This review was registered in PROSPERO (International Prospective Register of Systematic Reviews) (CRD42019124254) before screening search results, and was conducted according to the PRISMA (Preferred Reporting Items for Systematic Reviews and Meta-Analyses) statement [52,53].

Electronic databases—MEDLINE, Embase, PsycINFO, and Global Health—were searched between January 1990 and November 9, 2018, for studies that combined three concepts: telehealth, CVD, and implementation science. The search strategy was created for MEDLINE and modified for the other databases (see Multimedia Appendix 1).

Eligible studies were those that evaluated the implementation of cardiac telerehabilitation in routine clinical practice or assessed implementation outcomes of interest in experimental studies, including randomized and nonrandomized designs, among adults (aged ≥18 years) with CHD (ie, MI, angina, and coronary revascularization).

Cardiac telerehabilitation interventions were defined as those with at least 50% of the program delivered via ICT, including any mobile phone (ie, feature phone or smartphone), web-based platforms, or wireless devices such as sensors. Implementation in routine clinical practice was defined as interventions delivered as part of existing CR services, without significant ongoing input from a research team.

Experimental studies were included as we anticipated few studies examining real-world implementation, and experimental studies provide the next best available evidence to advance the field; eligibility was not limited to randomized controlled trials (RCTs) to allow the inclusion of translational studies that used alternative study designs. In addition to other criteria, eligible experimental studies were those that assessed constructs defined in a taxonomy of key constructs related to the effective implementation of evidence-based interventions, including acceptability, adoption, appropriateness, feasibility, fidelity, implementation cost, penetration, and sustainability [54]. To acknowledge the importance of multiple stakeholder levels in successful implementation projects [55,56] and to meet the aims of this review, we assessed these constructs at the consumer (ie, participant), individual provider (ie, CR practitioner), and provider (ie, health care organization or institution) levels [54]. Constructs were defined as follows:

Feasibility and pilot studies were excluded from this review. To meet the aims of this review, it was important to include only interventions that had already undergone preliminary testing for feasibility and were considered by their respective authors as feasible for testing in the trial or delivery in practice. Conference abstracts, nonhuman studies, non-English-language papers, and grey literature were also excluded. Systematic reviews and study protocols were not eligible for inclusion; however, relevant systematic reviews were searched for eligible studies and cited where appropriate, and results articles were sought for relevant study protocols.

Search results were exported to a reference manager, EndNote X8 (Clarivate), for duplicate removal, then transferred to Rayyan (Qatar Computing Research Institute) for screening [57]. Records were assessed by NS, verified by JR and HK, and underwent full-text review if the title or abstract identified the specified population and intervention components.

Data describing eligibility, study design, participant and intervention characteristics, risk of bias, and outcomes of interest were extracted by NS using a standardized electronic form and verified by JR.

Risks of selection, performance, detection, attrition, reporting, and other biases in included experimental studies were assessed using the Cochrane risk-of-bias tool [58]. Risks of bias in nonrandomized cohort studies were assessed using the Joanna Briggs Institute Critical Appraisal Checklist for Cohort Studies [59]. Risks of bias were assessed by NS and verified by JR. When available, risk-of-bias assessments were augmented with study protocols and clinical trial registrations.

A narrative synthesis of the data was carried out in this review.

In total, 2044 unique study reports were screened. From these, 21 underwent full-text review; 16 reports describing 10 studies (2250 participants in total) met the eligibility criteria and were included in the narrative synthesis [41-45,60-64]. Study selection is summarized in Figure 1.

Included studies were published between 2013 and 2018; studies were conducted in developed countries, including New Zealand [42,43,45], Australia [41,44], Canada [61], the Netherlands [60], Poland [63], and the United States [62]. One multi-country study was carried out in Spain, Germany, and the United Kingdom [64].

No eligible studies were identified that evaluated the effectiveness of the implementation of cardiac telerehabilitation in routine clinical practice. Of the 10 included studies, 9 (90%) used randomized controlled experimental designs [41-45,60-62,64], while 1 (10%) used an uncontrolled pre-post intervention design [63].

The mean age of study participants ranged from 55 to 65 years, and most participants were male (72%-93%). All studies recruited participants via hospitals or community-based CR centers. Detailed characteristics of the included studies (ie, study design, treatments, and primary and implementation outcomes) are presented in Multimedia Appendix 2.

Out of 10 studies, 3 interventions (30%) were delivered using a mobile phone, smartphone, or web-based platform alone [41,61,62], while remaining interventions used combinations of web-based content, mobile phones or smartphones, and sensors. The most commonly targeted lifestyle risk factors were physical activity, diet, tobacco smoking, and medication adherence. Intervention duration ranged from 30 days [62] to 24 weeks [41-43] (see Multimedia Appendix 2).

Most interventions (6/10, 60%) comprised a messaging component (eg, SMS or push notifications) [41,43-45,61,62] to educate or motivate participants to improve self-management behaviors. Out of 10 interventions, 7 (70%) enabled communication between providers and participants via a web-based program, mobile phone or smartphone, and/or telephone [42,44,45,60,61,63,64]. Out of 10 interventions, 6 (60%) included exercise monitoring [42,44,45,60,63,64], including 2 (20%) that provided live guidance [64] or real-time monitoring and coaching during exercise [45]. Out of 10 studies, 5 (50%) delivered telerehabilitation in combination with usual care (ie, center- or community-based CR) [41-43,45,60], 4 (40%) delivered cardiac telerehabilitation alone [44,61,62,64], and 1 (10%) delivered a hybrid intervention comprising center-based and telerehabilitation components [63] (see Multimedia Appendix 2).

The quality of the included studies in the review varied (see Multimedia Appendix 2). Risk of bias was judged to be low in 6 out of 10 (60%) experimental studies [41-43,45,60,61] and high in 3 (30%) studies [44,62,64]. High risk of bias was judged due to incomplete outcome data and lack of blinding of participants and outcomes.

The single nonrandomized study (1/10, 10%) had a high risk of bias due to lack of a control group, not identifying confounding factors, and inadequate reporting of follow-up time [63].

The primary finding of our review is that, despite encouraging evidence for effectiveness [33-37], there is a lack of evidence evaluating the translation of cardiac telerehabilitation interventions from research into routine clinical practice. It is unclear whether this suggests that evidence-based interventions have yet to be implemented in clinical practice, have been implemented without evaluation, or have been implemented and evaluated but not yet published in the scientific literature.

The next best available data comes from a small number of experimental studies that have assessed key constructs related to the effective implementation of evidence-based interventions. Almost all included studies reported factors related to intervention appropriateness, acceptability, and fidelity; however, adoption, fidelity, and cost have received less attention, and intervention penetration and sustainability had yet to be evaluated.

Moreover, while consumers (ie, CR participants), individual providers (ie, CR practitioners), and organizations (ie, health care services) all play critical roles in achieving successful implementation outcomes [54,72], the cardiac telerehabilitation literature we reviewed has focused only on the consumer perspective. This may reflect the lack of research conducted in routine clinical practice, as individual and organizational providers may have little involvement in the delivery or management of telerehabilitation interventions during experimental trials.

Cardiac telerehabilitation was generally well accepted among the majority of participants, even across a broad range of different interventions. While the small number of studies in our review makes it difficult to determine which interventions may be most acceptable to participants, intervention features that enable participants to communicate with practitioners and receive tailored or individualized support appear to promote high rates of acceptance. Unfortunately, acceptability has not yet been evaluated from a delivery perspective, so the perceptions of rehabilitation providers (ie, individual practitioners and organizations) remain unknown. Cardiac telerehabilitation was considered an appropriate delivery model by many participants, particularly those who value convenient, flexible, and accessible intervention support. Moreover, many participants reported that cardiac telerehabilitation was useful for improving their self-management of lifestyle behaviors and CVD risk factors. At the participant level, acceptability and appropriateness also appear to be moderated by intervention feasibility. Interventions that were simple to access, easy to use, reliable, and delivered through ubiquitous mobile, smartphone, and/or web technologies appeared to have higher acceptability and appropriateness.

As our review findings are drawn from experimental studies, it is unclear if they would generalize to the delivery of cardiac telerehabilitation in routine clinical practice. In particular, the predominance of randomized treatment allocation in included studies differs markedly from a recommendation that CR participants should be offered a choice of alternative CR delivery models that best fit their needs and preferences [73]. Studies that include preference-based treatment allocation are needed to mimic this key element of routine clinical practice or, better yet, evaluation studies should be conducted in parallel with the translation of cardiac telerehabilitation into routine clinical practice. The single non-RCT included in the review reported high acceptance and usability among participants who preferred telerehabilitation [63].

Reporting of participant-level feasibility was mixed, which may reflect the difficulty of documenting both the incidence and impact of usability and technical challenges. However, collectively the evidence suggests telerehabilitation is feasible for most participants. Feasibility among individual and organizational providers is also critical for successful implementation [74,75] but, similar to other implementation constructs, was not evaluated in the included studies.

Fidelity is one of the important implementation outcomes, as it contributes to intervention quality [54]. While 7 of 10 (70%) included studies reported high intervention uptake, some outcome measures of adherence may be confounded by self-reporting bias. Reassuringly, 3 (30%) studies indicate that adherence to cardiac telerehabilitation can be at least as high as center-based programs [44,45,60]. A key gap in the literature we reviewed is the lack of information about the fidelity of intervention delivery from the perspectives of individual and organizational providers, which is critical to understand when implementing new interventions to maintain intervention quality [76]. This may reflect known challenges in measuring implementation constructs and a lack of available validated tools [54]; however, we note that the delivery fidelity of interventions that require little, if any, provider input (eg, SMS) may be sufficiently assessed via software analytics.

There was little specific evidence that intervention delivery deviated from study protocols in the included studies, but it is unclear whether this indicates high fidelity intervention delivery or a lack of documentation to support such a conclusion. While future translational research could comprehensively evaluate the fidelity of intervention delivery, comparisons with preceding experimental research may be confounded by assumptions about equivalence between experimental and translational research contexts.

Intervention cost is a key contributor to low uptake of interventions by health care providers [77], and comparison of costs and cost-effectiveness is crucial for making evidence-based decisions about implementing and scaling new interventions [76]. Relatively few included studies reported economic analyses, and telerehabilitation costs varied markedly across different intervention designs. Our review indicates that telerehabilitation can reduce CR delivery costs and be cost-effective, but it is unclear if cost-effectiveness varies between different types of interventions. For example, interventions that require significant practitioner input, such as real-time remote exercise monitoring and coaching [45], may be substantially cheaper to deliver than center-based programs but more expensive than semiautomated interventions, such as SMS [43]. Whether or not an intervention represents good value depends on health effects and costs, as well as intervention and health care objectives. Therefore, examining the cost and health effects of telerehabilitation interventions in routine clinical practice is essential to provide more valuable information for implementation and scale-up.

The remaining implementation constructs of interest in this review—penetration and sustainability—were not assessed in any included studies. This was not surprising, as penetration and sustainability are more relevant to the mid- and later stages of implementation [54]. Short-duration experimental trials limit penetration to those who are willing to volunteer for research and preclude assessment of longer-term intervention sustainability. As a result, we lack valuable information on factors that could influence the sustainability of the interventions and how this might impact potential scale-up.

Evidence suggests the spread, scale-up, and sustainability of health care innovations are influenced by a broad range of factors related to the people who receive (ie, participants) and deliver (ie, health care practitioners and providers) the innovations as well as by numerous organizational and societal factors [55,56,74]. The experimental studies in our review focused on a relatively narrow range of outcomes, omitted the individual and organizational provider stakeholder levels, and could not replicate the complexity of implementation in routine clinical practice. Robust experimental evaluation of effectiveness and safety is critical before real-world implementation [78]. However, it is now critical to evaluate the implementation of proven interventions in routine clinical practice, preferably as a complementary adjunct to existing center-based programs, to incorporate the critical element of consumer choice that we and others have advocated [45,73]. Such studies should target all key implementation constructs across all relevant stakeholder levels, be embedded fully in routine clinical practice, and be evaluated for a sufficient duration to enable comprehensive assessment of all factors that contribute to successful, scalable, and sustainable implementation [74].

While it was beyond the scope of this review, an understanding of the relative importance of different factors on the implementation of telerehabilitation interventions in clinical practice is also needed. Although effective implementation is understood to be an interactive combination of factors [74], our review highlights the variability in the assessment and reporting of implementation constructs. It is unknown if this variability was due to, for example, researchers’ perceived importance of specific factors when selecting the study outcomes, evidence for the differential impact of implementation factors on outcomes, or the feasibility of evaluating multiple factors within a trial design. Improving the translation of interventions into routine clinical practice requires a greater understanding of the roles of implementation factors and consistent measurement of their impact. Research that explores the relative importance of such factors would greatly advance our ability to effectively scale interventions.

A major strength of this review was the use of a robust systematic review methodology to understand a novel research area. Secondly, while broad variability of intervention designs across a small number of studies makes it difficult to determine how to optimize telerehabilitation for translation into clinical practice, it provides a broad overview of potential issues that could be associated with implementation and scale-up.

The findings of our review are primarily limited by a lack of studies that have evaluated cardiac telerehabilitation interventions when implemented in routine clinical practice. The assessment of implementation-related outcomes during controlled experimental studies provides some insight, but marked differences within the context of real-world rehabilitation service delivery limit their generalizability. Additionally, while promising early evidence for effectiveness, safety, acceptability, and cost of cardiac telerehabilitation interventions [33-37] suggest they could play an important role in increasing overall participation in CR, it remains unclear whether positive trial outcomes will be retained following translation into clinical practice [76]. Our review includes a small number of studies with relatively small sample sizes and homogenous cohorts. This may limit generalizability to population subgroups who are typically underserved by CR, including older adults, women, people living in regional or rural areas, and people with diverse non-English-speaking cultural backgrounds [41-45,60,64]. Finally, there was a lack of studies from developing countries where telerehabilitation could have an even greater impact due to a low provision of traditional center-based CR [79-81].

Cardiac telerehabilitation interventions appear to be acceptable and appropriate for many participants in experimental trials and may be a cost-effective way to increase the reach and utilization of CR. However, explicit implementation studies are urgently needed to inform best-practice translation into routine clinical practice. When possible, such studies should implement telerehabilitation in parallel with existing center-based programs so consumers can autonomously match program delivery models to their individual needs and preferences.