Comparison of raw accelerometry data from ActiGraph, Apple Watch, Garmin, and Fitbit using a mechanical shaker table.

Area of Science:

• Biomedical Engineering
• Wearable Technology
• Human Movement Analysis

Background:

• Wearable devices offer promising avenues for objective physical activity monitoring.
• Evaluating the accuracy of raw accelerometry data from consumer wearables is crucial for reliable health research.
• Research-grade accelerometers are often compared against mechanical devices for validation.

Purpose of the Study:

• To assess the reliability and validity of raw accelerometry output from consumer and research-grade wearable devices.
• To compare device data against a calibrated mechanical shaker table as a reference.
• To determine the suitability of raw accelerometry data for movement estimation.

Main Methods:

• Compared raw accelerometry data from 40 devices (ActiGraph, Apple Watch, Garmin Vivoactive, Fitbit Sense) to a mechanical shaker table.
• Tested inter-device reliability using intraclass correlation coefficients (ICC).
• Assessed validity using Pearson correlation, Lin's concordance correlation coefficient (CCC), absolute error, mean bias, and equivalence testing.

Main Results:

• All tested devices demonstrated high reliability (ICCs ≥ 0.88).
• ActiGraph, Apple, and Fitbit showed excellent concordance (CCCs ≥ 0.83) with the reference, while Garmin showed moderate concordance (CCC = 0.59).
• ActiGraph, Apple, and Fitbit had similar absolute errors, while Garmin exhibited a higher error; no devices met statistical equivalence to the reference.

Conclusions:

• Raw accelerometry data from ActiGraph, Apple, Garmin, and Fitbit devices are reliable for estimating movement.
• While concordance is high for ActiGraph, Apple, and Fitbit, none of the devices achieved statistical equivalence to the reference shaker.
• Future research should focus on device-agnostic and harmonization methods for raw accelerometry data from wearables.