JMIR Publications

ABSTRACT

Background:

Physical activity trackers such as the Fitbit can allow clinicians to monitor the recovery of their patients following surgery. An important issue when analysing activity tracker data is to determine patients’ daily compliance with wearing their assigned device, following selection of an appropriate criterion to determine a valid day of wear. However, it is currently unclear as to how different compliance criteria impact the retention of data acquired from surgery patients during their recovery period. Investigating this issue can help to inform the use of activity data by revealing factors that may differentially impact compliance calculations.

Objective:

This study aimed to understand the impact of applying five different compliance criteria to the activity tracking data of ambulatory surgery patients. We further sought to investigate causes of variation in compliance outcomes through analysis of patients’ tracked data.

Methods:

62 patients who were scheduled to undergo total knee arthroplasty volunteered to wear a commercial Fitbit activity tracker over an 8-week perioperative period. Patients were asked to wear the Fitbit daily, beginning 2 weeks prior to their surgery and ending 6 weeks post-surgery. Of the 62 patients who enrolled in the study, 20 provided Fitbit data and underwent successful surgery. The Fitbit data was analysed using five different daily compliance criteria. These criteria consider patients as compliant with daily tracking if they either: registered >0 steps in a day; registered >500 steps in a day; registered at least one step in 10 different hours of the day; registered >0 steps in three distinct time windows; or registered >0 steps in 3 out of 4 six hour windows. The criteria were compared in terms of compliance outcomes produced for each patient. Data was interrogated using heatmaps and line graphs to identify causes of variation across the sample.

Results:

The five compliance definitions produce different outcomes when applied to the patients’ data, with a 24% difference in Mean compliance between the most lenient and strictest criteria. However, the extent to which individual patients’ compliance was affected by different criteria was not uniform. Some patients’ compliance was relatively unaffected whereas others varied by up to 72%. Wearing the activity tracker as a clip-on device, rather than on the wrist, was associated with greater differences between compliance calculations at the individual level (p = .004, r = 0.616). Age and gender were not associated with differences in individual compliance outcomes. Inspection of patients’ data suggested that changes in tracking routines are an additional source of variation between compliance outcomes. Finally, the analysis revealed the impact of surgery on patients’ compliance, with noticeable reductions in activity following surgery that impact patients’ compliance under stricter criteria.

Conclusions:

This study suggests that different compliance criteria cannot be used interchangeably to analyse activity data provided by surgery patients. The study also reveals the deleterious impact of surgery on patients’ assumed compliance under certain criteria. Patients’ reduced mobility post-surgery implies that the use of more lenient compliance criteria, such as >0 Steps or windowed approaches, can help to account for temporary mobility impairments while also tracking wear over the course of a day. Encouraging patients to wear the device at their wrist, and using secondary sources of data as ground truth, may also help to improve compliance by ensuring activity is detected and increasing actual wear time.