Wrist-worn accelerometers: Influence of decisions during data collection and processing: A cross-sectional study

Area of Science:

• Wearable technology and biomechanics
• Physical activity measurement
• Data science and analytics

Background:

• Objective physical activity assessment using accelerometers requires numerous data collection, processing, and interpretation decisions.
• The impact of these methodological choices on study outcomes is often underreported, hindering reproducibility.
• This study investigates how different data processing decisions influence the measurement of moderate-to-vigorous physical activity (MVPA), light physical activity (LPA), and inactivity.

Purpose of the Study:

• To quantify the influence of specific data processing decisions on accelerometer-derived physical activity metrics.
• To compare the effects of different cut-point criteria and epoch lengths on daily MVPA, LPA, and inactivity.
• To highlight the importance of transparent reporting in accelerometer-based research.

Main Methods:

• 156 participants wore wrist-worn ActiGraph wGT3X-BT accelerometers for 7 days.
• Data processed using the R-package GGIR, comparing default settings with literature-based options.
• Paired t-tests used to analyze differences in daily MVPA, LPA, and inactivity.

Main Results:

• Comparing MVPA cut-points showed minimal differences (0.4 min/day).
• Excluding bout criteria doubled MVPA estimates compared to bouted data.
• Changing epoch length from 5s to 1s significantly altered MVPA (-6.6 min/day), inactivity (-22 min/day), and LPA (+28 min/day).

Conclusions:

• Methodological decisions in processing wrist-worn accelerometer data significantly affect physical activity and inactivity outcomes.
• Variations in settings, such as epoch length, can lead to substantial differences in reported activity levels.
• Researchers are urged to report all data collection, processing, and interpretation decisions to enhance study transparency and comparability.