Challenges of using a Fitbit smart wearable among people with dementia

Limited research on using smart wearables such as Fitbit devices among people with dementia has shown favourable outcomes. The aim of this study was to explore the acceptability and feasibility of using a Fitbit Charge 3 among people with dementia, living in the community, who took part in the physical exercise component of the Comprehensive REsilience‐building psychoSocial intervenTion pilot study.


| Participants
People with mild to moderate dementia (n = 10) and their caregiver (n = 10) were recruited to the CREST intervention through General Practitioner (GP) practices and local dementia support and advocacy groups and networks (e.g., The Alzheimer's Society of Ireland, Western Alzheimer's). The sample size was determined mainly by pragmatism and what was feasible given the challenges in recruiting the cohort.
The recruitment process has been described in detail in the study protocol. 10 Inclusion criteria included adults with either 1. A formal diagnosis of mild to moderate dementia; or 2. Prescribed dementia medications; or 3. Their GP believes the person has memory problems and the person has a provisional diagnosis of dementia based on the DSM-IV criteria, community dwelling and aged 60 and above. All participants met inclusion criteria 1 and 2. They were also required to have their primary caregiver agree to participate in aspects of the CREST intervention. Ethical approval from the University of Galway Research Ethics Committee was obtained (Ref: . Consent to participate was also obtained from all participants.

| Intervention
A literature search was first conducted in relation to using activity trackers with older adults with dementia and it found support for using wrist-worn trackers, such as Fitbits, to capture data. 3,11 Researchers then explored different Fitbit devices for use in the CREST study, considering cost and requirements such as the ability to track sleep and activity, water-resistance, long battery life and design features such as screen size. The Fitbit Charge 3 was deemed the most appropriate for use in the CREST study.
The people with dementia were asked to wear a 'Fitbit Charge 3' every day and night for a nine week period between December 6th 2019-February 7th 2020 (a total of 64 days and nights). They were told that the purpose of wearing the device was to find out if the Fitbit was acceptable to them and to monitor their sleep quality and exercise patterns. They were not told especially to adapt their habits but as part of their participation in the CREST intervention, they were already taking part in the exercise component.
Prior to the first session of the exercise component, individual Fitbit wearer profiles were created for each person with dementia; consent was obtained from each individual to record details of their height, weight, and date of birth to set up these profiles. The research team also created new Gmail accounts linked to each new wearer profile, with a unique password only shared with that person. This was necessitated by data protection and privacy regulations, to ensure that researchers only had access to communications related to Fitbit data, and not to the participants' personal emails. Permission was also granted by each person with dementia for the research team to keep a record of their Gmail and Fitbit account passwords, to enable them to conduct data extraction of the Fitbit data. The researchers created a 'Fitbit User Manual' for each dyad (a person with dementia and their caregiver) with detailed step by step instructions including screenshots/pictures and large font etc. suitable for the participants.

| Data collection and analysis
Quantitative data relating to wear rates for the Fitbit were recorded. Data were extracted every 2 weeks from each Fitbit device during the study period (a total of four extraction points). The raw exported data was organised and cleaned manually within Excel before analysis was completed. Specific criteria were devised to outline a missing day/night. A 'missing' day was counted if the Fitbit had not been worn for one entire 24-h period, operationally defined as when the Fitbit data export showed '0 steps taken' and '1440 min of sedentary activity' (i.e., 24 complete hours) on any one calendar day. Fitbit devices use triaxial accelerometry to count steps and research has found they are likely to underestimate steps, however, this underestimation is small enough to be deemed acceptable. 12 As overnight data was exported over multiple columns, 'missing nights' were defined as a gap in wear data of one or more consecutive calendar days (i.e., the data showed the Fitbit had been worn on December 18 and again on December 20, leaving a one-night gap on December 19 so December 19 was recorded as a 'missing night').

| Procedure
At a set-up meeting organised with each person with dementia and caregiver, the researcher fitted the person with dementia with the Fitbit and they also gave them a short demonstration on setting up the Fitbit, logging in with the pre-created profile, enabling the Bluetooth connection to allow data synchronisation, how to charge the device and view the data on the device screen and in the mobile phone application. To facilitate data synchronisation, the caregivers were asked to keep Bluetooth enabled on their phone using the 'allday sync' option. Automated feedback settings (e.g., vibrations for reminders or exercise targets) were disabled, to avoid causing undue alarm for the person with dementia.

| RESULTS
Ten dyads were recruited into the CREST intervention. After the first day one of the participants had to stop due to illness. The remaining nine dyads completed the intervention. See Table 1 for participant demographics.

| Challenges around setting up the device
There were some issues with setting up the Fitbit for use by the person with dementia. 2. Setting up the device was time consuming in that only one of the nine caregivers was confident to do this without assistance from the research team; the others required additional meetings/phone calls during the set-up phase. Beyond set-up, follow-up phone calls/text messages were occasionally required with these dyads to ensure the device was still operating properly. Caregiver involvement was essential from the beginning and researcher input in the early phase of the study was underestimated.
3. Technical support needed to be provided throughout the intervention for issues such as slow Wi-Fi connections in the participant's home (which delayed the download of the Fitbit mobile application, or sync and upload of data); two caregivers had oldermodel smartphones which were unable to download the latest version of the mobile application; one caregiver already had a Fitbit (so was unable to sync a second device to the same phone); and a sporadic issue was reported with one Fitbit not syncing to the correct phone due to the presence of other Fitbits in the vicinity during set-up. Finally, some participants did not own a USB charger plug to recharge their Fitbit and so these were supplied to them. O'SULLIVAN ET AL.

| Feasibility of the Fitbit
Only one participant (P9) wore the Fitbit consistently (Table 2). Daily wear rates were inconsistent, with the device worn more frequently during the day than at night. There was, on average, 13 missing days per participant (range: 0-55 days) ( Table 2). Night-time adherence was particularly low, with an average of 21 missing nights per participant (range: 0-59 nights) ( Table 2). More specifically, two of the people with dementia (P4 and P8) recorded no data at all on 59 and 54 nights during the 64-night period (92% and 84% of total nights, respectively).
One of these participants was male and one female. Unfortunately, it was not possible to determine whether the device was malfunctioning during this period, or whether user error had been involved (i.e., they had not worn it, or wore it when the battery had no charge left).
However, in an interview with one of the caregivers following on from the intervention, they explained that in the case of P4 the strap had broken on the Fitbit and so they could not wear it. The optimal feasibility benchmark for adherence in this study had been set at 'wear rates of 70% or above' throughout this period. 10 Based on these criteria, adherence by people with dementia to wearing their Fitbit was confirmed as acceptable during the daytime (79.7%) but not acceptable for night-time (66.7%). However, when the two participants (P4 and P8) were removed from the analysis, the wear rates were 92% and 82.5% for daytime and night-time respectively.

| Qualitative interviews
One group interview, with seven of the people with dementia participating, was conducted halfway through the exercise programme. Individual interviews were held with each of the people with dementia (n = 9) and the caregivers (n = 9) at the end of the CREST intervention. Four key themes were identified: engagement with the Fitbit, important role of the caregiver, impact of the Fitbit on activity, and monitoring sleep. A summary of qualitative themes and example quotes can be seen in Table 3.

| Engagement with the Fitbit
During the interviews most of the people with dementia mentioned that they wore the Fitbit daily as requested and that it was easy to T A B L E 1 Demographic overview of the people with dementia and their caregivers. Sex Female 44% (4) 89% (8) Male 56% (5) 11% (1) Age 60-69 33% (3) 22% (2) 70-79 44% (4) 11% (1) 80-89 22% (2) (5) Other (e.g., professional) 11% (1) 22% (2) Employment status Employed 0 33% (3) Homemaker T A B L E 2 Number days and nights with missing data for each participant. were not interested in it. The remaining 3 dyads did not wish to keep the Fitbit and returned it to the research team for the data to be erased. These dyads were not explicitly asked why they did not wish to keep the Fitbit, but comments during interviews with these caregivers suggested it was due to a lack of engagement by the person with dementia (see Table 3 for supporting quotes).

| Important role of the caregiver
Some of the participants relied on their caregiver for help with using  Table 3 for supporting quotes).

| Impact of the Fitbit on activity
For some of the participants the Fitbit provided a focus for the caregiver to offer encouragement to the person they cared for, both to exercise more often and to view the details on the Fitbit.
Most of these caregivers felt that the person they care for was more motivated to exercise because of this encouragement. Other family members were also a positive source of motivation and family members who also owned a Fitbit were able to share activity levels and experience of wearing it (see Table 3 for supporting quotes).

| Monitoring sleep
While all caregivers reported that they were monitoring the daily activity levels on the Fitbits, only two reported that they were also monitoring the sleep quality of the person with dementia. These caregivers reported that they would check the sleep data of the person they care for to determine an appropriate level of activity for the coming day, or to confirm that the person's estimates of their sleep quality (e.g., whether they had awoken during the night) matched the sleep data on their Fitbit (see Table 3 for supporting quotes).

| DISCUSSION
This study intended to assess the feasibility and acceptability of using  7%). This was largely driven by two participants who missed over 80% of nights during this period. One of these participants had stated in an interview that they routinely wore the device overnight, but this was not corroborated by their device data.
It was not possible to determine the reason for this discrepancy (i.e., device malfunction, user error, or simple lack of wear). Technology malfunctions with wearables have been reported in previous research. 14 For another participant, the strap had broken and so they could not wear it. These findings fit within a pattern of inconsistent nightly wear in the wider literature among the general population. 15 It has been suggested that disruption to the normal nightly routine may at least partially explain the sudden drop-off in adherence at night; for example, if the person typically removes their watch before going to bed. 11 The high levels of support from researchers and caregivers required to ensure that the Fitbit devices were operational presented significant challenges to the overall feasibility during this study. None of the people with dementia owned a smartphone and so the Fitbit was linked with their caregiver's phone. Ongoing technical support was required at the beginning to assist the caregivers with set-up and troubleshooting. Both the lack of smartphones and the generally low technological ability in this population have been reported elsewhere. 1,3 Adoption of technology can be slow in older adults 16 and both this study and previous studies have reinforced the importance of caregiver involvement to provide day-to-day management and support for the device and to facilitate ongoing participation in these trials. 3,15,17 Within CREST, none of the caregivers mentioned feeling burdened by these tasks; however, caregiver involvement has the potential to increase burden, and must be weighed against the potential benefits of using smart wearables with this population when designing future studies. 15 Future studies should consider the caregiver involvement when planning their set-up and assessment, to ensure that caregivers can access technical support as required and minimise their potential burden. 3,15,17 Missing data has been reported as a problem with data collected from smart wearables. 18 In our study high levels of input were required at the data analysis stage: despite best efforts from the caregivers and research team throughout this study, the Fitbit devices were not worn consistently, and the available data were patchy and required additional time and effort to organise. This created a tradeoff between the perceived ease of data collection from smart wearables (continuous, objective, passive, and low effort) versus the extraction and analysis of the data (complex, time-consuming and high effort). Though still arguably less time-intensive than established qualitative methods such as interviews or questionnaires, the use of smart wearables to collect data requires improvements, if they are to give insight into ways to benefit people with dementia. 1 The objective nature of the data collected from smart wearables is, however, only as objective as the processing is valid. 19 The research team considered the purchase of Fitabase data management software (https://www. fitabase.com/) to automatically extract and collate the Fitbit data.

T A B L E 3 (Continued)
However, the significant cost of purchase was not deemed justifiable given the limited amount of data from the small sample, as manual cleaning by the research team was significantly cheaper. It is worth noting that the expense of automated data management software may benefit other studies with a larger sample, however, the cost could be prohibitive in resource limited settings and there should be cognisance of the research data being treated as a commodity.  11 as participants in that study suggested that this made the device more convenient to wear and subsequently more acceptable.
Overall, however, only a few CREST participants engaged with the features on the Fitbit, and this interaction mostly came after prompting from their caregiver rather than from their own initiative.
The people with dementia were not asked to rate their satisfaction with the Fitbit, but were asked about the device in their interviews, though the majority of comments were neutral. Perhaps the truest endorsement came at the conclusion of the intervention, when fewer than half (44%) of the people with dementia elected to continue wearing their Fitbit, with one noting they 'couldn't manage it at all' [P9]. This is considerably lower than retention rates in a group of older adults, wherein 68% would wear a similar device again. 14 It indicates that there may be limited benefit of using the Fitbit for personal use among people with dementia, but possible benefit for its use for research purposes as indicated by the acceptable adherence rate of wearing the device during the day, throughout the study.

| Limitations
One limitation of this study is that only one model was available to participants (Fitbit Charge 3). Though this model was deemed suitable due to its large screen size and robustness, a plethora of alternatives have been tested previously, with disparate results 3,9,14  have been considered to disentangle whether the challenges of using the Fitbit were related to cognitive challenges or unfamiliarity with the technology. Finally, the sample was small (n = 9) as they had been recruited for a pilot intervention (of which Fitbit was one element).
The data from the current study indicate that the perceived value of smart wearables remains somewhat limited in this population; for example, a typically sedentary group with lower levels of physical activity and lower levels of technical prowess than the typical owner of a smart wearable. 1 One common limitation of the O'SULLIVAN ET AL. existing studies is that all have assessed feasibility using commercially available smart wearables, often in conjunction with standard data collection methods such as questionnaires or interviews to cover gaps in the ability of the devices to track areas of interest. The next step in ensuring the feasibility of smart wearables in dementia research should ideally involve partnerships between academics, people with dementia, their caregivers, and device manufacturers, to develop dedicated wearable devices capable of monitoring behaviours specifically for people with dementia, which fit their own needs and wants. 8,9 There may be different features of the device which are pertinent to a particular cohort depending on the purpose such as monitoring sleep or physical activity or instigating behaviour change.
In addition, consideration should be given to the minimum wear time required to derive reliable estimates of the activity of interest. 20,21 Wearable devices are available whereby algorithms help determine accurate estimation of time spent in sedentary and active behaviours and these should be considered in future studies. 22

| CONCLUSION
Overall, adherence by people with dementia to wearing the Fitbit during the day was acceptable and caregivers were willing to support the use of the Fitbit during the study. However, there was little engagement with it by the people with dementia with fewer than half wanting to keep their Fitbit beyond the intervention and few engaged with the Fitbit features primarily only using it to check the time, showing that it had little added value to them. Our findings also suggest that significant researcher support and caregiver input is needed in studies using smart wearables among people with dementia. We conclude that there may be some utility in using the Fitbit device for research purposes, particularly for monitoring daytime activity.