Mobility refers to the ability to move freely and easily to carry out activities of daily living and refers to movement in all forms, from moving out of a chair to walking [1]. It is necessary for daily tasks, participating in social activities, and maintaining independence [2]. Mobility is a meaningful aspect of health for individuals across multiple health conditions and plays a large role in their physical, social, and psychological experience [3]. In conditions where motor symptoms cause functional impairment, understanding an individual’s walking, posture, and how they move is important for measuring and monitoring condition severity [4]. Specifically, walking is an important clinical measure whereby factors such as walking speed and stride length can predict functional decline and increase risk of frailty and mortality [5,6]. Hence, it is essential to be able to measure walking accurately and meaningfully convey these data to patients to provide insights into progress and changes in their condition over time.

Recent technological advances in wearable devices (such as body-worn sensors) offer new potential for measuring mobility continuously in real-world contexts (collectively termed digital mobility outcomes [DMOs]). This in turn presents the opportunity for novel insights into aspects of health that are important to patients, such as their real-world walking speed and daily step count [7]. Crucially, providing individuals with access to, and insights from, their own data from such devices is linked to patient autonomy, enhanced health professional-patient communication, and improved compliance with physician recommendations through the enhanced trust between patients and those involved in their care [8-12]. However, the effectiveness of this feedback hinges upon the meaningfulness of the outcome to patients, particularly whether their symptoms and symptom burden can be reflected by the outcome and be interpreted by patients [13-15].

Integration of the wearable devices into health care systems presents significant challenges in the visualization of the complex data and the insights these generate [16]. Mobility can be affected by multiple factors, such as changes in symptoms, environmental factors, and medication, which creates challenges and added complexity when communicating insights from the data [17-19]. Hence, it is essential to ensure that data visualizations are both accessible and understandable to both patients and clinicians and reflect the nuances of patients’ mobility while also representing meaningful aspects of their health [14,20,21]. Creating visualizations that enable an individual to come to a full understanding of a person’s overall health condition can be challenging; hence, several approaches have been attempted without one prevailing approach [19]. In previous visualizations from wearable devices, a range of traditional visualization approaches, such as radar plots, bar charts, line plots, and box plots, have been used, as well as more novel approaches, such as a calendar depicting 6 months of data in 1 visual [21,22]. Consequently, exploring data visualization preferences is important yet nuanced, and the creation of the visualizations requires an understanding of the patient experience.

This work builds upon the efforts of the Mobilise-D consortium to validate DMOs and provide novel insights into several long-term health conditions, including Parkinson disease (PD), multiple sclerosis (MS), chronic obstructive pulmonary disease (COPD), and proximal femoral fracture (PFF) [17,23-25]. Research has been carried out to establish the technical, clinical validity, and acceptability of wearable devices and the analytic pipelines applied to quantify DMOs [17,23,26]. An additional part of ensuring implementation of the methods and outcomes is to determine how best to visualize DMO data for patients. Consequently, this study aimed to explore patient visualization preferences by (1) determining which DMOs are meaningful to individuals living with diverse long-term health conditions and (2) identifying how to best visualize them from the perspective of end users [12].

Previous research on wearable device data visualization has highlighted the lack of consistent preferences regarding how best to visualize the data. Several approaches have explored the use of varying levels of granularity, with some studies attempting to visualize a day and others 6 months [21,22]. While this was highly variable depending upon the outcome, for mobility it is common to use a week of data [18,19,21]. A major point of interest in the literature was embedding contextual data within visualizations, particularly in relation to clinical use to enhance the inferences made [18,19]. Condition-related information, such as symptoms, emotions, and medications, can influence the lived experience, and the data from wearable devices should be able to reflect this, as well as the time and location of activities [27,28]. While we know specific design approaches and elements are key to the creation of comprehensive visualizations, it is also important to consider the individualized nature of preferences when developing meaningful visualizations [12]. Hence, this study aimed to map DMOs into the lived experiences of mobility and symptom burden and explore the visualization preferences of patients using a week of DMO data.

This study design was based on a Delphi methodology in which patients acted as subject matter experts to iteratively provide feedback on the cocreation of data visualizations through 3 rounds of questionnaires. It was designed in collaboration with academic researchers of Mobilise-D and the members of the Mobilise-D Public and Patient Advisory Group (PPAG) [29].

In total, 3 rounds were developed to explore the impact of mobility and symptoms and how DMOs can be visualized. A Delphi methodology was decided upon as it allows us to reach a consensus of patient preferences with an iterative, anonymous, multistage approach with controlled feedback of comments and scores on a 5-point Likert scale [30,31]. The Delphi methodology recommends 3 rounds of feedback; the first round is to generate qualitative data on a topic and the remaining rounds are to gain a consensus through Likert scales [31]. We followed the recommended 3 round; however, the methods used here will be described as a “Delphi-type process” due to deviations from Delphi-specific design elements. A key aspect of a Delphi methodology is ensuring participants completed previous rounds; however, due to ethical requests, this was not possible within this study [31,32]. This process involved a collaboration between patient feedback and researcher refinement, as described in Figure 1.

The data visualized in this study were collected as part of the Mobilise-D technical validation study (TVS) [23]. Participants with long-term health conditions wore a lower back wearable device (either a McRoberts MoveMonitor+ or the Axivity AX6) to provide real-world monitoring data for 7 consecutive days. Accelerometry data were collected and processed using the algorithms validated by Mobilise-D to produce DMOs [33]. In addition, GPS data were collected for participants, which allowed the data to be categorized into indoor versus outdoor. No activity or medication diaries were collected, which limited the contextual information available when visualizing the data.

This research was designed with members of the Mobilise-D PPAG [29]. In total, 4 international representatives (Belgium, Israel, Italy, and the United Kingdom) from the PPAG (2 with PD, 1 with a previous hip fracture, and 1 with MS) worked with academic researchers to develop each round of questions and visualizations to ensure they were appropriately worded and constructed and that the conclusions drawn from the data in each round were meaningful. Due to the international and remote work involved in the PPAG, perspectives of the digitally excluded unfortunately could not be included with the team.

People with PD, MS, COPD, or a previous hip fracture primarily from Europe (British, Irish, Norwegian, and German) were invited to participate in this web-based survey Delphi-type study. Patients were purposively recruited from newsletters sent out to various patient charities and organizations (including Voice, MS Society, and Parkinson’s UK), social media, and through existing Mobilise-D clinical sites. A sample size of between 10 and 15 people per health condition was proposed in line with guidance for Delphi studies [31].

There were 239 responses to this initial form, which was then used as the distribution list for all rounds of the study. However, because of this method of recruitment, participants were unable to be individually tracked or compared per round, and this meant the participation of individuals in previous rounds could not be confirmed, which is not typical for a Delphi methodology [32]. Data were collected between June 2023 and June 2024. As this study aimed to iteratively build upon previous findings, the lack of participation in previous rounds may have impacted the participant’s understanding of the context surrounding each round. However, each round started with a summary of the project and the findings from previous rounds to mitigate some of this confusion.

Ethics approval was granted by Newcastle University (2514/30280). To ensure anonymity in line with ethical requests, participants were asked to complete an anonymous form whereby their email address was collected and stored securely separate from demographic and response data from the study. These email addresses were stored securely on a password-protected computer and only accessed by researchers involved in questionnaire distribution. This was then used to send participants a link to take part in the study anonymously. Participation was voluntary, and responses were anonymous, with no personally identifiable data collected. All study participants received an information sheet about the study and were provided with the researcher’s emails in case of any questions. Participants were informed that the study should not take any longer than 10 minutes and were assured that they could withdraw without consequence at any time during the study. All participants gave informed consent. There was no compensation for participation in this study.

Participation was dependent upon a round being fully completed; the demographics of participants across the 3 rounds are presented in Table 2. Participation varied across rounds 1 to 3 (n=48, n=79, and n=78, respectively). Across all rounds and participants, 63.4% (n=130) were female participants, 29.8% (n=71) of participants were below degree level or preferred not to say, 94.1% (n=193) were White and 37.6% (n=77) wore a wearable device daily. Considering that the sample size for COPD and PFF was lower than expected, their qualitative responses were included in the analysis but were not considered to have reached consensus [31].

Examples of the mapping process per long-term health condition are provided in Table 3. The experiences described by patients for each health condition are briefly summarized here.

For people with PD, scenarios and examples that related to the themes of physical experience of walking, the context of the walking experience, and becoming aware of the personal walking experience were frequently reported. They described how their symptoms affected their balance and noted that they felt they were not able to do as much because of their symptoms, which forced them to slow down. Participants also described how their walking changed depending on the time of day and location. This impacts where and how far they might be willing to go. They also described feeling more conscious of how they walk in terms of speed and quality to counter the changes to their balance. These experiences could be directly related to DMOs describing stride length, step duration, walking speed, and step count (Table 3), which reflect altered balance, speed of movement, and amount of activity.

People with MS frequently described scenarios and examples that related to the themes of physical experience of walking, the context of the walking experience, and the walking experience as a link between individual’s activities and sense of self. MS symptoms led people to describe how fatigue impacts their lives, influencing their ability to walk distances and to plan activities. They also spoke about how the symptoms of MS influenced their capacity to undertake activities, depending on how they felt on a daily basis. Finally, they described changes to their balance and perception of distance, which impacted their ability to walk smoothly. The DMOs that most frequently mapped to these themes were step count, stride length, walking speed, and step duration.

Limited data were collected from people with COPD or PFF. However, both groups focused on their physical symptoms, which appeared to link most closely to the amount of activity that they could do. People with COPD described breathlessness as an important symptom that led to struggling to walk and use stairs, among others, thus limiting the amount or duration of their physical activity. For people with COPD, these experiences mapped onto step count and number of walking bouts. People recovering from PFF described changes and adaptations since their fracture, including the need to walk slower, increased fatigue, and having to rest because of this, which mapped onto DMOs, such as walking speed and step count.

Following this step, the issues and concepts that were most often mentioned were listed by the health condition and used to identify the data required to develop visualizations in line with these needs. The most relevant DMOs and contexts were extracted from the Mobilise-D TVS and were drafted into visualizations to represent varying mobility impairments across the 4 health conditions; this process is presented in Figure 2 and Multimedia Appendix 3 [23].

Researchers developed and presented 15 visualizations and justifications to participants based on the feedback from round 1. Every long-term health condition was presented with 2 visualizations of fatigue, PD was presented with 2 visualizations on balance, MS was presented with 3 visualizations on doing less, and PFF was presented with 2 visualizations on slowing down (Figure 3 gives the examples). Participants then rated the usefulness and ease of understanding of each visualization. The consensus was reached for the ease of understanding in people with PD for 2 graphs, namely the bar chart showing the indoor versus outdoor data for both balance and fatigue.

The qualitative responses to the draft visualizations were generally positive (summary in Table 4). Participants reported that they were able to understand the visualizations, and some were able to interpret what these visualizations represented. For some, particularly those with PD, more data and variables were deemed necessary to allow a fuller picture of the data being presented. Specifically, people with PD highlighted a desire for additional, contextual information and patient-reported outcome measures to accompany the visualizations. However, there were individuals in each long-term health condition that found the visualizations difficult to understand.

A range of feedback was provided that touched on different aspects of the data visualizations and their method of presentation (Table 5 gives the summary). This was used to improve the visualizations for round 3 (Multimedia Appendix 1). Feedback such as superimposing variables and improving the clarity, accessibility, and visuals of the visualizations was implemented, as seen in Figure 4. Some feedback could not be implemented due to software limitations, such as adding an interactive element to switch between meters and feet.

Researchers developed and presented 8 visualizations and justifications to participants based on feedback from round 2. Every long-term health condition was presented with 2 visualizations of fatigue, PD was presented with 2 visualizations on balance, MS was presented with 3 visualizations on doing less, and PFF was presented with 2 visualizations on slowing down (Figure 5 gives the examples). Patients rated the usefulness and ease of understanding of each visualization. No consensus was reached during round 3 for any visualization.

Qualitative responses to the draft visualizations were generally positive (summary provided in Table 6). Participants reported that they generally understood the visualizations and appreciated the comments surrounding the visualizations that explained the graph and how to interpret the outcomes. Participants found the differentiation between environment and time of day included in the visualizations to be important, as this influences the presentation of their symptoms.

A range of feedback was provided throughout the comments touching on different aspects of the data visualizations and how they were presented (aggregated summary in Table 7). This feedback could not be used within this study but serves as a recommendation for future studies wishing to visualize mobility data. The feedback mainly surrounded accessibility and explanations, as well as the need for additional data to fully interpret the graphs. Interestingly, this round provided contradictory feedback between participants, with some appreciating the infographic visuals while others found them irritating. This indicates the need for personalization of data visualizations.

This study explored patient preferences on how mobility data derived from wearable devices is visualized. Through the multistage cocreation of visualizations with patients as experts, we embedded the voice of patients throughout the research cycle so that the visualization recommendations would be meaningful to them. While these early insights cannot claim a consensus of patient preferences on mobility data visualizations, the study provides important learnings and recommendations for future mobility visualizations.

Participants were generally able to understand the visualizations and interpret what the data may indicate. The next step would be advancing and personalizing the data visualizations to empower patients to gain actionable insights from their inferences. Feedback consistently requested additional data to be presented to enhance the visualizations; this was achieved through embedding indoor versus outdoor categories in certain visualizations as well as superimposing DMOs on some visualizations in round 3. Participants appreciated this extra detail during round 3; however, this was not sufficient in terms of the information they felt was important to help them apply learnings or actions into their daily lives.

Contextual information, such as how the person felt, when medication was taken, and whether they used a walking aid, would be important when interpreting the visualizations. Previous studies have explored the use of calendars to present medication data over time using a similar approach; it would be possible to note when medication changes or interventional changes occur by indicating them with icons [19,21]. Alternatively, exploring the granularity of data may be helpful in addressing these comments; for example, exploring when medication was taken throughout the day may lend itself better to visualizing a single day’s data with icons added when medication was taken to identify any trends related to medication [18,35]. Previous patient-focused visualization studies within literature also discuss the importance of context and additional factors in determining the meaningfulness of a visualization; hence, it is important for these contexts to be embedded into visualizations [12,15,36]. Typical consumer wearable devices display a myriad of health data, such as number of steps and calories burned; however, these visualizations are often oversimplified and, consequently, do not demonstrate the relationship between multiple factors or acknowledge the presence of external, contextual factors [37]. It is important that researchers do not default to the basic bar charts of consumer wearable devices and instead recognize and highlight the intertwined nature of health, environment, and activity, particularly in those with health conditions whereby this additional information is vital.

Some participants found it difficult to interpret the information from the visualizations and apply it to themselves. This is potentially due to insufficient justifications, descriptions, and accessibility of the visualizations [12]. This indicates the importance of working alongside patient advisors to understand the lived experience of long-term health conditions and develop justifications for what and why the data are displayed [32,38]. A major barrier to understanding for participants was the rationale for why certain DMOs had been chosen to represent a symptom. This indicates the need for future qualitative studies to be conducted to map the patient experience to each DMO. Research to understand the link between patient experience and outcome measures is limited but will be essential for understanding the importance of where DMOs fit into the patient experience. Being able to use visualizations to make sense of the “chaos of lived experience” will be key to encouraging the use of wearable devices within health care [28]. While this mapping process was piloted within round 1 of this study, a larger and more in-depth mapping process is needed to bridge the gap of DMO meaningfulness to patients. Already, the benefits of this mapping process have been seen in other areas of digital health technologies, with additional studies currently ongoing [39-41].

An important barrier to interpreting the visualization was a lack of personal connection with the data. As the data presented to participants was not their own, they found it difficult to understand what was being visualized, as there was no context for what the data implied. This was a recurring theme throughout the qualitative analysis and something that should be addressed in future studies by personalizing visualizations to patients’ own data to address these barriers. Visualizing personal data may empower patients to make actionable insights that reflect their own lived experiences, which makes strides toward the potential for personalized health care [42]. This may also address the lack of consensus found in this study, as participants had no familiarity with the data presented, and this may have contributed to poorer understanding and usefulness of the data.

Patients were able to read and interpret bar charts, but their ability to interpret DMOs, such as stride length and walking speed, was hindered due to a lack of understanding of a healthy control’s “normal.” This is a common theme within clinical data visualizations. Providing a “normal” value may be helpful when goal setting using visualizations; however, issues arise regarding how that “normal” value is set and ensuring it reflects the age, gender, and other characteristics of the patient [21]. Other studies demonstrate that the use of visual cues, such as emojis, and user-friendly visual analogies can improve the understanding of the visualization [36,43]. The use of visual cues throughout data visualization should be of particular importance in future work with difficult-to-visualize outcomes. During round 3 we included comments and additional visual cues to better explain the visualization. The feedback implied that this was well received, yet there is scope for further enhancing these visual cues and comments. For example, the ability to relate to famous monuments as a visual cue was very individualized and should be personalized based upon participant location for future visualizations. Moreover, the accessibility of the graph is important. Due to software limitations, the graphs presented were small and difficult to read for some participants; this is something to be mindful of in future studies.

Future studies should explore alternative visualization strategies. Here, data were presented using graphs to display changes during a week and to represent the dynamic nature of patient experiences. However, certain DMOs may be better conveyed through other visual methods. For example, some studies have used motion capture and videos to visualize gait, which may be an interesting avenue for future mobility-related visualization studies [44,45]. Furthermore, for data to be relevant to a person’s immediate context, future work should focus on designing a tool that can provide real-time feedback that supports people’s understanding of their condition [27].

The COPD and PFF health conditions had a lower representation and did not meet the recommended number of participants for Delphi protocols (10 to 15 participants); this limits the understanding we have of how these groups prefer their data to be visualized [31]. This was due to our recruitment methods not being successful for these health conditions. While results indicate that DMOs and justifications should be personalized based on the symptoms and experience of a long-term health condition, the sample size was not adequate to explore any differences in visualization preferences between groups. Further research is necessary to fully understand any nuances that may arise from these long-term health conditions.

The sample of this study was not representative; participants were largely educated from Western countries (primarily the United Kingdom) and around a third of participants used a wearable device in their daily lives. Due to ethical requests, participant demographics were only collected after participants were recruited; it was not possible to retrospectively recruit and address this disparity. In addition, by nature of a web-based survey, it is likely that the digitally excluded were not well represented in this study. It is pivotal that research includes and represents the experiences of underserved communities to improve the generalizability of findings [46,47]. Previous data visualization studies have looked at preferences in people with lower educational attainment; this highlights the importance of diversity within data visualization samples, as preferences may vary between groups [48]. For the integration of wearable devices into health care, it will be paramount to understand how best we can visualize DMO data to ensure a wide range of demographics can access and understand the data provided. To address this, future research should aim to provide a more representative sample [49].

It is essential that future studies are conducted with clinicians and those involved in the care of patients. With the rise of commercial wearable devices, data are being presented to clinicians by patients and being dismissed, as there is a lack of understanding on the clinician’s side for how to interpret and act on the data [21]. It is imperative that the people involved in a person’s care (both the patient and care team) can extract meaningful data from wearable devices. It may be that different forms of visualization may be helpful for clinicians compared to patients; potentially clinicians will be more interested in longitudinal changes, such as baseline to 6 months, than patients. Already studies have explored this for clinic-based sensor data, and this should be explored more widely using real-world data to capture the nuances of mobility [50].

The use of remote mobility monitoring devices is predicted to increase. Hence, how to visualize this data in an understandable and meaningful way to patient populations is of key importance. This study indicates that visualizing these data is possible and can be done in a way that patients understand. While this study did not draft any visualizations that had a strong consensus of understanding and ease of use, we were able to provide several points of recommendation for future visualizations and studies. Particularly, we recommend that future studies explore individualized visualizations, whereby the participant’s own data are visualized both individually and in the context of their condition to better enhance understanding.

Despite the exploratory nature of this work, we have been able to develop several initial recommendations for future visualizations in tandem with existing principles that are applicable both for the visualization of mobility data as well as the visualization of health data in general, which include the following: