Sensitivity to Change of Objectively-Derived Measures of Sedentary Behavior

Abstract

The aim of this study was to examine the sensitivity to change of measures of sedentary behavior derived from body worn sensors in different intervention designs. Results from two intervention studies: Stand up for Your Health (pre-post home-based study with older adults not in paid employment) and Stand Up Comcare (non-randomized controlled trial in the workplace) were analyzed to quantify sensitivity to change of measures of total and accumulation of sedentary time obtained from hip-worn Actigraph and thigh-worn activPAL monitors. Sensitivity to change varied with intervention design and population considered. The activPAL was generally more sensitive but not consistently for all measures and designs. Measures of sedentary time accumulation, in particular half-life bout duration (W 50% ), were consistently more sensitive than total sedentary time. Measurement devices used in intervention studies need to be appropriately selected to be sensitive to changes in the behavioral target. For sedentary behavior interventions, measures of accumulation should be considered as outcomes.

Evaluating the effectiveness of organisational-level strategies with or without an activity tracker to reduce office workers’ sitting time: a cluster-randomised trial

“…The other activity outcomes, assessed during work hours and overall were: the average time per day spent in prolonged sitting bouts (sitting time accrued in continuous bouts of 30 min or more), standing, and stepping; the number of steps per day; and, the average time period between sitting bouts. This latter measure is a sensitive and responsive metric [35] of sitting or sedentary time accumulation.…”

“…Total time or steps per day were calculated then averaged over the valid days, and normalised to a 16-h waking day or 10-h workday (which were about average for this sample). Average time between sitting bouts was calculated as the mean duration of the upright periods between sitting bouts, using the maximum likelihood estimate of the mean for a log-normal distribution [35]. …”

An Evaluation of Accelerometer-derived Metrics to Assess Daily Behavioral Patterns

“…Next, we used conditional logistic regression to test whether the metrics were significantly associated with study condition, after adjusting for the main volume measures, total sedentary or MVPA time (min/day). This analysis assumes the experimental manipulations also changed the underlying daily behavioral patterns (i.e., frequency x duration of behavior) as well as the total volume of sedentary time and MVPA (8). We were therefore testing a broad range of sedenatary and physical activity pattern metrics for their ability to detect patterns of behavior associated with increasing or decreasing exercise, non-exercise activities and/or sedentary time, after accounting for total volume.…”

“…Given the dense and multidimensional data within a day of accelerometer data, there is a large, and possibly infinite number of metrics that could be extracted from accelerometer data to describe the interrelated components (e.g., frequency, duration, and intensity) of physical behavior (8). Together these components characterize the total volume of daily activity and inactivity (8, 43) and throughout this manuscript we will use the term daily behavioral patterns to refer to these components. Descriptive studies have used metrics that may reflect daily behavioral patterns, including time spent active or sedentary within various bout lengths (e.g., 1,5,10, 20 min) (3, 12, 20, 41).…”

On chronic cancer-related fatigue and physical behavior

“…Different adjustments are prone to fix that issue: on the reasoning perspective by allowing user input on the sensor 162 placement or type of activity, which would require validation studies, and clear instructions of the user. From a hardware perspective, the addition of a goniometer, which would require sensor placement on the upper leg such as used in the ActivPal, such as in Chastin et al (2015) [5], would also provide the input that is needed to estimate physical activity more validly for different types of behavior.…”