Measurement of physical activity in clinical practice using accelerometers

Abstract

Accelerometers are commonly used in clinical and epidemiological research for more detailed measures of physical activity and to target the limitations of self‐report methods. Sensors are attached at the hip, wrist and thigh, and the acceleration data are processed and calibrated in different ways to determine activity intensity, body position and/or activity type. Simple linear modelling can be used to assess activity intensity from hip and thigh data, whilst more advanced machine‐learning modelling is to prefer for the wrist. The thigh position is most optimal to assess body position and activity type using machine‐learning modelling. Frequency filtering and measurement resolution needs to be considered for correct assessment of activity intensity. Simple physical activity measures and statistical methods are mostly used to investigate relationship with health, but do not take advantage of all information provided by accelerometers and do not consider all components of the physical activity behaviour and their interrelationships. More advanced statistical methods are suggested that analyse patterns of multiple measures of physical activity to demonstrate stronger and more specific relationships with health. However, evaluations of accelerometer methods show considerable measurement errors, especially at individual level, which interferes with their use in clinical research and practice. Therefore, better objective methods are needed with improved data processing and calibration techniques, exploring both simple linear and machine‐learning alternatives. Development and implementation of accelerometer methods into clinical research and practice requires interdisciplinary collaboration to cover all aspects contributing to useful and accurate measures of physical activity behaviours related to health.

Stronger Association between High Intensity Physical Activity and Cardiometabolic Health with Improved Assessment of the Full Intensity Range Using Accelerometry

“…As objective measurement of physical activity (PA) have become widespread in epidemiological and clinical studies, a lot of work has been put into the development of these methods. In general, the methods consist of the following steps: collection of raw data, processing to useful metrics, calibration to represent PA and reduction to output variables to be used in further analysis [1]. The most common way of doing this is to capture raw data from an accelerometer, process this data to ActiGraph (AG) counts, calibrate these counts to energy expenditure and reduce this to the number of minutes at moderate-to-vigorous intensity (MVPA).…”

“…A main reason of PA assessment in research and clinical practice is its relation to mortality and cardiometabolic health [1,8]. Ultimately, each methodology has to be assessed by its relation to clinical variables and outcomes.…”

Discrimination of wear and non-wear in infants using data from hip- and ankle-worn devices

“…However, the resolution of the criterion measure of device wear available in this study was recorded to the nearest minute making it most appropriate to work the sensor data in 60-s windows. Migueles et al [ 31 ] highlights that it is generally an acceptable practice to collapse raw acceleration data to 60-s windows as this window length is considered sufficient resolution for many PA assessment applications as it makes for more simplified data handling, analysis, and interpretation [ 32 ]. Trost et al [ 33 ] has also noted previously that for classifying activity type in children, markedly greater agreement could be achieved when using a 60-s window length (88.4%) compared to using a shorter 10-s window length (81.3%) with machine learning methods.…”

“…While true that valuable information pertaining to fine movements exists in the raw sensor signal and is useful for tasks such as activity classification, working with the raw signal is not always the most practical as it is susceptible to signal noise not due to movement as well as device calibration concerns. Furthermore, some applications do not benefit from using the raw sensor signal which can complicate analyses if the statistical method chosen cannot sufficiently leverage the information available in the signal [ 32 ]. The study methods presented are intended to quantify gross device wear and non-wear time as a primary data cleaning step.…”

Measurement of sedentary time and physical activity in rheumatoid arthritis: an ActiGraph and activPAL™ validation study

“…Currently, several processing methods can be applied to raw accelerometer data, with the dominant approach being the use of thresholds or 'cut-points' that classify behaviour as sedentary, light-intensity PA (LPA), moderate-intensity PA (MPA) or vigorous-intensity PA. There is an absence of a consensus on the 'best' method, with this decision dependent on the research question, study resources and research team expertise [10].…”