The present study was a preplanned secondary analysis of data from the experimental arm of a feasibility randomized controlled trial (ClinicalTrials.gov: NCT0511195). The trial included individuals poststroke or TIA randomized into an experimental group receiving the mHealth version of the i-REBOUND intervention or a control group receiving mHealth consultations via videoconferencing [21,22]. The analysis focused on participants’ adherence to and engagement with the mHealth intervention and was therefore limited to the experimental arm. The data were collected between September 2021 and December 2023. This study was reported in accordance with the CONSORT-EHEALTH (Consolidated Standards of Reporting Trials of Electronic and Mobile Health Applications and Online Telehealth) checklist (Checklist 1) [23].

The study was approved by the Swedish Ethical Review Authority (dnr 2020‐05062 and 2021‐03622). Participation in the study was voluntary, and participants could withdraw at any time without consequence. No compensation was provided. Individuals who expressed interest received both verbal and written information before providing written informed consent. This consent also included approval for the secondary analysis conducted in the present study. A detailed description of the inclusion procedure and ethical considerations is available in the study protocol [21]. All data were anonymized to ensure participant confidentiality.

Participants were recruited through advertisements at collaborating clinics, social media, patient organizations, and the webpage of Karolinska Institutet. Inclusion criteria were (1) clinical diagnosis of stroke or TIA between 3 months and 10 years prior to study enrollment, confirmed by the participants’ physician, (2) living at home, (3) being able to walk a short distance indoors with or without a walking device, (4) being able to use a smartphone including e-signature identification with or without the help of a relative or carer, and (5) having access to a stable internet connection. Exclusion criteria were (1) already meeting the recommended physical activity levels of at least 150 minutes per week of moderate physical activity or at least 75 minutes per week of vigorous-intensity physical activity, (2) severe health conditions compromising engagement in the intervention, or (3) enrolled in a concomitant clinical trial or participating in rehabilitation (eg, aerobic exercises) at the time point of recruitment.

IBM SPSS Statistics software (version 24.0; IBM Corp) was used for statistical analyses. Baseline characteristics and descriptive statistics for aims 1 to 3 were calculated and presented as mean (SD), median (IQR), and frequencies (%).

For the first and second aim, descriptive statistics described the adherence to and engagement in the mHealth intervention across the study period. Participants who completed the introductory week and began intervention phase 1 were included in the analysis. Those who discontinued the intervention were considered lost to follow-up, with data analyzed up to the week of discontinuation.

For the third aim, between-group differences regarding demographic characteristics between the sub-groups with high and low adherence, and high and low app engagement, were analyzed using Mann-Whitney U test for continuous variables and chi-square test for binary variables.

For the fourth aim, logistic regression analysis was used to examine the associations between high adherence during the intervention, high app engagement during the intervention and postintervention periods, and maintained physical activity levels (categories 3 and 4, as described earlier) from baseline to the 6-month follow-up and from the 6-month to the 12-month follow-up. These analyses included age (years) and use of a walking aid as covariates, as they have been shown to be associated with the level of physical activity in people poststroke [36]. The α level was set at .05 for all analysis.

Of the 57 participants enrolled, 51 (89%) completed the intervention (Figure 3). The mean age of the participants completing the intervention was 71 (SD 8) years, and the median time since stroke or TIA was 2.4 (IQR 1.2-5) years. Of the 51 participants who completed the intervention, 34 (67%) were female, 27 (53%) had a university degree, 8 (16%) were employed, and 47 (93%) reported no or mild poststroke or TIA symptoms (ie, Modified Rankin Scale score 0‐1) (Table 1).

The mean adherence to supervised exercise across the intervention period was 79% (min-max: 29%‐100%) (Figure 4), with 911 of 1151 sessions attended in phase 1 and 489 of 618 sessions attended in phase 2. The mean adherence rate to individual counseling sessions across the intervention was 97% (min-max: 50%‐100%) and among those who completed the intervention, 82% (42/51) attended all counseling sessions.

App engagement was higher during the intervention phase, with an average of 84% of participants using the app weekly (min-max: 69%‐94%), compared to the postintervention period (average 38%; min-max: 18%‐63%) (Figure 5). The majority of participants (42/51, 82%) disengaged during the postintervention period, with an average time to disengagement of 8 weeks (min-max, 0‐23).

Engagement with the chat function and with the activity diary were the most common types of app engagement (Figure 6). The average weekly engagement with the chat function was 63% in phase 1, decreasing to 50% in phase 2 (Figure 6A). On average, participants sent 1.6 weekly chat messages during the intervention. Forty-three of 55 participants (78%) used the activity diary to some extent across the study period (Figure 6C). In total, 103 activities were monitored, and the most frequently monitored activities were outdoor walking (n=29, 28%), aerobic exercise (n=13, 13%), and unspecified physical exercise (n=12, 12%). Among the 43 participants using the activity diary, 14 (33%) demonstrated low usage (<19 wk), 17 (40%) sporadic usage (20‐39 wk), and 12 (28%) frequent usage (>40 wk). Fifty of 55 participants (91%) were prescribed individual exercises, and among those, 33 (66%) had limited engagement (<5 sessions), 12 (24%) sporadic engagement (6‐19 sessions), and 5 (10%) frequent engagement (>20 sessions). The mean engagement with educational videos was 67% (min-max: 0%‐100%), with each participant viewing 5 of 8 videos on average, and 53% (29/55) participants watching all content.

Of the 51 participants completing the intervention, 31 (61%) had high adherence (ie, ≥80%) and 15 (29%) had high app engagement (ie, ≥80%). Most personal and contextual factors, as well as clinical and functional measures, showed no significant differences between participants with high versus low adherence or high versus low app engagement in the mHealth intervention (Table 2). A higher proportion of females (24/31, 77%) demonstrated high adherence to the intervention compared to males (7/31, 23%, χ²₁=4.1; P=.04).

Valid physical activity data were available for 45 participants (88%) at 6 months and 42 participants (82%) at 12 months among those who completed the intervention. Twenty-four participants (53%) maintained their physical activity between baseline and the 6-month follow-up, whereas 19 (45%) maintained their physical activity between the 6- and 12-month follow-ups (Table 3). High adherence during the intervention was associated with maintained physical activity between baseline and the 6 months follow-up (Odds ratio [OR] 12.07, 95% CI 2-72.76; P=.01), and high app engagement during the postintervention was associated with maintained physical activity between 6 and 12 months (OR 5.10, 95% CI 1.02-25.52; P=.05). High app engagement during the intervention was not associated with maintained physical activity between baseline and the 6-month follow-up (OR 2.80, 95% CI 0.55-14.28; P=.22), nor was high adherence during the intervention associated with maintained physical activity between the 6- and 12-month follow-ups (OR 3.53, 95% CI 0.68-18.35; P=.13).

This study investigated adherence and app engagement during and 6 months after an mHealth intervention for people poststroke or TIA. Adherence to supervised sessions remained high (>75%) across the intervention, and more females adhered to the supervised exercise than males. Engagement with self-managed mHealth support was initially high but decreased postintervention. High adherence during the intervention was associated with maintained physical activity between baseline and the 6-month follow-up, while high postintervention engagement was associated with maintained physical activity between the 6- and 12-month follow-ups.

Adherence to supervised exercise during the intervention averaged 79%, which is comparable to adherence rates in nondigital [37] and digital interventions for people poststroke or TIA [20,38]. Adherence to counseling sessions (97%) was higher compared to a previous study showing decreasing adherence to long-term nondigital physical activity support in people poststroke [39]. The current study’s digital format provided an accessible and flexible solution that may have contributed to the high adherence rates. High adherence to supervised phases of exercise programs is common [40]. In this study, weekly interactions with a physical therapist as part of the mHealth intervention likely contributed positively to adherence. Additionally, most participants attended group exercise sessions, which may have further promoted adherence by creating a sense of accountability and commitment to the group [12,38]. Low app engagement within physical activity apps is a well-known challenge in mHealth interventions [41], yet our findings showed high engagement (84%) with self-managed support modules during the intervention, particularly the chat function and activity diary. However, app engagement postintervention decreased to 38%. As with adherence, app engagement during the intervention was likely promoted by regular interaction with a physical therapist, whereas the lack of such contact postintervention may have contributed to the decreasing patterns of engagement [19,42]. Similar patterns of declining engagement have been observed in unsupervised mHealth interventions for people with chronic conditions [43]. While the supervised sessions followed a structured schedule, engagement with the self-managed mHealth support was optional and guided by the participant’s preferences. A recent review suggests that clear guidance on the frequency and type of mHealth engagement can promote maintained participant engagement [43]. Thus, the lack of explicitly defined goals of app engagement in the current mHealth intervention might further explain the decreasing app engagement postintervention.

In line with a meta-analysis of digital physical activity interventions in adults [16], our results suggest that adherence during the intervention and app engagement postintervention is important for maintaining physical activity after the intervention period. This study found no significant differences in high adherence or app engagement between sub-groups, such as age or education level, which has previously been shown to influence mHealth intervention adherence [15]. However, our results suggest that females may have adhered to supervised sessions more than males, potentially because females tend to be more proactive and engaged in health-related issues [44]. The results from this study support that adherence to and engagement with an mHealth intervention might be important factors influencing the intervention’s impact on maintained engagement in physical activity.

Several aspects of the self-managed mHealth support could be improved to enhance the app engagement in the current mHealth intervention, as well as long-term engagement in physical activity beyond the intervention period. Personalized app content commonly benefits app engagement [41]. However, despite access to an activity diary with self-selected activities and individually prescribed exercises, only 28% (12/43) and 10% (5/50), respectively, frequently engaged with these modules in the present study. Engaging mHealth features, such as prompts, gamification, and regular content updates, have been shown to significantly enhance app engagement [19,41,42]. These elements were not incorporated into the current intervention; however, integrating them in future development could be a valuable strategy to encourage maintained app engagement among individuals poststroke or TIA. Another promising approach to enhance engagement in people poststroke or TIA, particularly those with lower digital health literacy, is to develop strategies that actively involve caregivers in participating in the mHealth intervention and mediating treatment delivery. This approach has been tested in on-site interventions [45] but has been rarely implemented in digital formats in stroke trials. Furthermore, a potential refinement of the intervention design could be development of distinct phases, each with specific goals for supporting behavior change. This might include an initiation phase focusing on establishing routines for exercise and physical activity (like the present intervention), followed by a maintenance phase in which therapist support occurs less frequently. This design could be especially important for people poststroke or TIA, who often require long-term support to establish and maintain behavior change [39]. Advancing toward more independent self-management of physical activity would also require extending the intervention’s focus to support participants in developing strategies for sustaining behavior change after the intervention, which should be embedded within the individual counseling and educational materials. We believe that refinement of the mHealth version of the i-REBOUND intervention should be cocreated with people poststroke or TIA and target multiple domains, including technology, intervention design, and the establishment of clear expectations for adherence and engagement [46].

The findings should be interpreted cautiously given the relatively small sample size. In addition, the participants were mostly female, highly educated, and with mild stroke symptoms, which limits the generalizability of the findings to other stroke or TIA populations [47]. Furthermore, the 7000 steps/day cut-off for physical activity, while clinically relevant for optimal cardiovascular risk reduction [35], is a crude measure for detecting change over time. Moreover, participants in the sample exhibited a notably high level of physical activity at baseline, exceeding the average daily step count of 4078 for individuals with stroke [6]. User logs yielded an in-depth examination of the participants’ mHealth engagement behavior but could not sufficiently capture the multifaceted nature of engagement. For a more comprehensive understanding of engagement, exploration of participants’ subjective experiences is required, which we acknowledge as a limitation of the present study. A strength was the 48-week study period, which allowed for a thorough investigation of changes in adherence and engagement patterns over time. However, adherence and engagement levels were classified solely as ‘low‘ (<80%) or ‘high‘ (>80%). For more substantiated levels, future studies should determine adequate levels for effective mHealth engagement and adherence and set predefined benchmarks [43].

Supervised mHealth support was well received, with high adherence. In contrast, the self-management modules for physical activity faced challenges in engaging participants. High adherence during the intervention was associated with physical activity maintenance immediately afterward, while high app engagement postintervention was linked to maintenance at the 12-month follow-up. These exploratory findings may be hypothesis-generating and inform future research. To enhance engagement with mHealth interventions and support maintained physical activity after the supervised phase, future studies could benefit from qualitative and cocreative approaches to better understand and refine self-managed mHealth support for individuals poststroke or TIA.