routines, with frequency and intensity tailored to SCI-specific guidelines and each participant’s functional abilities. The personalized program was then uploaded directly to the participant’s app account. Throughout the trial, participants could communicate with their physical therapist via in-app messaging and leave comments on their exercise sessions. The program was adjusted as needed in response to feedback and progress. Over the following 2-week trial, participants were instructed to complete and log their personalized exercise program at least twice per week and answer 4 short EMA surveys within the app daily. The EMA surveys were designed to capture participants’ real-time experiences with exercise, including perceived barriers (eg, low motivation, lack of equipment) and facilitators (eg, social support, energy levels), as well as exertion and symptoms during or after exercise. In this formative study, the EMA component also examined how feasible it was for users to complete several daily surveys within the app while also testing the EMA system’s functionality, scheduling, and delivery in a real-world home setting before moving on to the full efficacy study. However, due to technical issues with the study app, some participants did not receive the daily EMA surveys during their trial period. No other technical problems were reported by participants throughout the 2-week trial.

At the end of the 2-week trial, participants completed 3 standardized implementation measures [28]: the Acceptability of Intervention Measure (AIM), Intervention Appropriateness Measure (IAM), and Feasibility of Intervention Measure (FIM), along with the User Version of the Mobile Application Rating Scale (uMARS) to evaluate app usability and perceived quality [29,30]. Surveys were administered electronically via the REDCap platform. Research staff were available to help participants with survey completion if necessary. Specifically, AIM, IAM, and FIM each have 4 items rated on a 5-point Likert scale, from 1 (“completely disagree”) to 5 (“completely agree”), which measure users’ perceptions of the app’s acceptability, appropriateness, and feasibility related to the study, respectively. The uMARS uses a 5-point scale (1=“inadequate” to 5=“excellent”) to evaluate the objective quality of the app across four domains: engagement, functionality, aesthetics, and information, as well as the subjective quality of the app. uMARS also includes items evaluating the perceived impact of the app on users’ knowledge, attitudes, intentions to change, and the likelihood of actual change in the target behavior (ie, adhering to exercise guidelines) specific to the study. We defined a score of 4 or higher as a benchmark for high acceptability, appropriateness, feasibility, and quality [28,30].

Qualitative data were gathered through three focus groups conducted between May and August 2024, following participants’ completion of the postuse surveys. These sessions aimed to contextualize the quantitative findings and further explore participants’ experiences and perceptions of the intervention. Each focus group included 2 to 6 participants, lasted between 45 and 90 minutes, and was moderated by trained research staff. We adopted the focus group guide built from our prior development study [25] with questions guided by the Consolidated Framework for Implementation Research [31,32] to understand participants’ perceptions of app features, performance, personal opinions of the application, and suggestions for future improvement related to the study. Focus groups were conducted through Microsoft Teams, with both audio and video recorded, securely stored, and transcribed using Sonix software (Sonix Technologies). Deidentified transcripts were uploaded to a secure server and imported into NVivo software (version 13.1.7.1; Lumivero) for qualitative analysis. We had 1 participant who was unavailable to join the 3 scheduled focus groups. Instead, we interviewed this participant through Microsoft Teams after the 2-week trial.

Quantitative data were analyzed with descriptive statistics to summarize responses from the AIM, IAM, FIM, and uMARS. To explore potential variation in outcomes across participant characteristics, descriptive subgroup analyses were conducted by gender, primary mobility mode, injury level, and employment status. For each subgroup, mean (SD) values were calculated for all quantitative measures. Given the small sample size, no inferential statistics were computed, and all findings are reported as exploratory and descriptive only.

Transcripts were analyzed using an inductive-deductive hybrid thematic analysis approach, a qualitative analysis method commonly employed in mixed methods research [33]. To converge information from quantitative measures and qualitative focus groups, we first developed an initial codebook using uMARS domains as the high-level themes. Two primary coders, a project coordinator and a research trainee, conducted the qualitative analysis. Both had formal training in qualitative methods under the supervision of an experienced qualitative researcher and had prior experience in rehabilitation research. Neither coder had any personal or financial relationship with the app developer. To enhance analytic rigor and reduce potential bias, both coders independently reviewed and coded each transcript in NVivo, and they discussed and reconciled any discrepancies they encountered, refining the codebook as needed. The coders met regularly to review and harmonize their coding across multiple transcripts. When consensus could not be reached through discussion, a postdoctoral researcher with expertise in qualitative methods served as the final arbitrator. Through an iterative process, we refined the codes and organized them into themes and subthemes. Frequencies reported in the qualitative results reflect the number of coded references across all transcripts rather than the number of individual participants who mentioned each theme, consistent with classical content analysis frameworks for focus group data [34]. A single participant may therefore contribute more than one coded reference to a given theme.

In this process, we converged both quantitative and qualitative results by comparing the results of those measures with the codes and themes. This triangulation of findings helped us develop a richer, more nuanced understanding of participants’ user experiences with the app-based intervention.

The study was approved by the Institutional Review Board of Northwestern University (STU00216188). All participants provided written informed consent before enrollment. The study procedures adhered to the principles of the Declaration of Helsinki and institutional ethical guidelines. All data collected were deidentified before analysis to protect participant privacy and confidentiality. Participants received a US $100 honorarium upon completion of all study activities.

Of the 61 participants initially recruited, 33 underwent phone screening for eligibility. Fourteen met the criteria and agreed to participate in the study. Of these, 86% (12/14) completed all study activities and were included in the final analysis (Figure 1), indicating good retention and supporting the protocol’s feasibility. Two participants were unable to complete the required activities due to unexpected medical complications. Participants’ demographic and clinical characteristics are presented in Table 1. The cohort had a mean age of 52 (SD 11) years and was evenly split by gender: 50% (6/12) male and 50% (6/12) female. Most participants identified as White individuals (8/12, 67%), and 50% (6/12) had a thoracic-level SCI.

App usage data provided objective behavioral indicators of feasibility and engagement during the 2-week trial. Seventy-five percent (9/12) of participants accessed the app at least once per week, with a median of 1.75 (IQR 0.88‐2.63) active days per week. Additionally, 58% (7/12) of participants left at least one in-app comment during the trial period, with a median of 3.5 (IQR 0‐8, range 0‐30) comments per participant. In-app comments included notes on exercise modifications, pain or discomfort during activities, and self-directed progressions (Multimedia Appendix 1).

Sixty-seven percent (8/12) of participants rated the program as highly acceptable and feasible (scores of 4 to 5), and 50% (6/12) rated it as highly appropriate (Figure 2, left panel). Among the 4 uMARS domains of objective quality, participants rated information most positively, with 67% (8/12) of responses in the high range (4-5; Figure 2, right panel). Aesthetics followed, with 42% (5/12) of responses in the high range (4-5). Functionality showed a broader distribution, with most ratings (7/12, 58%) in the moderate range (3-4), followed by 25% (3/12) in the top range (4-5) and 17% (2/12) in the low range (2-3). Engagement received the least favorable responses, with only 8% (1/12) of participants rating it in the high range (4-5), and 33% (4/12) rating it in the low range (2-3). Ratings for subjective quality and perceived impact were generally favorable. Eighty-three percent (10/12) of participants rated the perceived impact above 3, with 42% (5/12) rating it in the high range (4-5), and only 8% (1/12) rating it in the very low range (1-2). Similarly, most participants (8/12, 67%) rated the subjective quality above 3, while only 8% (1/12) rated it in the very low range (1-2). Table 2 shows the descriptive statistics (mean and SD) of the AIM, FIM, IAM, and uMARS scores.

Female participants tended to rate the app more favorably across all domains, with the largest differences observed in perceived impact (female: mean 4.22, SD 0.87; male: mean 3.36, SD 0.99) and AIM (female: mean 4.21, SD 0.40; male: mean 3.46, SD 0.64). Wheelchair users rated the app higher than walking-only users across most domains, especially for the FIM (wheelchair: mean 4.38, SD 0.48; walking: mean 3.38, SD 0.48) and aesthetics (wheelchair: mean 4.19, SD 0.48; walking: mean 3.08, SD 0.76). Lumbar-level participants rated information (mean 4.33, SD 0.38) and perceived impact (mean 4.17, SD 0.58) higher than those of the cervical and thoracic groups. Employment status showed minimal variation across most domains. These patterns should be examined in future studies using larger, more diverse samples.

This study aimed to evaluate the acceptability, appropriateness, feasibility, and quality of a customized mHealth app designed to support home-based exercise participation among individuals with SCI. Using a mixed methods approach that converged quantitative survey data and qualitative feedback from focus groups, we gained a comprehensive understanding of users’ experiences, perceptions, and recommendations for improving the app specific to the study’s needs.

The intervention was generally viewed as acceptable, appropriate, and feasible, as reflected in both the survey responses and participant narratives. Among the uMARS domains, the information domain emerged as the highest-rated component and a key strength of the intervention. Users appreciated the app’s structured content, the relevance of the prescribed exercises, and the clear instructions provided through written guidance and demonstration videos. Participants valued the credibility of the app’s content because it was developed with input from SCI-specialized physical therapists and affiliated with a reputable rehabilitation institution, elements known to increase user trust and engagement in digital interventions [35,36]. These results align with the findings of Bernard et al [20] and Hayat et al [37], which noted that multimodal content can greatly enhance accessibility and comprehension for users with diverse functional needs.

Engagement varied across users. App usage data confirmed that the majority of participants accessed the app regularly and left in-app comments throughout the trial, indicating a meaningful level of interaction with the platform. These comments documented exercise modifications, physical responses, and self-directed progressions, providing the clinical team with ecologically valid, real-world insights into each participant’s exercise performance, information that can directly inform ongoing program refinement and personalization. However, the level of engagement varied considerably across individuals, potentially reflecting differences in motivation, physical capacity, or technology familiarity. Qualitatively, while some participants appreciated the flexibility of scheduling and progress tracking, others reported limited interaction and support for maintaining motivation. Technical challenges with reminders and EMA delivery, along with the absence of interactive feedback or adaptive goals, were seen as barriers to deeper engagement. Notably, issues with EMA delivery, in which some participants did not consistently receive daily surveys as intended, likely reduced opportunities for real-time interaction with the app and may have contributed to the comparatively lower ratings observed in the Engagement domain of the uMARS and the FIM. These findings should therefore be interpreted in light of this technical limitation. Participants expressed interest in enhancements such as motivational messages, reward-based features, and more personalized reminders. These observations align with the broader mHealth literature emphasizing that personalization and adaptive feedback are critical for sustained engagement in populations with chronic conditions or mobility limitations [25,38]. Similar findings have been reported in tele-exercise programs for populations such as individuals with stroke and multiple sclerosis, where user-centered, adaptive features were key to feasibility [39,40].

Feedback on the Functionality was mixed. While participants appreciated the step-by-step exercise instructions, they noted inconsistencies in navigation, difficulties with exercise terminology, and a need for onboarding tools, such as tutorial videos. These findings echo those of Haley et al [35] and Brown et al [41], emphasizing that ease of learning, early-stage guidance, and intuitive design are essential for successful mHealth adoption, particularly among users with varied levels of digital literacy.

Aesthetic design was generally viewed favorably, with users appreciating the layout and visual clarity. Suggestions for improvement focused on progress graphics and charts, consistent with prior evidence that visual analytics and feedback tools enhance motivation and reinforce behavior change in home-based rehabilitation programs [42]. These suggestions highlight that effective visual design focuses not just on appearance but also on promoting behavior change.

The perceived impact was positive among participants. Many described the app as helpful in increasing accountability and supporting the integration of exercise into their daily routines. The association with trained clinicians was highlighted as a factor that boosted participants’ confidence and the perceived value of the program. These findings are consistent with studies by van Eetvelde et al [43] and Nataletti et al [25] emphasizing the importance of tailored, clinically credible digital tools in supporting behavior change and self-management in rehabilitation populations. The higher ratings for perceived impact than for engagement reflect an important conceptual distinction. Perceived impact reflects participants’ beliefs about the intervention’s potential to improve their health behavior, shaped primarily by the quality and personalization of the app’s content and by the credibility of clinician involvement. By contrast, engagement depends on specific interactive features, such as notifications, real-time feedback, and adaptive goal setting, that sustain active app use over time. Participants may therefore recognize the value of an intervention while still facing barriers when these features underperform, including the EMA delivery issues documented in this study.

Exploratory subgroup analyses revealed preliminary patterns that warrant further investigation. Female participants and wheelchair users tended to rate the app more favorably across most domains, particularly in adoption intent and perceived impact, suggesting the app may resonate differently across gender and mobility profiles. Lumbar-level participants rated information quality and perceived impact slightly higher than those in other injury groups, while employment status showed minimal variation across domains. These patterns are preliminary, given the small and unequal subgroup sizes, and should be interpreted with caution. They highlight the importance of examining subgroup differences in future studies using larger, more diverse samples.

This study has several limitations. The small sample size, geographically limited recruitment, and missing neurological classification data may constrain sample characterization and reduce generalizability. The 2-week exposure period precludes conclusions about long-term adherence or sustained engagement with the app; these outcomes will be assessed in the subsequent efficacy study. Additionally, technical issues with EMA delivery prevented some participants from receiving the daily surveys as intended, potentially limiting engagement with the app and affecting the feasibility ratings.

The findings of this study will directly inform the refinement of specific app features by the app developer prior to a subsequent proof-of-concept efficacy study. Development priorities include resolving EMA delivery issues, enhancing notification and reminder features, introducing reward-based engagement mechanisms, and simplifying navigation. Future research should evaluate the effectiveness of the refined mobile app-based exercise intervention in promoting adherence to SCI-specific exercise guidelines and improving health outcomes among a larger, more diverse sample of community-dwelling individuals with SCI over a multimonth intervention period to better capture sustained engagement and long-term outcomes. Objective usage metrics and subgroup analyses across demographic and clinical characteristics should be expanded to better understand engagement patterns and identify who benefits most from the intervention. More broadly, this study underscores the value of centering end-user feedback in the early design stages to develop mHealth tools that reduce barriers to access, promote behavior change, and improve quality of life for populations with mobility needs.

This study highlights the value of a user-centered, mixed methods approach for evaluating an mHealth intervention in individuals with SCI. The intervention was perceived as informative, credible, and accessible and closely aligned with users’ physical activity needs. Meanwhile, participants recognized opportunities to enhance functionality and engagement based on their experience using the intervention as individuals with SCI. As a next step, this user feedback can guide further refinement of relevant app features by the app developers to better meet the usability and accessibility needs of individuals with SCI. Subsequent evaluation through a field test is necessary to determine if the intervention can increase adherence to SCI-specific physical activity guidelines and ultimately improve users’ health and quality of life. These findings align with the wider digital health literature and highlight the importance of customizing design and content to meet user needs. By incorporating these refinements, mHealth interventions can more effectively support exercise adherence, reduce barriers to rehabilitation, and support long-term health in the SCI population.