Concurrent validation of activity monitor based step-counts in people with Polymyalgia Rheumatica

Background: Physical activity is both a therapeutic intervention and an important outcome in people with rheumatic and musculoskeletal disease (RMDs). Subjective methods are prone to response bias and underestimate low-intensity activities, which are common in older adults. Activity monitors may provide an objective solution by measuring step-count, but limited previous research has identified reduced accuracy in people with RMDs. This study aimed to evaluate the concurrent validity of the ActiGraph with both standard and Low Frequency Extension (LFE) algorithms; Fitbit-Zip in both recommended positions; and explore which factors influence device error in people with polymyalgia rheumatica (PMR). Methods: The ActiGraph (±LFE), Fitbit-Zip (waist) and Fitbit-Zip (shirt) were concurrently evaluated using a two-minute-walk-test (2MWT) and stairs test in people with PMR (>1 month glucocorticoid treatment). Participants walked at their normal pace and with any assistive devices required. A GAITRite instrumented walkway collected temporal and spatial gait parameters. Gold standard step-counts were derived from a mean of three manual counts taken from half speed video replays by two investigators. Bland-Altman plots describing the mean bias (95% confidence intervals (CI)) and 95% limits of agreement (LoA) were used to determine monitor accuracy. Spearman’s rho described association of step-count differences with demographic, clinical, and gait parameters. Results: Data was collected on 27 PMR patients (24 female; mean (SD) age 69.2 (8.8); BMI 28.3 (5.6) kg/m2; median (IQR) steroid dose 9.0 (5.0, 12.5) mg/day; Health Assessment Questionnaire (HAQ) 0.63 (0.13, 1.25); FACIT-F 38 (33, 45)). The median (IQR) gold standard step-counts were; 2MWT 206 (197, 220); stairs 15 (15, 16). All monitors systematically underestimated 2MWT step-counts. The Fitbit-Zip (waist) achieved closest agreement to the gold standard step-count in the 2MWT (mean (95%CI) bias 10 (-3, 23), 95%LoA -55, 74). Conversely, the ActiGraph (LFE) performed best during the stairs test (mean (95%CI) bias 0 (-1, 1), 95%LoA -5, 5 followed closely by the Fitbit-Zip (waist) (mean (95%CI) bias 1 (-1, 3), 95%LoA -8, 10). The LoA were particularly influenced by 2 outliers: one participant had a severe level of impairment (HAQ=2.13), which affected all accelerometers during both tests. The second was a Fitbit-Zip (waist) measurement on the stairs test, which was out by an order of magnitude, possibly due to operator error. On all devices, step count difference correlated negatively with gait velocity (rho=-.37 to -.58) and stride length (rho=-.31 to -.72) and positively with HAQ scores (rho=.32 to .57) and duration of double-limb support phase (rho=.33 to .61). Conclusion: Close agreement with the gold standard step-counts suggests the Fitbit-Zip (waist) is a reliable measure of physical activity in people with PMR receiving treatment. The ActiGraph however, cannot be recommended without activation of LFE filter. Device performance reduces in participants with higher levels of functional impairment