Background
Type 2 diabetes is an increasing public health concern affecting > 400 million people globally [
1] and even more people affected by prediabetes. Prevention is of paramount importance and efficacy trials have shown that programs focusing on lifestyle modification can contribute to reducing diabetes risk [
2].
Several countries have implemented national diabetes prevention programs to attenuate the incidence of diabetes [
3‐
5]. Programs in Finland and China have helped people to sustain healthy lifestyle changes to diet and physical activity for at least a year [
3,
4]. Early findings from the UK’s National Diabetes Prevention Program have shown successful referral rates [
5] and that the program can promote health messages in a social setting [
6]. However, it is also important to consider how people at risk are supported outside of such programs, both during the programme duration (in-between sessions delivered) and after the programme ends.
The digital transformation of healthcare has grown rapidly over the past decade, with the traditional provision of medicine increasingly supported by digital tools [
7]. Sitting under the broader umbrella of electronic health (eHealth) technologies, mobile health (mHealth) technologies are increasingly capable of monitoring health status and encouraging changes in lifestyle behaviours. To help encourage people to change their behaviour to promote better health, a taxonomy of behaviour change techniques (BCTs) was developed to identify “active ingredients” and improve the reporting of intervention content to ensure evidence-based techniques are employed and recorded appropriately [
8]. Key techniques include feedback (such as biofeedback and haptic feedback), goal setting and self-monitoring (of behaviour and outcome). Evidence suggests that combining self-monitoring with at least one other self-regulation technique (such as goal setting) has been associated with improved intervention effects [
9,
10]. Employing self-monitoring of health and behaviour in parallel aligns with the sense-making perspective [
11]. Briefly, it involves evaluating new information in relation to existing understanding (perception) and, if the new information does not align with existing understanding, individuals engage in sensemaking (inference) and experimentation (action).
An important limitation to promoting behaviour change for better health outcomes has been the assumption that people are willing to make changes today to only see the benefit years or even decades later [
12]. Likewise, people often pay little attention to the cumulative consequences of small, repeated decisions which in combination have a marked impact [
13]. It is increasingly possible to observe the acute effect of lifestyle choices on health through technologies. However, there are several limitations to conducting research using mHealth technologies. Two key limitations often cited relate to low participant engagement in using the intervention [
14] and the critical time lag where technology can become outdated by the time the research trial finishes [
15]. Our aim was to explore how people at risk of type 2 diabetes engaged with real-time feedback on their physical activity and glucose levels over several weeks.
Methods
This paper is written in accordance with the Standards for Reporting Qualitative Research (SRQR) [
16] and the Consolidated Criteria for Reporting Qualitative Research (COREQ) [
17]. This study was approved by the Loughborough University Ethics Advisory Committee (R17-P049).
Context and study population
This qualitative work formed part of Sensing Interstitial Glucose to Nudge Active Lifestyles (SIGNAL) feasibility trial [
18]. All participants in the SIGNAL trial consented to be contacted about an interview, using purposive sampling to ensure representation of ages, genders and group allocations. Briefly, 45 participants could access feedback from a Freestyle Libre glucose sensor and Fitbit Charge 2 activity monitor over 6 weeks. Participants were aged ≥40 years and identified as being at moderate-to-high risk of developing type 2 diabetes using the Leicester Risk Assessment Tool [
19] in Leicestershire, UK. A more detailed description of how the technologies were deployed is provided.
Intervention
Briefly, 45 participants were randomized to one of three patterns of access to feedback from the Freestyle Libre glucose sensor (Abbott, Alameda, CA) and Fitbit Charge 2 activity monitor (Fitbit Inc., San Francisco, CA), over 6 weeks. No participants withdrew from the study.
In group 1, participants could access feedback from the glucose sensor for all of the 6 weeks but were also able to access feedback from the activity monitor in the final 2 weeks. In group 2, participants could access feedback from the activity monitor for all of the 6 weeks but were also able to access feedback from the glucose sensor in the final 2 weeks. Participants in group 3 could access feedback from both the glucose sensor and activity monitor for all of the 6 weeks.
The glucose sensor communicated with a smartphone application and showed feedback relating to glucose level, direction of glucose trend, time in range and daily patterns. Participants had to scan the glucose sensor to transfer data from the sensor to the application by hovering their smartphone over it temporarily at least once every 8 h to avoid data loss. The activity monitor communicated with a smartphone application too, but also presented feedback on a wrist-worn device. The data were transferred to the application via Bluetooth. The activity monitor showed feedback relating to the number of steps taken, flights of stairs climbed, calories, distance travelled and heart rate. Brief haptic vibrations were delivered through the wrist-worn device to remind the wearer to move regularly.
Data collection methods
Twenty-six semi-structured interviews between July–October 2017 were conducted at Loughborough University. Only the participant and researcher were present during the interviews. Interviews were scheduled to occur at the end of the 6-week intervention, taking place in evenings or the weekend. Interviews were conducted by MO and lasted < 60 min with no repeat interviews. Enrolment and interviews continued until thematic saturation was reached [
20]. Transcripts were not returned to participants for comment or correction.
For reflexivity, MO is a male Postdoctoral Researcher with expertise in wearable devices in patients with long-term conditions. MO received training prior to data collection and ongoing support from CB who is a Health Psychologist with expertise conducting qualitative research. Participants were introduced to MO as a member of the study team looking to understand participant perspectives of the trial (and otherwise independent to data collection). MO informed all interview participants that his main research interests lie in people with respiratory disease but has been involved in the use of technology to support long-term condition prevention and management. No field notes were recorded.
Data collection instruments and technologies
Interview questions were directed at revealing how participants intuitively engaged with the glucose and physical activity feedback presented by the two devices (Table
1). The schedules were initially developed by MO and CB and tested in the first couple of interviews. Interviews were audio recorded (Voice Recorder & Audio Editor smartphone application, TapMedia Ltd).
Topic Guide
Opening
Introduction Consent confirmed
Questions
How did you feel taking part in a study about your health? What were the reasons behind taking part? How did you find the devices? Can you describe how you used the wearable devices? How did you find the feedback provided by the devices? What did you think of the goals that were in place? How did you get on with the glucose feedback? How did you get on with the activity feedback? How did you find accessing the two types of feedback at the same time? How did receiving feedback make you feel? Is there anything you learned from taking part? Do you think anything has changed? What advice would you give to others using the technology?
Closing
Do you have anything to add? Do you have any further questions?
End
|
Data processing and analysis
Interviews were transcribed verbatim by a professional transcription service. All names within the transcripts were removed and pseudonyms allocated. Qualitative software (NVivo version 11) was used for data management and to support thematic analysis. The transcripts were read and then reread by both MO and FD; this helped familiarisation with the breadth and depth of content discussed. Initial codes were then generated systematically for text that appeared relevant. After all transcripts were analysed, codes were collated into potential themes by MO and FD independently. Potential themes were discussed and reworked with the additional involvement of MW and CB until key themes were generated for the entire data-set. Names for the master and sub-themes were agreed amongst all authors to represent the essence of each theme, including choice of quotes to represent each theme.
Techniques to enhance trustworthiness
Discussion
This study explored the perspectives of people at risk of developing type 2 diabetes receiving feedback relating to health and behaviour in a real-world environment. Our findings revealed that participants were driven to engage with the two devices either by themselves, device notifications or other people. Some participants could recognise a relationship between their behaviour and their glucose levels and behaviour change resulted. However, comments were raised that the data shown lacked meaning for several participants and there were barriers to making changes to diet and physical activity levels.
Participants made changes to their diet and physical activity levels as a result of recognizing the link between behaviour and physiology; driven primarily by the feedback provided by the glucose sensor rather than the activity monitor. This suggests that having access to physiological feedback can raise self-awareness and deepen understanding of how the body functions. It appears possible for people to interpret how their behavioural choices, such as going for a walk, immediately impact glucose levels and there did not appear to be any gender differences as to whether people recognised this effect. Real-time access to glucose levels may act as a silent persuader to encourage positive behavioural choices [
21]. The notion of seeing how glucose or blood pressure can vary in relation to other behaviours, including diet and exercise and extending to medication, has been observed elsewhere [
22‐
25]. Similar to present observations, studies involving self-monitoring of blood glucose have shown potential for people cutting down on sugary foods [
24], increasing activity levels [
23,
24], and improving medication adherence [
25,
26]. Being able to understand physiological data in the context of wider factors can help people assign meaning to the feedback [
27] to supplement specific physical activity feedback. Participants recounted how going for an after-dinner walk lowered their glucose faster than if they were sedentary and this observation was not facilitated by feedback from the activity monitor.
Participants naturally engaged with the glucose sensor around mealtimes or at set times of the day, facilitating the recognition of the relationship for some. This aligns with the importance of the frequency and timing of self-monitoring blood glucose levels [
28,
29]. However, some participants structured their engagement around avoiding data loss, bringing into question whether they would have engaged had this requirement not been in place. However, no gender differences were recorded. The variability in the approaches taken by participants demonstrates the importance of encouraging scans around opportunities to learn about physiological responses to behaviours. Scanning was also mentioned in the context of showing other people, reflecting findings of a text messaging intervention where people openly shared the messages with family members [
30].
The novelty effect of having access to new technologies supports existing phenomena; however, there were contrasting reasons behind this observed reduction in usage. Some participants became more efficient with interpreting the data or how they did not need to look at data as frequently because they become increasingly aware of bodily symptoms and signs [
28,
31]. Our findings emphasize the importance of understanding the reasons why some people use these technologies less frequently. It is worth noting here also that a reduction in use over time may be because our sample comprised people at high-risk, rather than people with a diagnosis of type 2 diabetes. Targeting high-risk populations can have implications. One in particular is that the information provided by such technologies may not show sufficient health risk so, despite being categorised as being at high-risk, as things stand their physiological parameters may be healthy. Maintaining normal physiological health is paramount but this information may not be motivational to make changes if no changes are visibly required. Another implication relates to the cost of targeting at high-risk groups; namely, the number of people living at high-risk far exceeds the number of people living with the condition [
32] and so an economic assessment would be needed to confirm a return on investment.
The reduction in usage reflected that many participants were unable to respond to the feedback being presented. Studies have previously described how people with diabetes often find glucose levels challenging to interpret [
31], and how more than half do not know what action to take [
24]. Digital health technologies may be appropriately placed to offer support during such events. However, data alone are unlikely to be sufficient. There was confusion caused by misleading insights into the immediate health effects of chocolate versus fruits. People could be misled into thinking less healthy foods might be better because they cause a better acute glucose response [
22,
31]. The off-the-shelf deployment strategy has identified a need for additional information or training beyond what is provided by the technologies.
Several participants discussed increasing their physical activity, interrupting sitting time and making changes to when and what food was consumed. This may be in part because participants found the glucose feedback motivating to make changes [
33,
34]. It could also be because the two devices provided feedback that was actionable and continuously available [
35] and offered information on the health consequences of behaviour [
36]. However, barriers to behaviour change were notable. Participants found that living with comorbidities, societal norms and weather restricted an opportunity to change their behaviour. This is not uncommon, with the wider literature citing barriers around health problems, lack of time and weather [
37], as well as coexistence of other poor lifestyle behaviours and misinterpretation of messages as barriers to behaviour change [
38]. With continued technological advances, it is increasingly feasible to overcome some of these barriers. For instance, taking into account the context of the person, integrated smartphone sensors could deliver notifications at times of the day where the weather is acceptable or when they are not in a work meeting to create a more receptive environment for behaviour change.
The small proportion of participants with prediabetes limited comparisons with at-risk individuals and recruiting people through community approaches may limit generalizability. Multiple interviews could have provided greater insight into what it was like for participants to use the technologies. Our findings are limited to short-term engagement with digital health technologies and would benefit from a longer duration of access.
Publisher’s Note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.