The validity and reliability of consumer-grade activity trackers in older, community-dwelling adults: A systematic review
- Publication Type:
- Journal Article
- Maturitas, 2018, 112 pp. 85 - 93
- Issue Date:
© 2018 Elsevier B.V. Objective: To understand the validity and reliability of consumer-grade activity trackers (consumer wearables) in older, community-dwelling adults. Methods: A systematic review of studies involving adults aged over 65 years who underwent physical activity monitoring with consumer wearables. A total of 7 observational studies qualified, identified from electronic databases: MEDLINE, EMBASE, Cochrane Library and others (2014–2018). Validity was interpreted using correlation coefficients (CC) and percentage error for agreement between reference devices or gold-standard validation methods Reliability was compared using mean differences or ranges (under- or overestimation) of step count and activity time. Results: Total sample size was 290 adults, mean age of 70.2 ± 4.8 years and females constituting 46.7 ± 26.1%. The studies evaluated eight different consumer wearables used by community-dwelling adults with a range of co-morbidities. Daily step count for all consumer wearables correlated highly with validation criterion, especially the ActiGraph device: intraclass correlation coefficients (ICC) were 0.94 for Fitbit One, 0.94 for Zip, 0.86 for Charge HR and 0.96 for Misfit Shine. Slower walking pace and impaired ambulation reduced the levels of agreement. Daily step count captured by Fitbit Zip was on average 7117 (±5,880.6), which was overestimated by five of the eight consumer wearables compared with reference devices (range 167.6–2,690.3 steps/day). Measurement of activity duration was accurate compared with reference devices, yet less so than step count. Conclusion: In older, community-dwelling adults, consumer wearables accurately measure step count and activity duration, as confirmed by reference devices and validation methods Further research is required to understand how co-morbidities, gait and activity levels interact with monitoring in free-living environments.
Please use this identifier to cite or link to this item: