Module 1: Accelerometer Devices and Studies

Materials

Site is rendered at https://jhuwit.github.io/wearabler/

Materials are at: https://github.com/jhuwit/wearabler

Can use git clone or download via: https://github.com/jhuwit/wearabler/archive/refs/heads/main.zip

Why Measure Activity?

• Physical activity important component of health and contributes to assessing and predicting mortality risk (Schoenborn and Stommel 2011)

• Objective assessment in free-living settings - more accurate than self-report (Prince et al. 2008).

• Shown as strongest predictor of mortality (Tabacu et al. 2020; Leroux et al. 2024)

Goals

Main Question

What does an accelerometer measure?

• Piezoelectric sensor - material generates an electric charge from mechanical stress/movement
• Capacitive micro-electro-mechanical system (MEMS) - mobile spring suspended mass with multiple capacitance plates

The Device Landscape

Commercial wearables (watches and fitness trackers) - designed for consumer experience.

• Proprietary step, sleep, or activity scores
• Raw signal may be unavailable and processing rules may change over time.

Research grade - for studies

• High-frequency raw acceleration and device metadata
• Trade off - decide on calibration, QC, and alternative processing

Example Studies

• NHANES Accelerometry (CDC) (2003-2006, 2011-2014)
• Large-scale population studies using wrist-worn devices
• Released raw data - we released processed version
• UK Biobank Wearable Data
• Over 100,000 participants with 7-day wrist accelerometer recordings
• All of US uses FitBit

Accelerometers Covered

• ActiGraph wGT3X+ (pictured)/GT9X: .gt3x extension (NHANES)
• Axivity AX3: .cwa file (UKBiobank)

This is a research-grade, tri-axial accelerometer—not a consumer step counter.

ActiGraph wGT3X-BT. Image: Ametris.

Axivity AX3. Image: Axivity.

What is Raw Accelerometer Data?

• Sensor that measures acceleration in 3D space: X/Y/Z
• Typically measured in g-forces (g, \(g=9.81m/s^2\)) or m/s²
• Sampled at regular(ish) time intervals (e.g., 30Hz, 100Hz)

Timestamp                  X        Y        Z
2025-02-17 12:00:00.0125   0.02     0.98     0.03
2025-02-17 12:00:00.0250   0.04     0.95     0.05
2025-02-17 12:00:00.0375  -0.01     0.99     0.02

Wrist-Worn vs. Hip-Worn Accelerometers

Feature	Wrist-Worn Devices	Hip-Worn Devices
Placement	Worn on wrist	Attached to belt
Compliance	Higher	Lower
Activity Types	Captures arm movements	Better for whole-body movement

Data Resolution

• Raw Accelerometer Data (high-resolution time-series)
  • 20-100+ samples/second (Hz)
• Wear Time Detection (non-wear vs. wear periods)
• Minute Level
  • Activity Counts, Activity Index, ENMO, MAD
  • Steps (estimation of a “step”)
• Sleep estimation

Course Learning Goals

• Read in accelerometry data
• Visualize the data
• Calculate minute-level summaries
• Calculate non-wear and inclusion
• Create daily summaries
• Discuss functional approaches

Packages

This work is based off the activerse, which is maintained by Johns Hopkins Wearable and Implantable Technology (WIT) group, with code at https://github.com/jhuwit.

If you can’t find a package, try pak::pkg_install("jhuwit/PACKAGE"), such as

• pak::pkg_install("jhuwit/asleep")
• pak::pkg_install("jhuwit/activerse")

Modules

https://jhuwit.github.io/wearabler/

References

Leroux, Andrew, Erjia Cui, Ekaterina Smirnova, John Muschelli, Jennifer A Schrack, and Ciprian M Crainiceanu. 2024. “NHANES 2011–2014: Objective Physical Activity Is the Strongest Predictor of All-Cause Mortality.” Medicine and Science in Sports and Exercise 56 (10): 1926.

Prince, Stephanie A, Kristi B Adamo, Meghan Hamel, Jill Hardt, Sarah Connor Gorber, and Mark Tremblay. 2008. “A Comparison of Direct Versus Self-report Measures for Assessing Physical Activity in Adults: A Systematic Review.” International Journal of Behavioral Nutrition and Physical Activity 5 (1): 56. https://doi.org/10.1186/1479-5868-5-56.

Schoenborn, Charlotte A., and Manfred Stommel. 2011. “Adherence to the 2008 Adult Physical Activity Guidelines and Mortality Risk.” American Journal of Preventive Medicine 40 (5): 514–21. https://doi.org/10.1016/j.amepre.2010.12.029.

Tabacu, Lucia, Mark Ledbetter, Andrew Leroux, Ciprian Crainiceanu, and Ekaterina Smirnova. 2020. “Quantifying the Varying Predictive Value of Physical Activity Measures Obtained from Wearable Accelerometers on All-Cause Mortality over Short to Medium Time Horizons in NHANES 2003–2006.” Sensors 21 (1): 4.