Part I: EPapers

5 The limitations of using commercial wearable activity trackers, such as FitBits, for the clinical monitoring of patient activity levels

Abstract

Objective There is increasing interest in remote monitoring of patients within the comfort and safety of their homes or care homes and became more pertinent during the COVID-19 pandemic to reduce hospital footfall and staff risk. While specifically designed medical devices exist, commercial wearable activity trackers (WAT), such as FitBits, are cheap, easy to use, and patients may already use them for lifestyle advice so their value in clinical intervention is of interest.

The feasibility of using commercial WAT for daily monitoring within a tertiary oncology centre was investigated, including limitations of non-medical devices, such as data collection and synchronisation errors.

Methods Participants were recruited for a study that investigated if remote monitoring of step counts was feasible and acceptable. Patients with advanced lung, upper and lower gastrointestinal cancer, or mesothelioma who were starting a new line of systemic anti-cancer treatment were recruited between December 2020 and December 2021.

Once recruited, participants were provided with a FitBit Inspire HR or Inspire 2 and asked to wear it every day for a 16-week monitoring period. Pseudo-anonymous accounts were created to register the FitBits without sharing patient identifiable data and the devices were set up to automatically synchronise data to the cloud-based platform, Fitabase, via their smartphone.

Steps were monitored on every workday and the ability to record heart rate was used as a proxy marker for compliance as it confirmed that the device was being worn. A day was considered complaint if the device was worn for >70% of waking hours, assumed for purpose of trial to be 7am to 10pm.

The manufacturer or age of the participant’s smartphone was not recorded. Previous discussions with FitBit regarding synchronisation issues had highlighted potential clashes with other Bluetooth devices preventing automatic synchronisation so use of other such devices was documented.

Results Forty-seven patients were recruited and 43 were eligible for ongoing monitoring. Average age was 66 (SD 9) and majority were men (72%). Twenty-nine patients completed the maximum 112 days of monitoring.

Patients were eligible for monitoring on 3855 days. Of these, synchronisation errors occurred on 482 days (13%) and all data from the previous 24 hours was missing on 275 days (7%) due to synchronisation not occurring on the day on monitoring. Only 5 (11%) of participants did not have synchronisation errors during their monitoring period. The median number of synchronisation errors per patient was 8 and maximum of 49, which accounted for 64% of that participant’s monitored days. One participant was withdrawn due to 100% synchronisation error over the first seven monitored days.

Twenty-two participants (47%) used other Bluetooth devices but there was no correlation between their use and synchronisation errors (r=-0.32), nor significant difference in synchronisation error rate (p=0.08).

562 days (15%) were considered non-compliant as heart rate was documented for less than 70% of the waking hour period. When synchronisation errors were removed, however, only 216 days (7%) were truly non-compliant due to the patient not wearing the device, rather than not having access to the data.

Conclusion This study has revealed a potential limitation of using commercial wearable activity trackers, such as FitBits, for clinical monitoring. While compliance with monitoring was good and matched previous reports on compliance at over 80%, the loss of data due to synchronisation errors reduced perceived compliance and, importantly for clinical interventions, reduced data available for immediate action.

Correcting these issues and restarting the automatic synchronisation was not a complex procedure but did necessitate a telephone call with the participant to manually synchronise the device, restart their smartphone or occasionally reinstall the app, which added to the participant burden of the investigation and overwhelmed the technological abilities of some participants. Currently, it is not clear what causes these synchronisation errors and, therefore, it is not possible to select patients who would be more suitable for this intervention.

The frequency of synchronisation errors mean that it is not feasible to use commercially available WAT for remote monitoring of patients and caution is needed if the results are used to guide clinical intervention, rather than simply offer lifestyle advice.

Article metrics
Altmetric data not available for this article.
Dimensionsopen-url