1 Introduction
Rehabilitation is vital to long-term recovery following an acute cardiac incident and is a key component in the management of cardiovascular disease. Rehabilitation programs are a secondary prevention model, shown to reduce mortality and the risk of recurrent events and improve quality of life (QoL) [
70]. The primary objectives are to help patients recover from the initial incident, regain autonomy, understand how to manage their condition, and ultimately reduce the risk of further adverse outcomes [
38,
72]. There are variations in the provision and organization of cardiac rehabilitation (CR) services from country to country. Broadly however, CR programs follow a consistent structure. CR is typically supported by specialist medical teams and includes clinical assessment, medication review, risk factor modification (e.g., dietary changes), psychological support, and supervised exercise [
38]. In the UK, for example, standardized CR is structured across four phases, starting with (Phase 1) acute in-hospital rehabilitation; (Phase 2) outpatient clinic follow-up; (Phase 3) structured exercise and education programs, targeting lifestyle factors, regular exercise habits, and patients understanding of cardiac health; and finally (Phase 4) longer-term maintenance and self-management [
10]. CR programs in the United States more often describe three phases [
38]. However, these phases are broadly consistent with phases 1-3 of the UK model, with longer-term maintenance and self-management considered separate from the rehabilitation process. This paper focuses on the design of technology to support Phase 3 of both models. This is a critical point in rehabilitation where patients still receive regular, structured support from a medical team, but the emphasis is on supporting the transition to longer term self-management of their health [
63,
70].
Previous HCI research has described how people face emotional and physical challenges during the transition from hospitalization to self-management [
53]. Studies suggest that technology supported CR programs have the potential to support this transition [
22,
47,
61,
72], but a recent systematic review [
62] also notes significant limitations, arguing for greater use of participatory and iterative approaches in the design of future systems. Key barriers identified include a lack of trust, technology becoming a burden or not addressing needs of both patients and clinicians, and a lack of technical knowledge [
6,
22,
62,
63]. In a detailed study of CR, Tadas et al. [
63] provide recommendations on key patient needs during the rehabilitation process, emphasising patients’ desire for a normal life, the value of shared experiences or connectedness, the need for physical and emotional safety, and the value of both generalised and in-the-moment knowledge.
In addressing these recommendations we focus on the use of patient-generated data. Many papers have argued that self-tracking data generated through wearable devices has the potential to ‘bridge the gap’ between formal medical settings and day-to-day life, capturing and sharing patients’ activities and symptoms outside medical settings [
26]. However, here to a wide range of barriers are observed in the routine and effective use of patient-generated data in clinical contexts [
17,
69]. Chung et al. [
13] found that clinicians, nurses, and specialists found patient-generated data hard to use due to time constraints and a lack of standardized formats. Andersen et al. [
3] identify three key challenges in designing eHealth systems. Systems should (1) make sense to both patients and clinicians, (2) be actionable to both clinicians and patients; and (3) be feasible within the organizational and social context.
This paper directly addresses these challenges in the context of patient generated data. It presents the findings of an 18-week field study undertaken in the cardiology unit of a large hospital. The tracking and data sharing system used in the field study was co-designed with clinicians and built on detailed patient engagement research. The co-design process and key design decisions are described. The overall aim was to understand how tracking and sharing both activity (objective) and experience (subjective) data could support CR patients and clinicians during a Phase 3 CR program. The paper provides a detailed description of how patients, physiotherapists, and other clinicians engaged with and used patient-generated data.
The paper makes several contributions. Firstly, we show how a combination of subjective experience data and objective wearable data can help to make patient generated data more meaningful for clinicians. We also show how a structured approach to experience data, when combining with summarisation and filtering, helps clinicians to track trends and makes data actionable. In contrast, an open approach to experience data provided patients with a channel to share experiences with clinicians, when they might be hesitant to do so in-person. We also demonstrate the importance of introducing technology at an early stage in the rehabilitation process. This helps patients to develop self-efficacy, both in regard to technology and through greater understanding of their physical safe zones. Finally, the paper provides insight on data sharing between patients. While patients were open to sharing their own data with peers, they were hesitant about viewing data shared with them. We discuss the reasons for this. Based on our findings the paper provides design recommendations for the use of patient generated date to support rehabilitation and the transition to self-care.
3 Overview of Methodology and Research Questions
As noted in the Related Work section, Andersen et al. [
3] have identified three key, unmet challenges regarding the use of patient generated data in clinical settings. To be useful data needs to (1) make sense to both patients and clinicians, (2) be actionable to both clinicians and patients; and (3) be feasible within the organizational and social context. The studies outlined below seek to directly address these challenges. In doing so, we also seek to understand how the rehabilitation stage can help to alleviate the longer-term tensions in technology supported self-management identified by Nunes et al. [
44], in particular the tensions regarding patient autonomy and the use of technology that integrates with everyday life. We also seek to building on research demonstrating the value of peer sharing, investigating patients attitudes to sharing data with both clinicians and other patients.
The work in this paper was undertaken in collaboration with a large hospital (the Beacon Hospital Dublin) which includes a specialized cardiology unit. To support patients following hospitalization for an acute cardiac incident, the hospital provides a 6-week Phase 3 CR program based on standard UK CR model [
10]. The research was divided into two stages. The first stage involved co-design with a CR team to develop a data system to support the Phase 3 CR program. The data system involved gathering and sharing objective and subjective data through consumer devices and self-report questionnaires. The second stage involved a field study, where the system was deployed as part the CR program. The field study ran across four separate iterations of the CR program. Each individual CR program lasted six weeks and had a separate group of patients. In total the field study lasted for 18 weeks (the individual programs overlapped) and involved 16 patients. Over the course of the field study, the data system was iteratively refined, based on ethnographic observations of its usage, clinicians’ feedback, and patient data. Ethics permission for all studies was granted by Beacon Hospital Research Ethics Committee (BHREC Ref: BEA0159) and the University College Dublin Research Ethics Committee (REERN: LS-E-21-145-Tadas-Coyle).
We address the following research questions: How did clinicians and patients use the data in and between rehabilitation classes? Is the combination of objective activity data and subjective experience data meaningful and actionable for clinicians? What were the patients’ attitudes towards self-tracking and sharing their data with both clinicians and peers? And finally, is there evidence that our approach is feasible and aided the transition to self-management?
4 Co-design of the Data Gathering and Sharing System
The co-design stage involved working collaboratively with a CR team to develop a data gathering and sharing system. Reflecting the recommendations of Andersen et al. [
3], we sought to design a system that would be (1) meaningful to both patients and clinicians, (2) actionable for both clinicians and patients, and (3) feasible within the organizational and social context. Understanding the CR team's experiences in running a Phase 3 CR program, including the activities involved, how they are run, the different clinicians involved in the program and their concerns and needs, was key to achieving these aims.
While patients were not directly involved in the co-design activities reported here, it is important to emphasize that this work built directly on prior research by our team, which did involve working directly with cardiac patients and focused on understanding their needs [
62,
63]. Throughout the co-design process a member of the research team acted to represent the views of patients based on this direct experience and related patient-centred literature. While the co-design reported here was undertaken in collaboration with clinicians, the emphasis was on addressing the needs of both patients and clinicians. Potential limitations in this approach are also addressed in the Limitations section of the paper.
4.1 Participants
The co-design process began with regular meetings between the first author and the chief physiotherapist of the CR program. This was important in building trust and understanding the workflow, organizational, and social context of the CR program. The full CR team then participated in the design process. This included two physiotherapists, including the chief physiotherapist, and a range of clinical specialists, including one dietitian, one pharmacist, one occupational therapist, and two CR nurses. Within this paper, the word ‘clinician’ is used to describe a range of roles that involve direct contact with patients.
4.2 Methods
The co-design process involved a combination of focus groups, small group meetings, regular discussion of design decisions, and shared prototypes. At the beginning of the first focus group, and to emphasise a patient perspective, findings from previous studies with cardiac patients in CR programs were presented. These included: patients need for physical and emotional safety; the importance of normalizing new experiences through social interactions and peer-experience sharing; the importance of both general knowledge of cardiovascular health and personalized insights; and a focus on capability rather than limitations [
62,
63]. Clinicians were informed that the aim was to understand their clinical needs and workflows and to develop a data system that would support the CR program, addressing both their needs and the needs of their patients. As the focus groups took place while COVID-19 restrictions were in force, they were held remotely over Zoom. The focus groups and meetings were moderated by the first author and notes were taken.
4.3 Key Design Decisions
This section highlights key design decisions that led to the data system used in the later field study. These design decisions were made on an iterative basis, through on-going discussions and focus groups with the clinical team. Regular discussions within the design team also took place to help ensure decisions also considered the needs of patients.
4.3.1 Organizational Context.
Figure
1 illustrates the structure of the CR program run by the Beacon Hospital Dublin. It is a 6-week program, based on standard UK CR model [
10] and consists of a one-hour supervised exercise class twice a week (Tuesdays and Thursdays) conducted by physiotherapists, and a 30-minute educational talk once a week after one of the exercise classes. Patients are encouraged to walk regularly and engage in physical activity of their choice between classes. Each educational talk is given by a different clinical specialist, including a physiotherapist, dietician, occupational therapist, pharmacist, and CR nurse specialists. The hospital did not have a dedicated full-time CR team. CR programs were managed mainly by two physiotherapists, who were also responsible for attending to other treatments provided by the hospital. Other clinicians were involved as required for educational talks. Due to the COVID-19 pandemic and the restriction on the number of people in indoor areas, the CR program had moved to a hybrid format. Patients had the option to take classes online through zoom video calls or in-person at the hospital. As this was a new format adopted by the hospital and due to the COVID restrictions on the number of patients in a class, the team had to increase the number of CR programs. Before the pandemic, the hospital ran one CR program every 6 weeks with 3 to 4 physiotherapists; now, the hospital had to run 2 concurrent CR programs during the same 6 weeks, with 1 to 3 physiotherapists per program. There was also a lack of staff for the administrative needs of the program. This had significantly increased physiotherapists' workload. Prior to our study the use of technology in classes was limited. Physiotherapists used blood pressure monitors, medical chest bands and watches to monitor patient's heart rate (HR) and blood pressure during the classes. They used a computer to connect to online patients through Zoom video conferencing software and to show presentation slides during the educational talks.
In initial discussions, the chief physiotherapist emphasized the importance of monitoring patients’ physical activity during exercise. This allowed physiotherapists to ensure the patient's HR stayed within safe limits, teach patients about their safe HR zones when exercising, and monitor improvements over the course of the program. They now wanted a way to monitor patients’ activity level and HR outside of the class, to see if they were complying with recommended activity levels and identify patients who needed immediate attention. They felt this would help reduce their workload, allowing them to check patients remotely and follow up with patients who needed further and immediate attention. With this aim in mind, the CR team had previously purchased a number of Fitbits. However, the Fitbits had not actually been used with patients. This was largely because no structured system was in place to support their use. Therefore, the initial focus of our data system was on developing a structured approach to collecting objective data related to patients’ physical activity between CR classes. Given their availability and familiarity to the CR team, and thus feasibility, we decided to use the available Fitbits. While more feature rich tracking devices are available (including research prototypes) Fitbit met clinicians’ core requirement for HR and activity data. From a patients’ perspective it was also important that Fitbits are a widely used consumer device. We felt they would more easily integrate with daily life, help to avoid over-medicalization of tracking and support patients’ preference for normalcy.
4.3.2 Meaningful Data.
In the first focus group the use of Fitbits was discussed with the full CR team. Each clinician shared their data collection needs, past experiences, and their role within the overall structure of the CR program. Physiotherapists directly involved in supervised exercise classes again expressed the need for more information about patients’ activity and rehabilitation progress between classes. In contrast, clinical specialists were mainly concerned with ensuring their educational talks were tailored to the audience. They wanted information about the mindset of patients, their medical condition, and their level of knowledge on the topic covered by their talk. For example, the pharmacist was interested in patients’ medical history, their attitude towards medication, whether they know how and when to take their medication, and who would they approach if they had a question regarding medication. While some of this information was available in their existing patient record system, the system did not provide information on patients’ mindset and understanding of their medications. The focus group led to an extension of the data gathering to include both Fitbit data and weekly patient questionnaires focused on their mindset and knowledge.
4.3.3 Actionable Data.
Subsequent meetings and prototyping focused on actionable data. Prototypes took the form of lo-fidelity paper-based sketches and computer mock-ups, which were shared in a second focus group and discussed on an on-going basis in small group meetings and via email. Several key decisions were taken in relation to Fitbit usage. Fitbits would be given to patients and set it up for them on the first day of the CR program. A clinician dashboard would allow clinicians to see HR, steps, calories, time spent performing intense activity, and sleep. It would provide an overview of all patients’ data on a single screen and then allow them to see hourly data ranging up to the last 7 days for individual patients.
Important decisions were also taken regarding the weekly questionnaires and several trade-offs were agreed. The first weekly questionnaire would collect general information focusing on patients’ confidence and concerns about engaging in physical activity, and their current health status. Clinicians emphasized that patients already receive many emails and questionnaires from the hospital after their surgery and that asking them too many questions could overwhelm them and affect engagement. A maximum of 12 questions per week was agreed. The format of questions was also considered. Clinicians emphasized the need for information which was structured and could be quickly interpreted. They emphasized the need to easily monitor changes in patients’ data through baseline and follow-up questions. Ultimately it was agreed that weekly questionnaires would be based on the language and content of standardised quality of life (QoL) questionnaires, including the SF-12 [
66], MacNew health-related QoL [
31], the HeartQoL questionnaire [
46], and the Dartmouth quality of life index [
42]. These standard questionnaires assess the impact of health on people's everyday life. While they had not previously been used as part of the CR program, they are widely used as QoL measures in medical studies. On the recommendation of the design team patients were given an option to elaborate their answers through open responses. It was felt that open responses would give patients greater scope to share emotional and practical concerns that lay outside the scope of standardized questionnaires.
Some questions were asked every week, with others included once or twice during the program. For example, questions on patients’ level of motivation and how informed they felt were asked every week. Questions relating to the educational talk were asked the week before the talk, with follow up questions asked during the week and after the talk is given. Questions relating to emotions, mental health, and social issues were are not included during the first few weeks as clinicians felt they were not a priority at this point. They felt such data was more useful during the later weeks of the program corresponding to the occupational therapist's educational talk (week 5). Questions that repeated every week were scaled questions ranging from 1 to 10 to allow easy tracking of weekly changes.
4.3.4 The Data System.
The final data system was implemented using off-the-shelf wearable technology (a Fitbit) and commonly used web design patterns. It included of a web application for patients, with functionality to answer weekly questionnaires and store their responses (Figure
2). Patients were provided with a Fitbit and the standard Fitbit app was installed on their phone. Alerts through SMS and email were sent to patients at the end of each week with a reminder to fill out the weekly questionnaire. Patients were given guidance on exercise and safe zones for their HR in rehabilitations classes and encouraged to monitor these zones between classes. Patient's Fitbit data was consolidated and presented as a dashboard on a clinician facing web application. Weekly questionnaire responses were consolidated in Excel spreadsheets and shared through emails. This decision was based on discussions where clinicians said email and Microsoft Excel were part of their weekly workflow.
The system was designed and implemented in such a way that it could be adjusted over the course of the field study. In many respects our approach mirrored the Wizard-of-Oz technique used in HCI studies. Traditional lab-based Wizard-of-Oz studies enable the observation of a user interacting with an interface without knowing that the responses are being generated by a human rather than a computer [
68]. This allows researchers to gain an early understanding of user experience and uncover limitations in the technology prior to developing a fully operational system. We applied this approach in our study. Core functionality of the system, including the weekly reminders sent to the patients, presentation of patient's weekly questionnaire responses, data extraction and sharing of patient's Fitbit data was handled manually by the first author. However, it was done in a way that replicated an automated system and from the patient and clinician perspective it was an automated system.
Not implementing a fully automated system allowed flexibility for the system to be easily modified and iterated throughout the CR programs. The aim was to investigate how patients and clinicians used the system and make ad-hoc changes based on their experiences. In practice, physiotherapy classes and educational talks were usually conducted on Tuesdays and Thursdays of each week. The first author, as the human working behind-the-scenes, extracted the data from patients’ Fitbit devices and weekly questionnaire responses and shared them in a consolidated format to the CR clinicians on Mondays.
4.3.5 Allowing Patients to Share and See Other Patients’ Data.
As noted in the Related Work section, prior research, and our own work with patients going through CR, suggests that shared experiences can be very helpful during the rehabilitation process. During the co-design process the possibility of allowing patients within CR programs to share their own data and view the data of other patients was openly discussed. The first author described the results of prior work and the potential benefits of shared experiences in helping patients to both normalize their own experience and learn from the experience of others. Clinicians expressed a strong opinion that it could be harmful to share Fitbit data between patients and felt it could have a negative impact. They explained that exercises in CR classes are tailored to each patient and progress of each patient could be misinterpreted if shared without detailed explanation. Given this concern it was agreed that patient data sharing should not be included in this study. Instead, we used the post field study interviews to develop a more detailed understanding of patients’ attitudes towards sharing data with their peers. While prior research suggests that clinicians’ concerns around peer-sharing can be overstated [
33], we believe this was a balanced overall approach, respecting the concerns of clinicians, but also giving us the opportunity to further explore patient attitudes towards peer data sharing.
5 Field Study
In the field study the data system was used as part of a Phase 3 CR program. The overall structure of CR programs and data collection are outlined in Figure
3. The field study involved direct observations of the CR program, together with feedback from CR team and semi-structured interviews with patients. Research ethics approval was granted by the Beacon Hospital Research Ethics Committee (BHREC Ref: BEA0159) and the University College Dublin Research Ethics Committee (REERN: LS-E-21-145-Tadas-Coyle). The study ran from Aug-Dec 2021 at the Beacon Hospital Dublin.
5.1 Recruitment and Participants
During the field study the group size for CR classes was limited to 4 people exercising together in-class due to COVID-19 restrictions. Some programs had patients who opted for online attendance in classes via Zoom video calls. With the help of the chief physiotherapist, we recruited participants through purposive sampling. All patients attending CR programs during the field study were potentially eligible to participate. The final decision on whether patients were invited to participate was based on the judgement of the chief physiotherapist, giving consideration to whether participation could be detrimental to their rehabilitation. Ultimately all patients were invited to participate. The study ran across 4 separate CR programs. A total of 16 patients participated in the study, with a mean age of 59.25 years, and including 13 males (Table
1).
5.2 Data Collection
After consent was received, patients completed a pre-study questionnaire to collect demographic information, current smartphone and smartwatch usage, a reflection on their current health condition, including physical and emotional health, and finally their attitudes towards sharing Fitbit and weekly experience data with their peers and CR clinicians. Thereafter, a Fitbit Charge 2 and web App was set up for each participant. The Fitbit was used to collect their physical activity and HR data. The web App was used to collect weekly CR experience data through questionnaires. The standard Fitbit app was also installed on patients’ phones.
It is important to note that the Fitbit and questionnaire data was shared only with the CR team. The authors did not have ethics permission to analyse this data or directly share it with other patients in the CR program. This data was thus not subject to analysis by the research team.
5.3 Ethnographic Observations
The field study also involved ethnographic observations of the CR programs. The first author attended CR programs and observed in-person and online classes. Observations were collected by taking notes, which were guided by a checklist (Supplement file). The aim was to understand the actual usage of collected data during classes. The author did not interfere in any proceedings of the classes during these observations. Feedback was also collected from the CR team in debriefing meetings after each CR program.
5.4 Semi-structured Interviews
After each of the four CR programs, semi-structured interviews were conducted with patients. All the patients from the field study participated in the interviews. They were asked about their experience of wearing the Fitbit and answering weekly questionnaires for the entire duration of their CR program and the impact it had on them, including reflection on their physical and emotional health condition. The full interview guide is included in supplemental files. The interview also included questions to understand their attitudes towards a future patient-to-patient (peer) experience sharing application and their willingness to share their experiences with peers via such an app. While patients were not shown actual peer data, they were given illustrative examples of such data during the interviews.
Audio recordings of the semi-structure interviews were transcribed verbatim. The initial coding framework was jointly agreed by the first and third author and was derived from prior literature and our prior work with patients [
62,
63]. Initial codes included “normalizing new experiences”, “personalized insights”, “shared experiences”, “safe zone for physical exercise”, “emotional support”, and “focusing on capabilities”. A combination of inductive and deductive thematic analysis was applied to this data [
11]. This analysis was led by the first author who reviewed and coded all transcripts. Regular meetings were held with the third author, who also reviewed and directly coded a smaller portion of the transcripts. New codes were discussed and refined based on joint agreement to create a final set of codes, which were applied to all transcripts. Following this the data was iterated to produce high-level themes. Again this work was conducted jointly by the first and third author. Finally these themes were abstracted out to three main themes presented in this paper.
7 Discussion
The field study allowed us to investigate the use of patient-generated data as part of a CR program, which itself is designed to support the transition from hospitalization to self-management. In the Related Work section, we noted the recommendations by Andersen et al. [
3] that systems should (1) make sense to both patients and clinicians, (2) be actionable to both clinicians and patients; and (3) be feasible within the organizational and social context. Here we reflect on whether our approach supported these recommendations. We also return to our key research questions: How did clinicians and patients use the data in and between rehabilitation classes? Is the combination of objective activity data and subjective experience data meaningful and actionable for clinicians? What were the patients’ attitudes towards self-tracking and sharing their data with both clinicians and peers in their CR program? And finally, is there evidence that our approach is feasible and aided the transition to self-management or self-care?
Throughout this section we highlight key recommendations for the design and use of patient generated data to support rehabilitation and the transition to self-care.
7.1 Use of Data by Patients and Clinicians
Table
3 summarises how activity and self-report data were used by patients and clinicians. It also captures the frequency at which each group engaged with data. Important to note here is a contrast in the use of the data. Clinicians viewed the data less frequently and took a more task orientated approach, with an emphasis on supporting pre-existing tasks in their workflow, e.g., checking compliance with exercise. This is perhaps unsurprising given the time constrains they face. In contrast patients viewed their Fitbit data more frequently and took a more exploratory approach, investigating the functionality of Fitbit app and asking questions in classes about different ways in which it might be used (e.g., exploring guided breathing exercises and Fitbit visualisations and goal setting).
Design recommendation:•
Support both task oriented and exploratory use of patient generated data and the associated self-tracking systems.
7.2 Making Patient-Generated Data Meaningful and Actionable in a Clinical Context
Pantzar and Ruckenstein suggest that due to the automated and standardised collection, data from tracking devices can be perceived as essentially objective [
48]. They argue that the meaning of tracking data is in fact deeply tied to the particular contexts in which it was collected. A study by Ada et al. on the use of Fitbits to support the treatment of patients with post-traumatic stress disorder supports this view, again arguing that context was crucial in supporting patient-clinicnan interactions around data [
43]. Pantzar and Ruckenstein propose the concept of “situated objectivity”, which combines mechanical objectivity with the important role of context in knowledge formation.
Findings of this study suggest that, in the context of CR, experience data collected through weekly questionnaires played a vital role in supporting “situated objectivity” and in making data meaningful. As an example from our findings, Fitbit data might indicate low activity levels by a patient in a certain timeframe. However, there could be multiple reasons for this: the patient's ability to function could be impaired due to anxiety and stress; low motivation due to external factors; or it could reveal a fault with the Fitbit, or the patient simply not wearing the device. Collecting experience data along with physical activity data on a regular basis provided context to the objective data that was collected automatically using Fitbit. It allowed clinicians to get a better understanding of what limited their patients’ physical activity, why their recommended HR levels were not achieved outside of class, and their motivation levels to maintain physical activity outside of class.
Of course, to be clinically useful data must also be actionable and fit with a clinical workflow. The research in this paper was motived by studies such West et al. [
69] that identified barriers to the use of patient-generated data in clinical setting. Alongside data lacking context, barriers included: incomplete data, insufficient time, unfamiliar structure, and misaligned objectives. They report that clinicians want patient-generated data to be interpretable ‘at a glance’ in a short amount of time. This is possible if the data is presented to them in a familiar representation [
69]. In our study we addressed these needs in two ways: through structured self-report, and using summarisation. The structured approach to subjective data collection was grounded in clinicians’ workflow and based on standardised medical questionnaires. These questionnaires were adapted and shortened to reduce the demand on patients. The release of questionnaires was timed to concide with relevant points in the CR program and made use of repeat measures to allow clincinans to track progress.
Importantly, the inclusion of more flexibly or open approach to data collection also served a valuable purpose. In our case an open approach was supported through open questions around patients’ experiences. This gave patients greater flexibility to describe their experiences and provided a channel share experiences they might have been reluctant to discuss face-to-face. In future studies this open approach could be extended beyond simple open questions, considering for exampe a wider range of expressive media (e.g. audio, image, drawing or video). In exploring this more open approach it will however be important to recognise and balance the time demands and workflow constaints of clinicians. In our study patient data was filtered and summarised for clinicians. This was enabled through a Wizard of Oz approach, but could be automated in future systems. For example, it was possible to filter out the important information present it to the clinicians in the form of a summary or highlights. This made the captured experience data quicker to interpret and glanceable for the clinicians. After experiencing the system in this study, clinicians proposed for a future App that could show dashboard for Fitbit data along with the corresponding experience data highlights with an ability to select timeframe and an option to see more information of a specific experience data highlight if needed. There is also potential for making the highlighting process automated through natural language processing [
16,
30].
Design recommendations:•
Collect both subjective experience data and objective tracking data. Subjective data can help to make the objective data more meaningful and supports ‘situated objectivity’.
•
Structured experience data, combined with filtering and summarisation, can help to make patient generated data actionable for clinicians.
•
An open approach to patient experience data is also helpful. It gives patients an additional channel to communicate issues they might be reluctant to discuss in-person with clinicians.
7.3 Designing for Peer Sharing
Our intial (and ultimately naïve) assumption was that patients would welcome the opportunity to learn from the data of other patients, but might be reluctant to share their own data. This was not the case. Instead participants were open to sharing their data, but they had reservations about seeing peer data. Key to this finding is that rehabilitation is highly personalized journey. Patients felt that viewing shared Fitbit data would only be useful if their peers were closely match in terms of physical fittness and medical condition. In this, patients were echoing the doubts expressed by clinicians during our initial co-design process. Clinicians felt strongly that sharing Fitbit data between patients could have a negative impact. They have the expertise to carefully tailor exercise programs to each patient and progress is understood in the context of personal capability. Without careful and potentially time-consuming explanation, shared patient data could be misinterpreted or could cause other people to become demotivated.
We now believe that level of detail is critical in peer sharing. Tracking technologies enable the sharing of detailed information. This is not what rehabilitation classes actually facilitate. Classes provide a space for participants to give and receive general social support and encouragement, based on a broadly common overall experience. Our ethnographic observations support this view. Patients did not have detailed conversations about their health or emotions. Nor did they have detailed conversations about their Fitbit data or the weekly self-report questionnaire. Aside from general conversations about Fitbit usage, patients mainly talked about casual happenings in their life and general news topics. Ultimately, from a peer support perspective, it appears that broad high-level support is beneficial, but detailed sharing is less desirable, or only becomes useful if it can be clearly contextualised or matched against individual cases. As described in the Related Work section, Hartzler & Pratt [
29] investigated of the differences between information shared by patients and clinician. They found that while it is less comprehensive, information shared by patients is often more narrative, practical, and emotionally supportive, in comparison to the more formal information shared by clinicians. This was mirrored in the informal chats we observed during rehabilitation classes.
We do not want to dismiss the potential value of more detailed peer sharing, certainly not based on this one study. It is possible that our findings reflect the limitations of our system. Prior literature provides strong support for the value of peer sharing [
45,
67]. It is also interesting that patients were willing – and often enthusiastic – about sharing their own data. It is possible the act of sharing is valuable in and of itself, e.g., as an opportunity to validate their experience. Timing was also important in patients’ attitude to sharing data. While patients felt immediate sharing could have a negative impact, delayed sharing was viewed more positively. Following this study clinicians expressed an interest in understanding the impact of making their patients’ Fitbit data more visible during the CR class, so that patients can see each other's in-the-moment data, but in a structured and supervised manner. Such work could draw on HCI research that highlights visual alternatives to the numeric presentation of data. This has been found to emphasize multiple possible meanings of data, potentially reducing the emphasis with quantified achievements [
35,
59,
65].
We also see an opportunity to build on research in recovery narratives. Recovery narratives are structured, first-person accounts of recovery [
37]. They are rich in personal and contextual detail, and often make use of text and multimedia content (e.g., photos and audio recordings). The patient generated data in our study was not rich in this manner, focusing instead on Fitbit data and brief experience questionnaires. Recovery narratives have been explored in detail in mental health literature, with evidence for the benefits of both creating personal narratives and viewing other people's narratives [
37,
57]. They have also been explored in stroke rehabilitation [
21]. We believe they are worthy of further research in the CR space, as they could provide a structured, narrative, and practical approach to peer sharing.
Design recommendations:•
Carefully consider the level of detail and timing when designing peer sharing features.
•
Sharing detailed information may not be beneficial and should be approached with caution.
•
Delayed sharing, use of a narrative format, and a focus on lived experience may be beneficial, but again careful consideration is required.
7.4 Supporting the Transition from Clinical care to Self-care
The final research question in this paper asked if our approach to tracking and sharing data is feasible and aids the transition to self-management? Based on our findings we believe the approach is feasible. It made use of a widely available tracking device and patients responded positively. The approach also integrated well with the workflow of the CR program. Whether the approach aided to the transition to self-management is more difficult to address. Strong evidence would require a longer-term study, with direct measurements of improvements (or a reduction) in self-management. This was beyond the scope of our study, but the initial evidence is promising.
Pollack et al. [
53] identified three important challenges for patients recovering from illness: (1) lack of support for health knowledge, (2) lack of opportunities to access resources, and (3) lack of opportunities to promote self-efficacy. Tadas et al. expanded on the need for health knowledge [
63] highlighting CR patients initial need for general knowledge, shifting towards a need for detailed personalized knowledge as they transition to self-care. Traditional CR programs are well suited to supporting general health knowledge. Self-tracking data enhanced the potential for detailed personal insight. Sharing patient data with clinicians also enhanced patient-clinician communication and supported clinicians’ preparation. Self-report questionnaires provided channel for patients to share experiences they might be reluctant to discuss in-person.
A recent article on what it means to empower a patient identifies self-efficacy as one of the essential components of empowerment [
20]. Greater self-efficacy is ultimately dependent on patients’ independent activity between classes. Our study suggests that self-efficacy was supported in several ways. Both the timing and context in which tracking was introduced were important. The Fitbit provided a connection between in-class and independent exercise, enabling patients to mimic the safe exercise zones identified in class and build their confidence. Even those who were new to self-tracking technology became confident by the end of the CR program and could continue achieving the recommended HR and activity levels the help of their smartwatch. Introducing the technology at the beginning of the program was important in this regard.
Design recommendations:•
Use self-tracking and patient generated data to provide a connection between clinical and non-clinical settings. This can help to bridge the gap between supervised and self-managed care.
•
Self-tracking technology should be introduced at an early point in the rehabilitation process.
8 Limitations
As mentioned above, patients were not directly involved in the co-design process of the data system. Although a member of the research team acted to represent patient views throughout the design process, based on our own direct experience of working with patients, we acknowledge the possibility of bias towards clinicians’ priorities in the design. We accept the design of the system might have been different if patients were directly actively involved in the co-design process described in the paper.
We further acknowledge that, as the study took place in a single hospital in Europe, all our participants were of a western European background. All participants were classified as middle-income. We acknowledge that the future work should look at participants with more diverse backgrounds. However, we also note that the participant demographics reflect the real-world situation in the Beacon Hospital Dublin.
Finally, our findings were over representative of patients who took the CR program in-person compared to online. We recommend that future work balances this representation to get a richer perspective and possibly compare perspectives of remote and in-person rehabilitation.
9 Conclusion
This paper contributes to our understanding of the use of patient-generated data in a clinical context, derived from co-design research and a field study conducted as part of a CR program. Our findings suggest that subjective data can help to make objective data more meaningful. A structured approach to collecting subjective data helped to make data actionable, providing information clinicians could quickly understand and act on, e.g., through conversations with patients and tailoring educational talks. Findings suggest the approach is feasibly and supported clinicians’ workflow. Modifications to the clinician dashboard, increased use of email and data summaries were also important in supporting this workflow. The use of a commercial activity tracker was well received by patients. It provided a connection between in-class and independent activity. This helped patients to stay motivated, understand their cardiac health, and build confidence in their ability to exercise safely. Both activity data and self-report subjective data further bridged the divide between clinical and independent use by helping clinicians to prepare for classes, gain an insight into patients’ mindset, and by supporting in-class conversations.