Cardiovascular disease, including coronary heart disease (CHD), is the leading cause of death in Australia and globally [1-3]. Exercise-based cardiac rehabilitation for people with CHD is associated with significant risk reductions in cardiovascular mortality, hospitalizations, and repeat cardiac events [4]. Even so, 1 in 3 cardiac events are repeat events, and repeat cardiac events increase the risk of premature mortality [5].

Cardiac rehabilitation aims to reduce morbidity through positive lifestyle changes, including increasing physical activity and decreasing sedentary behavior [6,7]. Despite this, the sedentary behavior levels of cardiac rehabilitation participants remain high before, during, and after the program [8-16]. Sedentary behavior is associated with an increased risk of morbidity and all-cause mortality [17-21]. Accelerometry-measured sedentary times of ≥9 waking hours per day place healthy individuals at a significantly higher risk of death [18]. Television viewing times (a self-reported marker of sedentary behavior) in adults with diagnosed CHD or stroke of ≥4 hours per day is associated with a 52% increased risk of all-cause mortality compared with those watching television for ≤2 hours [21]. Breaking up sedentary time more frequently is also associated with decreased systolic blood pressure in cardiac rehabilitation participants [8]. Cardiac rehabilitation participants are likely to benefit by engaging in interventions that reduce sedentary behavior, and further options to support participants may be needed.

Evidence suggests that cardiac rehabilitation participants are interested in support via the internet and mobile phones [22,23], and they have the potential to reduce hospital readmissions [24-26]. Widmer et al [25] reported that cardiac rehabilitation participants using a digital health intervention had significant (P=.04) reductions in weight and blood pressure and a 28% reduction in rehospitalizations and emergency department (ED) presentations compared with those who received traditional cardiac rehabilitation only. We hypothesized that better secondary prevention management of risk factors through cardiac rehabilitation plus the digital health intervention can lead to reduced hospital admissions. Similarly, smartphone apps that reduce sedentary behavior may be a feasible option to reduce hospital admissions. Few studies have targeted sedentary behavior change through smartphone apps in people with CHD [27-29]. These studies are generally small (≤50 participants), short in duration (≤3 mo), and aimed at examining the feasibility of such interventions [30]. Effect sizes have varied in these studies, with one reporting no change in self-reported sitting time [27], and in the feasibility trial preceding this study, a medium reduction in accelerometer-measured sedentary behaviors was reported [28]. One of the largest studies to date targeting sedentary behavior involving a smartphone app, pocket-worn activity tracker, and participants with CHD reported no significant between-group reduction in device-measured sedentary time [29]. However, cardiac rehabilitation participants using the SIT LESS protocol had reduced odds of sitting >9.5 hours per day [29]. Despite varying results, these studies concluded that smartphone apps may be feasible in reducing sedentary behavior in people with CHD and that larger-scale randomized controlled trials are warranted to determine their effectiveness. Therefore, we aimed to test the effectiveness of a sedentary behavior change smartphone app (Vire app and ToDo-CR program) as an adjunct to cardiac rehabilitation for hospital admissions and ED presentations over 12 months. As secondary aims, we examined the effectiveness of the Vire app and ToDo-CR program in decreasing accelerometer-measured sedentary behavior and cost-effectiveness.

Using an assessor-blind, parallel randomized controlled trial design, participants were recruited from 3 phase-2 hospital-based cardiac rehabilitation programs in Canberra, Australia, between January 2020 and December 2021. The study duration for each participant was 12 months (Figure 1). Following the baseline assessment, participants were randomly assigned 1:1 to either usual care cardiac rehabilitation or the intervention: cardiac rehabilitation plus the 6-month behavior change smartphone app commencing at the start of cardiac rehabilitation (Vire app and ToDo-CR program) and followed up at 6 and 12 months at the University of Canberra by a blinded research assistant. The study protocol has been published elsewhere [31]. Participants were aged ≥18 years, had stable CHD, owned a smartphone, had no serious medical or functional impairments, and had adequate English language and cognitive skills [31].

All participants provided written informed consent. Ethics approval was obtained from the Australian Capital Territory Health (2019.ETH.00162), Calvary Public Hospital Bruce (20-2019), and the University of Canberra (HREC-2325) Human Research Ethics Committees. Reporting was informed by the CONSORT (Consolidated Standards of Reporting Trials)–EHEALTH [32] and CONSERVE statements [33] and guidelines for studies affected by the COVID-19 pandemic [34].

The Vire app and ToDo-CR behavior change program was informed by the Do Something Different approach focusing on breaking existing sedentary behavior habits and becoming behaviorally flexible [31,35,36]. The Vire app integrated data from a Fitbit Inspire wearable activity tracker provided to the participants and smartphone GPS data through machine learning to create a comprehensive digital profile of the participants’ current behavior. Using the data, the Vire app sends short, personalized behavior change messages known as Do’s in the form of push notifications 2 to 3 times per week at random over 6 months. The Do’s targeted sedentary behavior, suggesting microbehavioral alternatives designed to disrupt usual habits and encourage small lifestyle changes. Participants were encouraged to download the app, and technological support was provided throughout the trial.

Sample size was based on detecting a significant difference in hospital admissions and ED presentations between usual care and a digital health intervention. A similar study [25] noted a 28% difference in rehospitalization and ED presentations between usual care cardiac rehabilitation (a standard rate of 44%) and cardiac rehabilitation plus a digital health intervention (a standard rate of 16%). Using a 2-sided significance of P<.05 and power of 88% (calculated using G*Power 3.1.9.4), a minimum of 108 participants were needed. Accounting for a 25% dropout rate, 72 participants per group (144 total sample = 108 / [1 – 0.25]) were calculated.

To comply with public health recommendations, there were variations in the methods used in the protocol [31]. The trial was extended by 12 months owing to the closure of the cardiac rehabilitation programs. The required sample size was also reduced from 144 to a minimum of 108 (removing the 30% dropout buffer) owing to the impact of the COVID-19 pandemic [31]. In addition, all participants experienced varying restrictions and closures of nonessential health care services. Due to COVID-19 pandemic restrictions, some follow-up assessments were completed via telehealth (Zoom video call). Participants used their own blood pressure monitors, scales, and tape measures under the instruction of a blinded assessor. The method of assessment was documented throughout the study. All changes were approved by the overseeing ethics review committees.

Data were analyzed according to group assignment using intention-to-treat analyses. Missing data were handled by bringing the last value forward (carryover approach). An on-protocol analysis was also performed. All descriptive statistics were reported using means and SDs, medians, and IQRs or proportions, as appropriate. Normality was assessed using the Kolmogorov-Smirnov test for samples ≥50.

The primary analysis was the comparison of rates of nonelective hospital admissions and ED presentations. Binary logistic regression for “yes” versus “no” admissions (odds ratios) and negative binomial with log link regression for the rate or number of admissions (incidence rate ratio) were completed. Survival analyses (Cox regression) were completed to consider the time frame to admission (hazard ratios). Adjustments were made for sociodemographic variables (eg, age and sex) and other covariates (eg, diabetes and other chronic diseases).

Linear mixed-effects models for repeated measures were used to analyze all other secondary outcomes. The maximum likelihood method was used for parameter estimation. Time and between-group comparisons were explored as fixed effects while adjusting for demographic characteristics (eg, age and sex) and other covariates (eg, education, employment, and accelerometer counts/d) [49-51]. Participants were treated as random effects (ie, random intercept models), and the intraclass correlation coefficient was reported. The model with best fit was informed by the Akaike Information Criteria [52]. All estimated effects (β) are reported with their associated 95% CIs. The correlation between outcomes and engagement with the Vire app and ToDo-CR program was also explored.

Sensitivity analyses were completed for outcomes self-administered by participants under telehealth conditions (eg, waist circumference and blood pressure) and for participants wearing their accelerometer belt during COVID-19 pandemic lockdowns and restrictions. Subanalyses were completed for participants excluded versus participants consented, dropouts versus those who completed the study, and those with prior experience with a physical activity tracker versus those with no experience. Data were analyzed using SPSS (version 27; IBM Corp).

Effectiveness was measured in terms of the secondary outcomes. The incremental cost-effectiveness ratios (ICERs) were calculated using the following formula:

A total of 120 participants were recruited for this trial (Figure 1). Participant characteristics are reported in Table 1. The majority were male (93/120, 77.5%), tertiary educated (95/120, 79.2%), and employed (62/120, 51.7%). Approximately half (55/120, 45.8%) of the participants had prior experience using physical activity tracking apps or wearable trackers. The participants who were assessed for eligibility and excluded (n=114) were significantly older (67 vs 63 y; P=.003), from public hospitals (P=.001), and less likely to have had a percutaneous coronary intervention and more likely to have had a myocardial infarction (P=.01; Table S1 in Multimedia Appendix 1). The main reason for exclusion was declining to participate, with the primary reason for declining being “not interested in smartphone apps” (28/93, 30%). The only reasons for not meeting the inclusion criteria were not having a smartphone (14/21, 67%) and having an incompatible smartphone (7/21, 33%; Table S2 in Multimedia Appendix 1). Retention rates were high in both groups (intervention 49/60, 82% vs control 56/60, 93%); however, there was a significant difference between the 2 groups’ retention rates (P=.02) but no difference in characteristics (Table S3 in Multimedia Appendix 1).

Nonelective hospital admissions and ED presentations for all participants are shown in Table 2. The most frequent cause of admission was chest pain (AR-DRG code F74B). The results of the logistic regression models on the likelihood that participants have a nonelective hospital admission or ED presentation are reported in Table 3. After adjustment for age, sex, and the presence of diabetes and other chronic diseases, those in the intervention group were 1.54 times more likely to have an admission, 3.26 times more likely to have a cardiac-related admission, and 2.07 times more likely to have an ED presentation than those in the control group.

Group allocation was not a significant predictor of the incidence (number) of all-cause or cardiac-related admissions or ED presentations (Table 4). The admission rate among participants in the intervention group was 1.69 times higher than the rate in the control group for all-cause admission, 1.56 times higher for cardiac-related admissions, and 1.90 times higher for ED presentations following adjustment for age, sex, and the diagnosis of diabetes or other chronic diseases (Table 4).

Results of the Cox regression analysis to determine if there were differences in time to admission or ED presentation between the intervention and control groups are presented in Table S4 in Multimedia Appendix 1. The intervention group participants had 1.52 times the probability of experiencing an all-cause admission in 12 months compared with the control group. They also had 3.14 times the probability of experiencing a cardiac-related admission and 1.84 times the probability of experiencing an ED presentation when adjusted for age, sex, and diagnosis of diabetes and other chronic diseases (Table S4 in Multimedia Appendix 1).

Sedentary behavior and physical activity measures are reported in Tables 5 and 6 and Table S5 in Multimedia Appendix 1. Both the intervention and control groups showed an increase in sedentary behavior over 12 months, spending approximately 10 hours per day in sedentary behaviors (Table S5 in Multimedia Appendix 1). There were no significant between- or within-group differences in any sedentary behavior measures, although there was a small effect size for the reduction in sedentary behavior (min/d) at 6 (Cohen d=0.11) and 12 months (Cohen d=0.21) in favor of the intervention group compared with the control group. At 6 and 12 months, the control group engaged in 15 minutes (P=.54) and 22 minutes (P=.33) more sedentary behavior than the intervention group when adjusted for age, sex, VM counts, employment, and education (Tables 5 and 6). In subanalyses, those in the intervention group who had prior experience with physical activity trackers (n=30) spent a lower percentage of the day in sedentary behavior (mean difference 6.13%, 95% CI 0.97%-11.28%; P=.02), had shorter sedentary bouts (mean difference 2.32 min, 95% CI 0.36-4.27; P=.02), and had lower overall sedentary minutes per day (mean difference 65.22 min, 95% CI −13.29 to 143.72; P=.10) at 6 months on the completion of the intervention (Table S6 in Multimedia Appendix 1). There were no significant differences in the control group based on prior physical activity tracker use (Table S7 in Multimedia Appendix 1).

There were no significant between-group differences at 6 or 12 months in any physical activity measures, except for wear time at 12 months (Table S5 in Multimedia Appendix 1). Nonetheless, the control group showed a significant within-group increase in VM counts at 6 months and light-intensity physical activity at 12 months. The results of the linear mixed-effects models are reported in Tables 5 and 6, and all models were nonsignificant.

All other secondary outcomes are presented in Table S8 in Multimedia Appendix 1. The control group had a higher BMI (Cohen d=0.30), waist circumference (Cohen d=0.43), and waist-to-hip ratio (Cohen d=0.33) at 6 and 12 months (Table S8 in Multimedia Appendix 1). Using linear mixed-effects models, BMI, waist circumference, and waist-to-hip ratio remained higher in the control group after adjustment for age, sex, employment, and education (Table S9 in Multimedia Appendix 1).

There was no significant between-group difference for systolic and diastolic blood pressure, 6-minute walking distance, anxiety, or depression (Tables S8 and S9 in Multimedia Appendix 1). There was a significant between-group difference in favor of the control group for quality of life (Cohen d=0.14; P=.03; Table S8 in Multimedia Appendix 1). In linear mixed models, the control group experienced a significantly higher overall quality of life at 6 and 12 months when adjusted for age, sex, employment, and education (Table S9 in Multimedia Appendix 1).

On the basis of the University of Rhode Island Change Assessment Scale readiness to change scores, participants were in the precontemplation stage throughout the study regarding physical activity change. There were no significant within-group or between-group differences (Table S8 in Multimedia Appendix 1) or significant differences over time (Table S9 in Multimedia Appendix 1). No significant differences were observed in sensitivity analyses for secondary outcomes between those wearing accelerometers during the COVID-19 pandemic lockdowns or completing telehealth measures.

The median completion rate of Dos was 3 of 55 (IQR 0-37.75) over 6 months. There were 27% (16/60) of participants who engaged with the app for the entire 6 months (ie, completed at least 1 Do/mo), 33% (20/60) of participants who engaged with the app less than once per month, and 40% (24/60) of participants who did not complete any Dos. In logistic regression modeling, those who had used apps or wearable activity trackers before were 1.04 times (P=.01) more likely to complete Dos (χ²₁=7.1). Those who were employed full time or part time were 0.97 times (P=.01) less likely to complete Dos (χ²₁=6.6). There were no statistically significant correlations among the total number of Dos completed and age (r=−0.07; P=.60), sedentary behavior (r=−0.15; P=.25), physical activity (r=0.21; P=.11), quality of life (r=0.18; P=.17), anxiety (r=−0.12; P=.35) or depression (r=−0.15; P=.26) at 6 months (Table S10 in Multimedia Appendix 1).

The usability of the Vire app (Unified Theory of Acceptance and Use of Technology) is presented in Table S11 in Multimedia Appendix 1. Participants were relatively satisfied with the usability of the Vire app, with a median score of ≥4 in all constructs except for habit (score=3.54/7) and use (score=3.29/7).

The use of the Vire app and ToDo-CR program was not effective in reducing hospital admissions and ED presentations nor did it significantly decrease sedentary behavior compared with usual care over 12 months. Participants in the intervention group were more likely to have a cardiac-related hospital admission; however, the costs of these admissions were markedly lower than those in the control group. Although the intervention was costlier to implement, it was also more effective at reducing sedentary behavior, BMI, and anxiety and increasing quality of life and light-intensity physical activity. Retention rates were high in this study; however, engagement with the Vire app and ToDo-CR program was low. Furthermore, there was no correlation between age and engagement with the Vire app and ToDo-CR program; instead, there was a correlation between prior experience with physical activity trackers and apps.

Intervention participants were 50% more likely to have a hospital admission, twice as likely to have an ED presentation, and 3 times more likely to be admitted to the hospital for a cardiac-related hospital admission. This contrasts with the studies by Widmer et al [25] and Rivers et al [26], who each noted an approximate 30% decrease in the rate of hospital admissions in the groups using an app-based intervention in cardiac rehabilitation. These studies targeted a range of lifestyle risk factors (eg, physical activity, smoking, diet, and medication adherence) rather than 1 risk factor. Perhaps by targeting only sedentary behavior, the strength of the intervention was not enough to produce significant changes in the primary outcome, noting that multiple factors contribute to hospital admissions and risk factor control such as medication and comorbidity management [53].

Cardiac rehabilitation is associated with significant reductions in hospitalizations and repeat cardiac events [4]. Nonetheless, of those who are referred for cardiac rehabilitation in Australia, only 28% attend [54]. Alternate technology-based methods (such as the Vire app and ToDo-CR program) may therefore be better suited to reach those not attending traditional cardiac rehabilitation who are at a higher risk of hospitalizations compared with those who attend traditional cardiac rehabilitation. This group is missing the behavior change advice and support provided in cardiac rehabilitation and may have the most to gain by engaging. Providing the option of an app-based cardiac rehabilitation program is associated with increased overall cardiac rehabilitation participation rates [26] by removing barriers such as the need to travel. In addition, app-based cardiac rehabilitation programs have been shown to achieve outcomes comparable with traditional programs [24,55,56]. One study reported readmission data comparing no cardiac rehabilitation, traditional cardiac rehabilitation, and app-based cardiac rehabilitation [26]. Although this study was not designed to evaluate differences in readmissions, cardiac-related admissions were lower in those using the app (1/23, 4%) versus those receiving no cardiac rehabilitation (5/39, 13%) [26]. Larger-scale studies are required to investigate the effect of smartphone apps in those declining cardiac rehabilitation and the possible benefits to hospitalizations.

Those in the intervention group attended the ED more frequently than those in the control group, and the most frequent cause for admission was chest pain (AR-DRG code F74B). Presenting to the ED for chest pain is particularly relevant with advice given in the Vire app and ToDo-CR program to seek medical attention if participants experienced these symptoms. These participants may have become more proactive in self-managing their CHD because of the education provided. Public education campaigns to raise awareness of the signs and symptoms of acute chest pain have been shown to be effective in increasing the rate of ED presentations for early medical intervention [57,58]. Despite having more admissions, the intervention group costs associated with cardiac-related admissions were lower (intervention, Aus $252.40 vs control, Aus $859.30), meaning that when they were admitted, they potentially had shorter length of stays for less-severe diagnoses, requiring less-costly medical intervention and overall less impact on the health care system. Delays in presenting to the ED with acute chest pain, a potential indicator of a myocardial infarction, contribute to patient morbidity and mortality [57,59]. Smartphone apps are being developed to support patients with CHD and incorporate educational features regarding symptoms of chest pain, with a key focus being to support participants rather than just monitor them [60]. These types of features and providing education may be especially important in future studies for those not attending face-to-face cardiac rehabilitation programs.

Although the effect size for reduction in sedentary behavior was small in this study, an approximate 20-minute reduction may still be clinically meaningful. In the limited studies available that test the effect of smartphone apps on sedentary behavior [30], greater reductions have been seen compared with this study in cardiac rehabilitation participants (accelerometer-measured 100-min reduction/d at 4 mo [28]; accelerometer-measured 96-min reduction/d at 3 mo [29]); stroke (accelerometer-measured 60-min reduction/d at 6 wk [61]); and chronic stroke (self-reported 180-min reduction/d at 3 mo [62]). There is no well-established, minimal clinically important difference for sedentary behavior in people with cardiovascular disease including CHD. Studies in the general population have reported that for every 30 minutes of sedentary behavior reallocated to light-intensity or moderate to vigorous intensity physical activity, there is a 2% to 25% improvement in cardiovascular disease risk biomarkers [63], and a 1- to 2-hour reduction in television viewing time is associated with reductions in cardiovascular disease risk (eg, waist circumference, BMI, triglycerides, insulin sensitivity, and blood pressure) [64,65]. In addition, a break in sedentary behavior as short as 1 minute can reduce waist circumference and improve C-reactive protein (inflammatory marker) independent of total sedentary time [65,66]. This study showed minimal change in sedentary bouts and breaks between groups or over time. Sedentary behavior bouts and breaks are also infrequently reported [30]. Future iterations of apps such as the Vire app and ToDo-CR program may benefit from providing specific advice to reach such targets and help with creating more substantial levels of change [67].

The waist-to-hip ratio (0.03) and BMI (1.64 kg/m²) significantly reduced in the intervention group; however, they were unlikely to be clinically significant changes. The intervention also had a statistically significant 6-cm reduction in waist circumference, which could be considered clinically significant after accounting for measurement error (≥2 cm) [68]. These results combined may indicate an overall lower central adiposity and lower cardiovascular disease risk profile, in line with previous app-based studies in CHD participants [25].

A high percentage of the intervention group did not complete a single Do message (24/60, 40%), which was used as the marker of engagement. There may have been a misunderstanding of the need to tick off the Dos after viewing them or, alternatively, the intervention did not appeal to them. Previous studies on cardiovascular disease and smartphone apps have reported similarly low engagement and adherence levels [25,62,69-71] and that levels tend to decrease with time in the intervention [30]. Similar studies have also reported technical difficulties as a key reason for low engagement [30,62,69,70,72]. Low levels of engagement, as seen in this study, could result in an engagement level that is insufficient to achieve the intended effect [73], with increasing evidence that digital health apps for chronic disease self-management require ongoing patient engagement as a key determinant of overall clinical impact [74-77].

In line with the feasibility trial preceding this study [28], not having a smartphone was a major reason for exclusion. Among those who declined to participate, the main reason provided was not being interested in smartphone apps. Our study was consistent with previous studies, with people declining to participate being significantly older [28,30] and younger cardiac rehabilitation participants being more likely to use smartphones [22,23,30,78]. Patient preference for the type of intervention may be key to their success [26,67]. In a study comparing multiple delivery methods for cardiac rehabilitation including app-based cardiac rehabilitation, approximately 25% of the group declining the app-based approach listed technology issues as the main reason for nonparticipation [26]. Perhaps key messaging moving forward is that those who want to use smartphone apps will engage regardless of age [67].

Although smartphone apps have the potential to be a cost-effective solution [79], further work is required to ensure that the health benefits offset the initial increased costs of setting up and implementing such interventions. There is 1 relevant comparison study that evaluated the cost-effectiveness of a smartphone app in people with heart failure [80]. This number highlights the relative infancy of research in this area. The CardioManager app achieved greater savings in the management of heart failure with an ICER of €9000 (US $9595.8) per patient, equating to large reductions in costs associated with hospital admissions [80]. Future research in this area would benefit from reporting the economic evaluations of smartphone apps to determine their cost-effectiveness and improve research translation and real-world implementation.

This study has several strengths being one of the first published studies to explore the impact of a sedentary behavior change smartphone app on hospital admissions and its cost-effectiveness. Device-measured sedentary behavior was analyzed, personalization and clinical guidelines were championed, and the study included both public and private hospitals. This is critical for improving the implementation of research apps in real-world settings.

Despite these strengths, there were several limitations. First, this trial was interrupted during the COVID-19 pandemic, impacting the recruitment and follow-up of participants and ultimately resulting in incomplete data. The number of participants who completed outcomes remotely have been reported, and sensitivity analyses were completed in line with guidelines for reporting trials affected by the pandemic [34]. No participants reported dropping out owing to the pandemic; however, there were difficulties recruiting, and hence, the target sample was not reached. Despite this limitation, the final dropout rate was lower than expected; therefore, the required sample size was maintained. Second, the majority of participants were male, tertiary educated, and working, and those excluded were significantly older, had more severe diagnoses, and were predominantly from the public health system, limiting generalizability. There were multiple assessors that may have affected the measurement error of outcomes such as waist circumference. Measuring engagement with the Vire app and ToDo-CR program was also limited. Further investigation into app engagement using back-end data is needed to better understand the relationship between app use and changes in sedentary behavior. Although half of the participants had experience using physical activity trackers, there is evidence that inexperienced users may not use all app features and therefore may not obtain the full anticipated benefits of behavior change smartphone apps [81]. In addition, although the use of ActiGraph accelerometers provides an objective measure of sedentary behavior, they are less sensitive to postural changes than devices such as ActiPal [13]. Future research would benefit from the use of inclinometers to detect changes from sitting to lying to standing and from combining this information with heart rate monitors and GPS technology to better inform the holistic picture of individual behavior surrounding sedentariness [82]. Finally, bringing the last value forward for missing data and the lack of participant blinding inherently introduced bias to the methods used.

It does not appear that the Vire app and ToDo-CR program targeting sedentary behavior is an outcome-effective or cost-effective solution to reduce all-cause hospital admissions or ED presentations in cardiac rehabilitation participants. Although those using the Vire app and the ToDo-CR program had more hospital admissions, these admissions were less costly. Further research is warranted to improve engagement and implementation with age appearing to be less of an impacting factor and prior experience with apps correlating more with engagement. This type of intervention may work as a better resource for those not already attending cardiac rehabilitation, who are at greater risk of hospitalizations, to influence sedentary behavior and potentially reduce costs associated with cardiac-related hospital admissions.