Influence of epoch length on measurement of light physical activity in the elderly: A technical analysis

Aims: To clarify how epoch lengths of accelerometers affect measurement of physical activity in elderly people. Methods: The data was based on 70 elderly people (72.6 ± 5.4 years) living in Japan between 2017 and 2018. Furthermore, we used data obtained from triaxial accelerometers that the subjects wore for more than 10 hours every day for 7 days. We evaluated light physical activity (2.9 Mets or less) and further grouped it into sedentary behaviour (SB, 1.5 Mets or less) and Light Intensity Physical Activity (LIPA, between 1.6 and 2.9 Mets). We also compared 10-second epoch lengths (ELs) to 60-second ELs (%) by Bayesian estimation. Results: In 2017 and 2018, SB at 10-second ELs was longer than at 60-second ELs (55.5 vs. 50.4% in 2017; 55.7 vs. 48.9% in 2018); however, LIPA at 10-second ELs was shorter than at 60-second ELs (35.0 vs. 42.3% in 2017; 35.0 vs. 44.5% in 2018). The Bayesian factor varied between 3.0x10 12 and 3.2x10 24 . The robustness of the Bayesian factor was confirmed by the robustness check. Furthermore, effect sizes were between |1.68| and |3.00|. Conclusion: ELs of the accelerometer may affect measurement of physical activity in elderly people. Thus, SB at 10-second ELs may be longer than at 60-second ELs and may be the reverse for LIPA.