We investigated the combined influences of accelerometer processing parameters across a range of epoch lengths, cut-points and bout durations on time spent in total and bout-accumulated MVPA in adults. Our findings highlight the need to make informed decisions concerning the processing and interpretation of accelerometer data. The regression models reveal how time spent in total and bout-accumulated MVPA are affected by epoch length, cut-point and bout duration.
The standardisation of accelerometer-derived PA measurement is a challenge within the field of objective monitoring. We have demonstrated that the data collection and processing decisions contribute, both as independent and interacting factors, to the estimated amount of time spent in MVPA within a given population. Our analyses provide insight into how participants tend to accumulate their activity. For example, if total and bout-accumulated time spent in MVPA is very different using shorter epochs, as is the case in this current investigation, it can be concluded that participants accumulate their MVPA in a relatively sporadic nature. We therefore recommend that these analyses are conducted for other populations, providing information on the effects of these parameters.
Epoch length and bout duration were found to interact; there is a reverse association for epoch length between total and bout-accumulated MVPA. Longer epochs result in greater accumulation of MVPA in bouts due to a buffering effect, which results in a reduced likelihood of bout termination by short interruptions. Conversely, shorter epochs are more likely to end bouts of MVPA. The effect of longer epochs on bout-accumulated time spent in MVPA is attenuated by the use of higher cut-point thresholds or longer bout durations. We used a continuous bout definition in the present investigation and did not include the concept of bout interruption allowances [15, 20]. It is a matter of definition if for example a 21-min segment of MVPA with an interruption of 1 min in the middle is considered a single bout of 20 mins or two bouts of 10 mins.
Gabriel et al.  found more than 15 mins difference in total time spent in MVPA between 60-s and 10-s epochs in female adults which is comparable with the results of this study in men and women. Caution must therefore be given to comparisons between adult PA investigations using accelerometry without adjustments for epoch length, as these differences are likely to impact on adherence to PA guidelines and associations with clinical outcomes. As suggested by Gabriel et al. , intermittent MVPA is likely to be misclassified as light intensity when summed over a full minute. It is therefore not only the sporadic nature of children’s activity  but also the intermittent activity of adults which makes epoch length a significant factor on the interpretation of accelerometer data.
Nilsson et al.  investigated the role of epoch length in children, demonstrating that the magnitude of the epoch effect is amplified among individuals with higher activity levels. In children, this was predominantly seen in the vigorous and very vigorous intensity categories and only to a lesser degree in moderate intensity. It seems likely, based on both total and bout-accumulated time spent in MVPA and previous findings , that this amplification is also present in adults but the effect is shifted towards less vigorous intensities.
The finding that a longer epoch length results in increased bout-accumulated time spent in MVPA is supported by Nilsson et al. . For continuous 10 min bouts of MVPA, large differences were found between 5-s, 15-s and 60-s epochs, as few daily activities require moderate intensity sustained for at least 10 consecutive mins. Using a 60-s epoch allows activity to drop below the threshold of MVPA temporarily, as counts are summed over the minute. Shorter epochs will capture this instance of lower intensity, thus the observed difference between total and bout-accumulated MVPA will depend on the relative proportion of MVPA undertaken in steady-state for any given population. As our results show, this epoch influence becomes important relative to the selection of shorter or longer bout durations. This interaction between epoch length and bout duration on MVPA should be considered when comparing bout and non-bout MVPA estimates.
The extent to which epoch length impacts on bout-accumulated MVPA serves to highlight an important issue arising in the field of accelerometry. There is a growing interest in the field with regard to the accumulation of MVPA in bouts. However, it is governed more severely by epoch length compared to total time spent in MVPA. Consequently the processing of bouts using shorter epochs, although more accurately detecting lapses in activity, may not reflect the underpinning physiology. In parallel to developing appropriate definitions and methods for determining bout-accumulated MVPA using shorter epochs, further research into the health effects of varying bout durations, and the allowance of interruptions within them, is required.
The comparison between 1952 cpm  and 2020 cpm  cut-points on total and bout-accumulated time spent in MVPA found the two sets of cut-points to be reasonably comparable across epochs. Fewer than two minutes difference (~5%) between the cut-points was observed in all epochs for both total and bout-accumulated MVPA, thus this difference is unlikely to greatly influence the conclusions of studies assuming their comparability. Cut-point values which are more distant on the movement intensity continuum, however, result in significant differences in time spent in MVPA as also observed in other populations; for example every 500 cpm increase in cut-point between 1000 cpm to 2000 cpm resulted in a 50% reduction of MVPA in older American adults taking part in NHANES 2003–6 (60-sec epochs), a pattern which was consistent across age decades despite differences in absolute levels . The cut-point also interacts with epoch length in determining the likelihood of counts averaging above the chosen MVPA threshold. The effect of selecting a higher cut-point diminishes bout-accumulated MVPA to a lesser degree with longer bout durations.
Wear time was found to be associated with time spent in MVPA. It appears that wear time bias or selection effects may exist in relation to time spent in MVPA. However, the extent to which wear time contributes to the overall effects of accelerometer processing criteria may depend on multiple factors, including the activity level of the population. It may be that those who do more MVPA are more likely to wear the monitor for longer on average each day. It could also be that wearing the accelerometer for longer simply provides greater opportunity to accumulate MVPA. This finding adds to that of Pettee Gabriel et al.  and Schmidt et al.  linking wear time with sedentary behaviours. It is therefore recommended that researchers not only specify non-wear detection criteria for the time-series processing (e.g. zero-string length) and minimum wear time criteria, including monitored days, but also report the (truncated) mean wear time of the sample and how it influences the activity outcome in question.
This study highlights the need for researchers to consider sample characteristics when investigating bout accumulated MVPA. In populations such as ProActive, time spent in MVPA differs substantially when bouts are introduced. The ProActive cohort is relatively inactive; of the 98,556 days of data analysed (differentiated by different settings of epoch length, cut-point or bout length for each participant), 48,581 days (49.3%) contained no MVPA whatsoever and 64,636 days (65.6%) had fewer than 10 mins of bout-accumulated MVPA.
There are important limitations to consider when interpreting the results of the current investigation. We only examined the influence of epoch in the 5-s to 60-s range, the effect of MVPA cut-point in the 1500 cpm to 2500 cpm range, and the effect of bout duration in the 1 min to 15 min range; it is unlikely that the established relationships hold true outside those ranges. In addition, the relatively small sample size means analyses using larger, more heterogeneous populations are needed to establish the generalizability of these findings. However, the relative inactivity of this population means the estimated scaling effects of these processing criteria are likely to be conservative.
Technological advancements in accelerometry have increased the capabilities of available data collection, processing methods and interpretation of activity estimates. Accelerometers are now capable of collecting data using 1-s epochs or even raw acceleration. This can then be used in conjunction with emerging signal processing and pattern recognition models such as artificial neural networks [7, 24] to infer activity type and potentially also enhance the assessment of MVPA. Despite these advances, the issues raised in the current study will remain applicable as the challenge of developing standardised methods of categorising population activity levels using accelerometry will still exist. The advancements in accelerometer technology means researchers are faced with a large number of decisions with regards to the processing and interpretation of PA data. Cut-point selection, epoch length and bout duration have all been identified as influencing MVPA outcomes. We would encourage researchers to collect PA data in the highest available resolution and replicate the analyses of the current investigation to establish regression coefficients for these three parameters in other populations. This enables greater understanding of how populations accumulate activity of a particular intensity, duration and frequency. The present analyses also highlight the need for investigations into the physiological impact of bout-accumulated MVPA in parallel to advances in accelerometer processing.