Table 1.

Methodological factors which can affect PA metrics derived from the Actigraph GT3X during device initialization, study conduct, and data processing.

Implementation phase	Methodological factors	Description
Device Initialization	Sampling frequency	The Actigraph measures acceleration a set number of times per second (i.e., the sampling frequency), generating raw data in three axes, which correspond to movement along the body's longitudinal, sagittal, and frontal axes. The sampling frequency is selected before the device is given to the participant (16).
Study Conduct	Sensor placement	The Actigraph can be worn on either hip, wrist, or ankle. The locations do not produce equivalent data, and data processing methods are generally validated with respect to a specific sensor placement (17, 18).
Data Processing
Filter data	Filter	During processing, a bandpass filter is applied to the raw data to remove non-physiological movement artifacts. Two settings are available: a standard filter for healthy gait and the low-frequency extension (LFE) for populations with impaired gait. The LFE is designed to be more sensitive to light, slow movements (16).
Aggregate into epochs	Epoch length	Filtered data is aggregated into epochs, which can be seconds, minutes, or longer in length. The resulting data is expressed in terms of the number of activity counts or steps which occurred during each epoch. Many algorithms are designed to be used with specific epoch lengths (16).
Identify non-wear time	Wear time algorithm Inactivity period	Data is then further processed through wear time algorithms to differentiate between inactive periods when the wearer was sedentary, and those when the device was not worn. This typically consists of identifying extended periods with zero activity counts, with or without a spike tolerance (19).
Remove invalid days	Minimum daily wear time	A minimum daily wear time is typically specified to ensure that PA metrics are valid representations of the wearer's daily activity. Days which did not reach this minimum are usually considered invalid and removed from the dataset (16).
Remove invalid measurements	Minimum valid days Minimum weekdays vs. weekend days	Daily PA metrics are frequently averaged over the course of a week to account for day-to-day variability. A minimum number of wear days is typically specified to ensure measures are not confounded by this variability. Studies sometimes require that measurements do or do not contain a weekend day, as activity can differ from weekdays (20).
Calculate physical activity metrics	Data source Algorithms, cutpoints, and metric definitions	Algorithms and cutpoints (i.e., limits used to differentiate between PA intensity levels) are applied to the data to calculate PA metrics. These methods may use raw or epoch-level data, vertical axis activity counts, vector magnitude counts, step counts, or other data sources to calculate PA metrics. Further aggregation and processing are then applied to calculate outcome measures, such as average daily metrics (16). Researchers may opt to adjust PA metrics for wear time if wear time is contributing to intra- and inter-participant variability (21).