WHAT IS A PHYSIOLOGICALLY RELEVANT OUTCOME MEASURE AND EPOCH LENGTH
IN OBJECTIVELY QUANTIFYING SEDENTARISM?
Wullems J. A.1, Morse Ch. I.1, Degens H.1, 2, Verschueren S. M. P.3, Onambele-Pearson G.1
1Manchester Metropolitan University, UK, 2Lithuanian Sports University, Lithuania,
3University of Leuven, Belgium
Relevance of the research. Total daily activity can be classified in terms degree of sedentary behaviour (SB) or physical activity levels (PA). Both SB and low PA have distinct negative effects on health and it is therefore important to accurately monitor daily mobility behaviour to obtain insight into a person’s long-term health prognosis (Lord et al., 2011; Gorman et al., 2014). Although accelerometry is preferred in most studies, there is no current consensus for a gold-standard device, or method of data analysis (Pedišić, Bauman, 2015). Indeed, use of inappropriate devices or data analysis has the potential danger of misinterpreting the true pattern of daily behaviour (Gorman et al., 2014). Accurate measurement of SB and PA is key to designing individualised lifestyle recommendations (Rezende et al., 2014). This is of importance in older adults (≥65 years of age) since they are the most sedentary and less physically active age group (Gennuso et al., 2013). We believe that using thigh-mounted triaxial accelerometry combined with an algorithm that includes a physiologically relevant outcome measure and epoch length can monitor objectively and accurately SB and PA. This objective approach will eventually help to understand how SB and PA are related to healthy ageing (Visser, Koster, 2013).
The aim of the research is to refine an algorithm to monitor objectively SB and PA in elderly, and the objective is to determine the physiologically relevant outcome measure and epoch length to be included.
Research methods and organization. Triaxial accelerometer data (thigh-mounted bilaterally; 60 Hz sampling rate) and expired gas were collected from six participants (algorithm-refining group: n = 5, aged 67‒82 years; 2 women; body mass index (BMI) 21.6‒35.8 kg·m-2 & algorithm validation group: n = 1, aged 72 years; female; BMI 23.8 kg·m-2) during a set of laboratory-based standardised activities of daily living (three minutes each) of different intensities; such as lying down, sitting, standing and walking. Expired gas was collected during the final minute of each activity. These samples were used to estimate energy expenditure (EE) and calculate the metabolic equivalent (MET) of the simulated activities of daily living. The accelerometer data acquired during the same minute was analysed using 18 different combinations of epoch lengths (1, 5, 10, 15, 30 and 60 seconds) and outcome measures (activity counts (AC; summed acceleration signals divided by device resolution), sum of vector magnitude (SVM) and total movement (TM). The outcome of each combination was plotted against EE to 1) explore correlations, and 2) calculate algorithm cut-off points according to 1.5 and 3.0 MET thresholds. For these purposes, data from the algorithm-refining group was used only. Next, all 18 algorithms (using both thigh orientation and cut-off points) were applied to the accelerometer data from the algorithm validation group only. The applied algorithms classified each epoch as either, SB, standing, light-intensity PA (LIPA) or moderate-to-vigorous PA (MVPA). To investigate which algorithm (and thus outcome measure and epoch length) was most valid, agreement with the actual performed activity per epoch was determined.
Results and discussion. Correlations coefficients found for SVM and TM were > 0.70 regardless of epoch length, whilst AC showed a correlation coefficient of 0.79 for the 1 second epoch length, but < 0.56 for the others. Excellent agreement (100 %) with the actual performed activity per epoch was shown when classifying SB, irrespective of outcome measure or epoch length. Standing was difficult to detect when using AC (highest agreement 7 %, while 100 % agreement was found for both SVM and TM regardless of epoch length, with the exception of using TM/30 seconds epoch (75 %). High agreement was found for classifying PA, independent of epoch length (AC: 75‒86 %; SVM: 96‒100 %; TM: all 100 %). When focusing on PA intensity, LIPA seems more difficult to correctly classify than MVPA, regardless of epoch length (AC: 0‒34 % vs. 85‒100 %; SVM: 37‒75 % vs. all epochs 100 %; TM: 0‒36 % vs. all epochs 100 %). Inferior results when using AC could be due to the lack of overall variation in outcome measure, resulting in overlapping activity type clusters. The fact that preliminary data were used might explain the under- and overestimation of LIPA and MVPA respectively.
Conclusions. The preliminary results of this study suggest that the optimal epoch length for determining sedentarism is dependent on the eventual outcome measure.