Heart failure affects approximately 40 million people worldwide [1]. A diagnosis of heart failure portends a poor prognosis, with a 12-month mortality rate of 17% for patients who are hospitalized and 7% for patients who are stable or ambulatory [2]. Hospitalization is associated with a 3-fold increased risk of death [3,4] and is preventable with good quality self-management [4-6], including symptom monitoring and taking prompt action when deterioration begins [4,7]. However, there are several barriers to achieving good quality self-management, such as lack of knowledge, symptom recognition, motivation, and confidence [8]. Addressing these can improve outcomes; yet, delivering such models for support at scale is challenging.

Mobile health (mHealth)—medical and public health practice supported by mobile devices, such as mobile phones, patient monitoring devices, personal digital assistants, and other wireless devices [9]—has excellent potential for cardiovascular disease prevention [10-14]. In particular, app interventions seem promising as they can automate the self-monitoring of physiological data, facilitate symptom and medication tracking, and provide reminders and personalized feedback to promote patient engagement [15-17]. To date, no systematic reviews have focused exclusively on mobile apps to support self-management of heart failure. Previous mHealth systematic reviews on heart failure have mostly reported remote monitoring interventions using older technologies such as phone calls and interactive voice response to transmit data to health care providers, rarely focusing on supporting patient self-management [18-27]. A total of 3 nonsystematic reviews evaluated the content and quality of existing commercial heart failure apps and mHealth interventions without assessing their impact or patient perspectives [28-30].

This systematic review aims to examine the role of mobile apps in heart failure self-management, specifically, their impact on improving (1) clinical outcomes, (2) patient-reported measures, and (3) self-management knowledge and behaviors and in addition, examine the acceptability and feasibility of these interventions, as well as patient perspectives, needs, and preferences for specific app features.

A systematic search of the literature was performed in October 2019 and updated in December 2021 on PubMed, Embase, CINAHL, and PsycINFO, using several search terms such as mobile apps, heart failure, and self-management (Multimedia Appendix 1). The reference lists of relevant articles and gray literature such as dissertations, theses, and conference proceedings were also screened to ensure that all eligible studies were captured. The search was limited from 2008 onward as app stores were launched in that year [31]. No language limits were applied.

Studies were included if they (1) focused on adult patients with heart failure, (2) involved an intervention comprising a mobile app to support heart failure self-management (ie, provision of education and support to increase patients’ skills and confidence in managing their disease [32])—the mobile app could be a single component in the intervention or be combined with other intervention components (eg, wireless devices for remote monitoring)—(3) included any type or no comparison (eg, qualitative studies), (4) reported impact on any health outcome or patient-reported measure (eg, self-management and medication adherence) or focused on patients’ perspectives, and (5) were a primary research study involving the use or testing of the mobile app intervention. Studies were excluded if they (1) did not involve the use of the app by patients with heart failure and (2) assessed interventions without a clear component of heart failure self-management (eg, patients using the app only to input data to be analyzed by health care professionals).

The screening form was piloted by 2 investigators before beginning the screening process. The 2 investigators independently screened studies based on the information in their titles and abstracts and then performed the full-paper screening. Disagreements were resolved through discussion between the reviewers or by a third reviewer. Cohen κ statistic was used to measure intercoder agreement in the initial and full-text screening [33].

One of the reviewers extracted the following information from the included studies: author, year of publication, country, study design, sample size, population characteristics, study duration or intervention use time, intervention characteristics (eg, technology components and others, mobile app features, and presence or absence of personalization), comparison, outcomes, and main results. The 2 investigators reviewed the data extraction form for consistency. The coding of behavior change techniques (BCTs) according to the BCT taxonomy [34] was conducted by 1 researcher and reviewed by another. Studies’ quality and risk of bias were appraised by 2 researchers using Cochrane’s risk of bias tool [35] for randomized controlled trials (RCTs) and the Risk Of Bias In Nonrandomized Studies of Interventions [36] tool for other experimental studies. Disagreements were resolved by a third reviewer. We performed a narrative synthesis of the studies. The PRISMA (Preferred Reporting Item for Systematic Reviews and Meta-Analyses) 2020 statement was followed (Multimedia Appendix 2) [37], and the protocol was registered on PROSPERO (International Prospective Register of Systematic Reviews; CRD42020158041).

The database search retrieved 1689 citations, from which 458 (27.1%) duplicates were removed (Figure 1). After title and abstract screening of the 1689 articles, 1189 (70.4%) were excluded. Full-text screening was conducted for 42 articles, and a further 26 (62%) papers were excluded (see Multimedia Appendix 3 for reasons for exclusion). A total of 12 additional papers were identified—1 (8%) from the reference list of the included studies and 11 (92%) from database alerts and search updates—leading to the inclusion of 28 articles [38-65] for final analysis (corresponding to 27 studies, as 1 study was published in 2 different articles [38,65]). The Cohen κ statistic was 0.81 (excellent agreement) for the title and abstract screening and 0.53 (fair agreement) for the full-text screening before the consensus agreement was reached [66].

All 28 included articles [38-65] were published from 2012 onward and covered 27 studies and 23 interventions (n=4, 17% interventions were evaluated in ≥1 paper, using different study designs [39,44,47,52,56,59,63-65]). Of the 28 studies, there were 18 (64%) experimental studies [38-55], (n=10, 56% RCTs [38-47] and n=8, 44% quasi-experimental [48-55]; n=7, 39% with a qualitative component [38,44,46,49,51,52,54]; Table 1), 9 (32%) qualitative-only studies, 5 (18%) that included interviews [56-60], and 4 (14%) that involved a survey with open-ended questions (Table 2) [57-64]. Most studies were conducted in the United States (15/28, 54%) [39,40,44,45,48,50,52,54,56,58,60-64] and Canada (4/28, 14%) [47,51,55,59], and most were single-center, except for a few (5/28, 18%) [38,42,43,55,62]. There were 1397 participants (n=8-232 in experimental studies and 5-37 in qualitative studies), mean age was 63.4 years, 30% were women, 68% were White (from 15/28, 54% studies that reported on ethnicity), and the average education level was high (Multimedia Appendix 4 [38-65]). The study duration in the experimental studies ranged from 2 weeks to 12 months (average of 3.2 months). The 10 RCTs had a moderate risk of bias [35]; the quasi-experimental studies were of lower quality (Multimedia Appendix 5 [36,38-55]) [35].

Across the 23 apps, the app was provided via a smartphone in 17 (74%) [38-54,57,59,63,64] and via a tablet in 6 (26%) interventions [42,55,58,60-62]. In addition to the app, 35% (8/23) of interventions included telemonitoring (ie, remote monitoring), with transfer of data to health care providers [41,42,46,47,49-51,54], and 65% (15/23) were solely focused on self-management support [38-40,43-45,48,52,53,55,57,58,60-62]. Approximately 9% (2/23) of apps provided patient access to electronic medical records [42,49], and 22% (5/23) of apps allowed direct clinician communication [39,40,42,49,53]. Approximately 48% (11/23) of apps involved patient or clinician co-design [38,40,42-44,47,51,55,57,58,60]. For 39% (9/23) of apps [38,40-42,48,51,53,57,58], the authors reported the use of personalization, mostly in the form of feedback to self-monitored measures (Multimedia Appendix 6 [38-65]).

The most frequent app features were weight monitoring (19/23, 83%), [38-42,44,46,47,49-54,57-60,62-64], symptom monitoring (18/23, 78%) [40-42,44,46-55,57-60,62-64], and vital signs monitoring (blood pressure and heart rate: 15/23, 65%; Figure 2; Multimedia Appendix 7 [38-55,57-65]) [38,41,42,44,46,47,49-54,57-60,63,64]. Automated monitoring through external wireless devices (eg, weight scale, blood pressure, and heart rate monitor) was present in 43% (10/23) of apps [38-40,42,44,47,49-51,54,59,64]. Of these 10 apps, 6 (60%) were part of a telemonitoring system (ie, the apps were connected to a health care service or clinical provider) [42,47,49-51,54,59]. None of the interventions included implantable cardiac devices. Most apps recommended daily monitoring of symptoms and vital signs, and reminders for monitoring were mentioned in 52% (12/23) of interventions [38-42,44,47,51,53-55,57]. Few studies detailed the format or specifics of symptom monitoring except for 22% (5/23) of interventions [41,46,54,55,62], which allowed for the recording of the presence or absence of specific symptoms, with 20% (1/5) of them based on a validated questionnaire [54] and 60% (3/5) of them also providing symptom severity scales [41,54,62].

The most common BCTs presented in the studies were instructions on how to perform the behavior in 91% (21/23) of interventions [38-55,57-59,61,63,64], followed by self-monitoring of outcomes of behavior in 83% (19/23) [38-42,44,46-54,57-60,62-64], behavioral practice or rehearsal in 78% (18/23) [38-42,44,46-51,53-55,57,59,60,62,64], prompts or cues [38-42,44,45,47-51,53-55,57,59,60,64] and feedback on outcomes of behavior [38-42,44,46-51,53,54,57-59,62,64] in 74% (17/23) of interventions each. Feedback was active in 48% (11/23) of apps (ie, the app gave specific instructions to the patient in response to the individual information inputted by them) [40-42,44,46-48,51,53] and passive in 65% (15/23) of apps (ie, display of measurements in graphs) [38-40,42,44,47-51,54,57,58,62] (Multimedia Appendix 8 [38-55,57-65]).

The 10 included RCTs were small, often underpowered, with main outcomes self-reported, and the results were inconsistent. Approximately 20% (2/10) of RCTs found significant improvements in their primary outcomes: heart failure knowledge [43] and self-care [41]. One of the RCTs [47] reported several primary outcomes, showing improvements in self-care and quality of life. Approximately 40% (4/10) of RCTs did not show significant improvements in their primary outcomes (quality of life [38,39], self-care [38], achieving optimal medical therapy [42], medication adherence [45], and heart failure–related hospital days [46]). Approximately 20% (2/10) of RCTs indicated that their main aim was to assess feasibility [40,44].

Key clinical outcomes in heart failure were seldom reported (ie, mortality [46,47], emergency department visits [39,43,46,47], and hospital readmissions, [39,43,47,50,53]), with only 4% (1/28) of RCTs [39] and 7% (2/28) of quasi-experimental studies [50,53] showing a reduction in readmissions. Approximately 20% (2/10) of RCTs reported higher health care services use in the intervention groups than the control groups, including a higher number of unplanned clinic visits [46,47] and higher use of nurse resources (time and calls) and medication optimization [46].

Other significant improvements were inconsistently reported across experimental studies: heart failure–specific knowledge [40,44,55], self-care [38,52], hospital readmissions [50,53], depression and anxiety measures [38], quality of life [40], systolic blood pressure [42], diet [49], self-monitoring (blood pressure and weight) [49], medication adherence [49], 6-minute walking test [53], and pitting edema [53]. Engagement with the mobile app was reported in 11% (3/28) of studies, 67% (2/3) indicating that less than half of the participants accessed the app daily as recommended by the investigators [40,44] and another showing that 60% of participants used the app more than once a week, as recommended [49].

In this first systematic review targeting exclusively mobile apps for heart failure self-management, we identified 23 unique apps evaluated in quantitative and qualitative designs, with 8 (35%) being part of telemonitoring systems and connected to health care services. Common features of apps were weight, symptom, and vital signs monitoring and provision of education, medication reminders, and graphical visualization of data. Overall, few had robust efficacy evaluation frameworks—only 10 RCTs involving 717 participants, with ≤6 months of follow-up, substantial heterogeneity in interventions and outcomes, and hence little quantitative evidence to indicate efficacy. Few studies involved patients and clinicians in the design of apps, and few quantified app engagement metrics such as frequency of use during studies. Qualitative studies identified the automation of self-monitoring tasks and feedback, personalization of content and format, communication with clinicians, and data sharing and integration capabilities as key enablers.

Similar to previous systematic reviews of other digital technologies in heart failure (focused on nonapp mobile technologies, such as SMS text messaging, personal digital assistants, interactive voice response, and phone calls), our findings were mixed, with high heterogeneity and lack of detailed reporting of intervention characteristics [18-27] likely because of poor evaluation frameworks. In these reviews, the interventions did not commonly offer self-management support (eg, education and feedback), merely involving remote monitoring with regular digital transmission of physiological and other disease-related data from the patient’s home to a health care center. In addition, previous nonsystematic reviews seemingly with a focus on apps for heart failure self-management either only assessed the content and quality of commercially available apps [28-30] or broadened their inclusion criteria, including studies where the intervention was some type of mHealth technology but not an app (eg, SMS text messaging) [30]. In contrast, our systematic review is the first to focus exclusively on mobile apps for heart failure self-management (with or without clinician involvement via telemonitoring).

Despite the focus on heart failure self-management, the studies included in this review varied considerably in the types of self-management support features available in the apps. Core components of heart failure self-management programs, as defined in existing guidelines [2,5,70], include education, symptom monitoring, medication support, and physical activity support. Nevertheless, only 26% (6/23) of apps provided all these features [44,52,54,55,57,58], with more apps including features less supported by evidence in regard to their benefits in heart failure [5], such as daily weight monitoring. As a road map for future studies in this area, we encourage researchers and developers to follow the best available evidence [2,5,70] when designing and evaluating heart failure apps for self-management, focusing on features that have been systematically associated with improved outcomes. In addition, better reporting of intervention features is crucial to avoid what has been named as the black box of home telemonitoring [20], where the specific effective components of these interventions remain unknown.

Key outcomes in heart failure were seldom assessed in the included studies, hampering a complete evaluation of the impact of heart failure self-management apps. Overall, 1 RCT [39] and 2 quasi-experimental studies showed a significant reduction in readmissions [50,53], corresponding to the evaluation of 1 self-management app with telemonitoring and 2 without telemonitoring. Furthermore, 30% (3/10) of RCTs evaluated health care system use [43,46,47], with 67% (2/3) of them finding a higher number of unplanned clinic visits and medication optimization for participants in telemonitoring programs [46], although without significant changes in mortality, emergency department visits, or hospitalization [46,47]. Higher health care use may reflect earlier actions in the face of signs of worsening heart failure and provide opportunities for medication optimization. Such results may help explain the positive outcomes of telemonitoring interventions [26]. Longer and adequately powered studies measuring key clinical outcomes are needed to fully assess whether the potential benefits of self-management apps outweigh the costs of increased health care use.

Self-reported measures were commonly assessed in experimental studies, including validated questionnaires to measure heart failure knowledge, self-care, and quality of life [67-69]. Heart failure knowledge was significantly improved in 14% (4/28) of studies, all of which involved apps without telemonitoring [40,43,44,55]. Self-care was improved in 14% (4/28) of studies [38,41,47,52], 50% (2/4) of which involved apps with telemonitoring [41,47], and quality of life improved in 7% (2/28) of studies [40,47], 50% (1/2) of which involved telemonitoring [40]. There has been increasing recognition of the importance of including patient-reported outcomes as end points when evaluating interventions, as well as the benefits of collecting them routinely to improve care [71-73]. Digital technologies such as mobile apps can facilitate the capture of patient-reported outcomes, such as symptom status and severity [71], which can then be used by clinicians to guide care. Nevertheless, only one of the apps used a validated questionnaire for symptom monitoring [54]. The potential of mobile apps to collect patient-reported outcomes should be further explored in future studies, given their ability to promote patient-centered care and improve the quality of care for patients.

Overall, the evidence on the use of mobile apps for heart failure self-management is still lagging behind the large body of work supporting mHealth for remote monitoring, where significant reductions in all-cause mortality have been reported [19-22,26,27]. In our review, all included studies focused on supporting heart failure self-management, with 44% (8/18) of experimental studies including a telemonitoring component with clinician involvement [46,47,49-51,54]. Unfortunately, the small number, size, and quality of these studies do not enable us to draw conclusions regarding potential differences in efficacy between these 2 different types of mobile app interventions for heart failure self-management—with or without telemonitoring. Given the demonstrated benefits of self-management interventions more broadly [74] and remote monitoring [18-27], future research should explore the possibility that their combination may result in synergistic effects and higher efficacy in improving heart failure outcomes.

Personalization was valued in the studies included in this review, particularly personally relevant feedback and tailoring of the intervention to different levels of education and digital literacy. These findings are similar to those involving apps for other chronic diseases, showing that enabling customization (eg, editing information and choosing which aspects to track) is among the most appealing features and may enhance the usability, motivation, and engagement with the apps [17,75,76]. Future studies may explore the delivery of core BCTs (self-monitoring, feedback, and instruction on how to perform the behavior) and provide other techniques in a personalized manner, according to patient preferences and self-reported information [77] or based on machine learning algorithms using patient data collected over time (eg, from smartphone sensors or wireless monitoring devices) [78,79].

Limited experience in using technology can be a barrier to using mobile apps and may affect the utility and perceived benefit of mobile apps, as shown by our findings. The lack of confidence in using technology and perceived capability to benefit from it, as well as the workload required to learn how to use an app, are particularly challenging among older patients [80,81]. A study conducted to understand the main facilitators of and barriers to the use of mobile technology among older adults found that the most often mentioned barrier was the lack of knowledge on how to use it, whereas having previous experience of use was a facilitator [82]. However, older patients are willing to learn how to use mHealth technology and feel it may help them improve and maintain self-care behaviors [82,83]. Given that a large population of patients experiencing heart failure involves older adults, future app development needs to take into account specific characteristics of this population to design apps with simple navigation and ease of use [81].

This study presents several strengths. The PRISMA protocol was systematically followed. The screening process was pilot tested before its start, and there was good agreement between the independent reviewers. We also included both experimental and qualitative studies, enabling a better understanding of the impact, acceptance, and user preferences regarding mobile apps for heart failure self-management.

Some limitations should be considered in the interpretation of our results. First, given the heterogeneity between interventions and the small number of RCTs, a meta-analysis was not conducted. Second, the heterogeneity of study designs, sample sizes, follow-ups, interventions, and outcome measures among the experimental studies did not allow for consistent conclusions on the effectiveness of mobile apps in heart failure. Third, some studies in this review included analysis of adherence, acceptability, or usability of their interventions; however, although favorable trends were reported, the different measures and definitions used hindered reliable conclusions. Fourth, the socioeconomic and clinical characteristics of participants were rarely reported in the included studies; however, when reported, they suggested a high educational level and mild to moderate disease severity, potentially limiting the generalizability of the findings. Finally, the nature of this kind of research hampers the proper elucidation of the sociotechnical aspects of the interventions, which should be further evaluated in future studies (eg, using realist review methods).

Despite growing interest in the use of mobile apps for heart failure self-management, critical gaps remain in their design and evaluation, with lack of patient and clinician involvement and lack of robust evaluation to determine the populations that may benefit the most. Given the importance of patient preference and engagement in the successful delivery of heart failure interventions [26,27], co-design processes involving clinicians and patients and process evaluations assessing engagement and acceptability of the interventions are likely to improve intervention quality and consistency. Future studies should follow existing evidence in designing apps with features most likely to improve key patient-reported and clinical outcomes, adhering to recommendations derived from this study (Textbox 1). In addition, they should explore the efficacy and cost-effectiveness of mobile apps for heart failure self-management with and without a telemonitoring component. It is possible that self-management interventions without telemonitoring may be sufficient to improve outcomes in the early stages of disease in patients with a low risk of premature morbidity and mortality.

Research is needed to better understand how these interventions can be implemented in the real world and integrated into existing models of care, such as collaborative care models involving shared care between heart failure nurses, general practitioners, and cardiologists [84-86]. Integrating these interventions into such services may increase their benefits and leverage partnerships between patients and clinicians, possibly leading to a more seamless implementation in practice. Perhaps a future model of care for heart failure patients can involve using mobile technology to improve patients’ confidence and ability to manage their condition with greater autonomy, coupled with telemonitoring with clinician support for higher-risk patients.

This systematic review showed that research on the use of apps in heart failure self-management is still at an early stage, with limited evidence supporting its efficacy. RCTs are needed to fully ascertain the impact of these interventions. Future research should encompass greater involvement of end users and comprehensively measure patient engagement with the intervention.