Measuring Time in the Office Using Bluetooth Sensors: Feasibility and Validity Considerations

in Journal for the Measurement of Physical Behaviour
Restricted access

Purchase article

USD $24.95

Student 1 year subscription

USD $37.00

1 year subscription

USD $50.00

Student 2 year subscription

USD $71.00

2 year subscription

USD $93.00

The office is a key setting for intervening to reduce sitting, therefore office-specific activity measures are needed to evaluate interventions. We tested whether valid measures of office time and office-specific activities could be obtained using Bluetooth sensing with a variety of sampling intervals, receiver wear positions, and beacon placements. Workers from one building (n = 29, 72% female, age 23–68 years) wore, for one workday, the activPAL3 on the thigh (measured sitting, standing and stepping) and the Bluetooth-enabled ActiGraph Link on the wrist and thigh. Location (office/not) was estimated by Bluetooth signal presence/absence at two beacons in the wearer’s office (desk, wall), with chest-worn video cameras as the criterion. Accuracy in location classification was assessed and compared across 60-s, 30-s, and 10-s sampling intervals. The validity of Bluetooth-derived measures of total time in the office and in office-specific activities was assessed. For both the wrist and thigh-worn Link, with various beacon placements, accurate classification of location (office/not) was obtained, with a significant (p < .05) but trivial difference in accuracy across sampling interval options (F scores all ≈ .98). With the 60-s sampling interval, mean absolute percent error was very small for office time and office sitting time (<5%), but higher for infrequent activities: standing (17%–23%), incidental stepping (30%–49%), and purposeful walking (57%–86%). The ActiGraph Link can be used to validly measure office time and office location of activity with a 60-s Bluetooth setting. Higher resolution improves accuracy but not to a meaningful degree.

Clark, Hadgraft, and Winkler are with School of Public Health, The University of Queensland, Herston, Queensland, Australia. Hadgraft and Sugiyama are with the Swinburne University of Technology, Melbourne, Victoria, Australia. Hadgraft is also with the Baker Heart and Diabetes Institute, Melbourne, Victoria, Australia. Sugiyama is also with the Australian Catholic University, Melbourne, Victoria, Australia.

Clark (b.clark3@uq.edu.au) is corresponding author.

Supplementary Materials

  • Supplementary Table 1 (PDF 100 KB)
  • Supplementary Table 2 (PDF 91 KB)
  • Supplementary Table 3 (PDF 141 KB)
  • Supplementary Table 4 (PDF 107 KB)