Meta-analyses have shown that telemonitoring for patients with heart failure (HF) can improve patients’ health outcomes and reduce health care utilization [1-4]. However, when one considers that HF directly impacts 1 million Canadians [5] and that in 2013 only 5000 patients across all disease types were enrolled in a telemonitoring program [6], it is clear that the diffusion of telemonitoring is lagging. This can partly be explained by higher than anticipated costs of implementing these programs and the lack of user input in the conception of such interventions [7]. In addition, although meta-analyses generally conclude positive outcomes, inconsistencies at the individual study level, particularly with respect to the economic impact, are difficult for stakeholders to ignore [8]. We have proposed in a previous work that this heterogeneity is caused by variances in the characteristics of (1) patients enrolled, (2) the intervention (eg, telemonitoring system used, clinician involvement, and supporting health services), and (3) fidelity with which the intervention is administered over time [8].

Although differences in the way telemonitoring interventions are delivered can lead to contradictory evidence, understanding these differences and how they might influence outcomes could hold one of the keys to scalability. This is because theories of diffusion of innovation have suggested that to be sustained and scaled, interventions must be able to adapt if they are to be embedded within local conditions [9,10]. This notion of adaptability is a prominent theme in studies of delivering digital health interventions at scale, which reinforce the view that a one-size-fits-all approach does not work [11]. The challenge is determining how to undertake necessary adaptations without compromising program fidelity [12].

A useful analogy used by theorists to discuss the notion of adaptability is the idea that health interventions have a hard core and a soft periphery [13-15]. The hard core represents the essence of an intervention, in other words, the central mechanism(s) for producing desired health outcomes in the intervention’s theory of change [12]. When considering telemonitoring, the hard core can be conceptualized as an intervention that leverages technology to enable the collection and transmission of patient biometric data to be viewed and acted upon by a clinician at a distant location [1].

In contrast, the adaptable soft periphery represents the different ways this intervention can be delivered in practice. Adaptability of this soft periphery to local contexts allows innovations to spread without negatively impacting the intervention’s ability to yield desirable outcomes [15], thus maintaining intervention fidelity. As it relates to telemonitoring, elements of the soft periphery may include differences in the hardware used, intensity of clinician monitoring, duration of a telemonitoring program, and format of training or technical support services. However, many of these program components are essential for a telemonitoring program to function. Therefore, delineating the line between the hard core and soft periphery of complex interventions such as telemonitoring programs is particularly difficult. Despite this challenge, implementation and scaling require a clear definition of a program’s core components to ensure that fidelity is maintained when adaptations are undertaken to ensure implementation and scaling success [10,12].

In fall 2016, an HF telemonitoring program was made available to patients of a heart function clinic at an urban hospital in Toronto, Canada. A previous study concluded the initial implementation to be a success based on the degree of integration within the clinic, number of patients enrolled, and fidelity of program delivery as part of the standard of care [16]. However, this study also identified important barriers related to the cost of the equipment and supporting human resources, which could hinder the sustainability and scalability of this program [16]. The objectives of this paper were to (1) understand which components of the telemonitoring program could be modified to reduce costs and adapted to other local contexts while maintaining program fidelity and (2) describe the changes made to the program to enable its sustainability within the initial implementation site and scalability to other health organizations.

This qualitative study was designed to elicit insights from end users to better understand the hard core and soft periphery of an existing telemonitoring program. These insights would inform adaptations required to reduce costs of delivering the intervention. Semistructured interviews were conducted within the context of a larger quality improvement program evaluation [17], which was approved by the University Health Network (UHN) Research Ethics Board (16-5789).

Separate interview guides were developed for patients and clinicians to inform possible strategies for lowering costs and improving program efficiency by gaining a better understanding of the program’s soft periphery. Specifically, participants were asked to comment on the topics presented in Table 1. In their responses, participants were encouraged to consider HF telemonitoring in general, and not just the Medly program. The Medly software (with embedded rules-based algorithm) was developed around the program’s theory of change [19]. As such, it is considered part of the program’s hard core; thus, no probes related to this component were included.

Patients (n=23) were identified through purposeful sampling based on age, gender, and time since enrollment in the Medly program to ensure a variety of perspectives. This included patients who were interviewed immediately after enrollment and, thus, had no prior experience being monitored in the program. Interviews with patients were conducted until theme saturation was reached (no new themes or perspectives were found in the data) [20]. This was achieved by setting an a priori target of 20 patient interviews. Three additional interviews were conducted, which yielded no new findings, thus confirming theme saturation. All clinicians actively monitoring patients using the Medly system at the time of the interviews (n=4) were invited to participate. In addition, 4 clinicians within the UHN Heart Function Clinic who had not yet begun using the system were also interviewed to obtain the views of nonusers. Written informed consent was obtained from all participants.

Patients had the option of being interviewed in a private room at the UHN Heart Function Clinic during one of their regularly scheduled visits or over the phone. Clinicians were interviewed in their private offices. Interviews lasted 15 to 60 min and were recorded and transcribed verbatim. Transcripts were analyzed using conventional content analysis [21]; PW and KG each independently coded the transcripts and then met to discuss the results and discrepancies with themes. Once a finalized coding scheme was agreed upon, it was used to code the transcripts before a final analysis of themes. NVivo version 11 (QSR International, Doncaster, Victoria, Australia) was used to organize the data analysis.

On the basis of the qualitative findings, PW and KG interpreted the degree to which each of the Medly program components explored in the interviews could be adapted without impacting program fidelity. Ideas for redesign were discussed during biweekly operations meetings and prioritized for implementation according to their (1) potential to impact sustainability and scalability through cost reductions and optimization of clinic resources, (2) feasibility of implementing the change, and (3) perceived risks of negatively influencing program fidelity (and ultimately effectiveness).

The demographic characteristics for the patients interviewed were representative of the patients enrolled in the Medly program. The average age was 60 years (SD 15) and 74% were male (17/23); see additional demographic characteristics and clinical variables (NYHA and left ventricular ejection fraction) in Table 2.

At the time of the interviews, 13% (3/23) of patients had been enrolled for 12 months; 48% (11/23) had been enrolled for 6 months; and 9% (2/23) had been enrolled for 1 month. In addition, 22% (5/23) were interviewed immediately after receiving training on their first day and, thus, had no prior experience being monitored in the Medly program. Of the 8 clinicians who participated, 2 NPs and 2 cardiologists had 9 to 12 months of experience monitoring patients with the Medly system. The remaining 4 cardiologists had no first-hand experience monitoring patients in the Medly program.

The following is a discussion of participants’ perceptions of opportunities for adapting existing program components aimed at reducing costs and optimizing clinic resources. Themes included were as follows: (1) Bring Your Own Device (BYOD), (2) technical support, (3) clinician role, (4) duration of patient enrollment, and (5) intensity of monitoring.

Qualitative results related to opportunities to modify components of the Medly program were interpreted and classified according to the degree to which they could be adapted while maintaining program fidelity. As shown in Figure 2, the format of technical support and the peripheral equipment used were considered highly adaptable and, thus, clearly part of the Medly program’s soft periphery. The participation of a clinician (role and intensity of monitoring) and the monitoring of patients over time are central components of any telemonitoring program theory of change, indicating some overlap with the intervention’s hard core. However, the interviews suggest some degree of adaptability depending on contextual factors, which explains why intensity of monitoring, clinician role, and duration of enrollment were classified as moderately adaptable and part of a fuzzy boundary between the hard core and soft periphery. These findings informed the decisions to adapt the Medly program as described in Table 3.

This qualitative study is the first to describe adaptations to an existing HF telemonitoring program aimed at enabling its sustainability and scalability. The redesign was informed by interviews with clinicians and patients to identify which program components could be adapted while maintaining program fidelity. User perceptions helped identify that the type of peripheral devices used and the format of technical support were highly adaptable, making them ideal targets for cost reduction measures. This led to the decision to move forward with a hybrid BYOD model and lower-touch technical support services, which would substantially reduce the cost burden to the clinic for delivering the program. In addition, findings related to the clinician role confirmed that frontline alert management should be done by someone within the patient’s immediate circle of care rather than being outsourced to an offsite telehealth clinician. This informed a more cost-effective model in using an RN to replace the existing NPs as the frontline manager of telemonitoring alerts and to absorb the technical support functions previously performed by the THA. The notion of an RN playing both the central clinical and operational roles within telemonitoring services is supported in the literature [22,23]. In this case, hiring of an RN made sense as a resource optimization measure because of the existing program structure, which involves escalating alerts to an MRP. In sites where a physician is not as readily available, a professional with an ability to make medication changes (eg, NP) might be more appropriate to lead a telemonitoring service.

Because no generalizable dose with respect to duration of enrollment and intensity of monitoring could be established, no changes were made at the existing program site. However, the moderately adaptable nature of these components may reveal opportunities for scaling as they might be tailored to allow for program integration within sites with different patient populations, resources, and objectives. For example, although rapid feedback from a clinician is often described as the most important component of a telemonitoring service [24], a clinical site serving patients with a lower disease severity may not require 7 days per week monitoring. In addition, a site with resource constraints may wish to prioritize improving patient self-care, in which case, a 3- to 6-month duration might represent an optimized duration of enrollment. Alternatively, sites with available resources and different organizational values may wish to prioritize the patient’s experience in addition to improving self-care and decide to monitor patients indefinitely. Finally, although the clinicians in this study felt comfortable receiving alerts during off hours, the medicolegal implications of continuous monitoring must be considered on a site-by-site basis. For example, it is possible that 7 days per week monitoring is deemed important for a specific patient population but that receiving alerts during off hours represents a medicolegal concern that cannot be addressed through the hiring of additional staff or on-call personnel. Such a situation may leave a site no choice but to offer a weekday-only telemonitoring program, with the understanding that the impacts of the program may be suboptimal.

Several studies have explored the barriers to and facilitators of implementing telehealth systems [25,26] but few have described the process of adapting an existing program to ensure its sustainability and scalability. One multiple case study by Taylor et al describes a participatory approach to implementing solutions for expanding a telehealth program [27], but the description of these activities remained high level without concrete examples, leading to limited transferability of results.

Many authors cite the ubiquity of smartphones as an opportunity for delivering telemonitoring services at a lower cost [28,29]. However, most studies of mobile phone–based telemonitoring have provided patients with this mobile equipment [30], and little is known about clinicians’ and patients’ perceptions of a BYOD model. From a usability perspective, there is a clear preference among patients, both in the literature [24,31,32] and in this study, for using Bluetooth-enabled peripheral devices. However, what appears most important is that patients can access telemonitoring services using devices they are most familiar with (ie, personal smartphones and tablets) [33-35] and that the perceived advantages of a telemonitoring program are greater than any usability inconveniences caused by manually entering biometric data [36-38]. To our knowledge, ours is the first study to confirm that BYOD is perceived by both clinician and patient users as a viable option for delivering telemonitoring services with a caveat that considerations are required to ensure universal accessibility.

The finding that clinicians believe patients could exit a telemonitoring program after they have stabilized or gained self-care skills is supported by other studies [39]. We found a similar perspective among clinicians in this study, but we also found that many patients grow accustomed to being remotely monitored and would like to continue over a longer term. Until now, considerations about the duration of telemonitoring interventions have primarily been driven by costs. However, this perspective ignores the natural history of HF, whereby although patients may stabilize for a period, they will rarely improve [40]. Therefore, as opportunities are leveraged to deliver telemonitoring interventions at lower costs (including clinicians’ time through the development of more sophisticated decision-support capabilities), it is conceivable that the costs of delivering certain telemonitoring programs will become sufficiently low so that it removes the need to ration their use. Thus, future work should seek to answer whether it is better (from the patient, clinician, and health system perspectives) to (1) remove patients from a low-cost telemonitoring program when they have stabilized only to reinstate them in the program when their condition has worsened or (2) leave them enrolled in the program indefinitely.

First, although participants were asked to consider their responses with respect to telemonitoring in general, it is likely that their responses were influenced by their experiences with the Medly program. In stating this limitation, we emphasize that our intent was to describe adaptations to a specific telemonitoring program rather than to provide a detailed blueprint for implementing all HF telemonitoring interventions in any given clinical context. We argue that the context-dependent nature of implementing complex interventions makes the creation of such a blueprint impossible. Rather, we have sought to provide foundational considerations for developers of telemonitoring programs and for implementation scientists who wish to sustain and scale existing telemonitoring programs to other clinical sites and health care organizations. Second, the pragmatic nature of this study meant that patients can be enrolled in the Medly program without consenting to participate in the evaluation activities, making them ineligible to participate in the interviews. Our inability to purposely sample these patients may have led to selection bias. Third, the interview guides were developed to probe the opinions of users on specific program components. We recognize that our approach for compartmentalizing and defining the various components of this complex intervention was subjective and context specific; this should be considered when determining the transferability of results to alternative settings. Finally, although the resulting user-guided adaptations are expected to maintain the fidelity of the intervention, the true impact of these changes was not empirically tested in this study. This important question will be evaluated as part of a subsequent publication on the overall impacts of the Medly program as well as patient adoption and adherence to the intervention.

Theories of diffusion of innovation suggest that one of the keys to scaling health interventions lies in adapting elements of its delivery to better fit the implementation context. However, this is only true if fidelity of the intervention can be maintained and its potential effectiveness is not compromised. This concept has informed the implementation of cost reduction measures of an existing HF telemonitoring program to ensure its sustainability. Our findings suggest that the peripheral devices used in telemonitoring programs and the format of technical support are highly adaptable, making them ideal targets for cost reduction measures. Duration of enrollment and intensity of monitoring are inextricable components of a telemonitoring intervention, but the dose of these components required to yield expected outcomes is highly context dependent. Our efforts provide a user-centered example of how necessary actions can be taken to improve the sustainability and scalability of telemonitoring programs.