A Clustering Approach for Modeling and Analyzing Changes in Physical Activity Behaviors From Accelerometers

A Digital Educational Intervention With Wearable Activity Trackers to Support Health Behaviors Among Childhood Cancer Survivors: Pilot Feasibility and Acceptability Study

“…To complement our assessment of time spent in MVPA before and after the intervention, we used the richness of the accelerometer data and explored the continuity and duration of various physical activity bouts. Physical activity bouts (eg, frequent short bouts or infrequent long bouts) are important to understand how participants achieve their physical activity levels and how they are distributed over the day or week [ 21 , 54 ]. Our cluster analysis showed that 27% (3/11) of the participants moved to a more active cluster, indicating that they increased the frequency and duration of MVPA bouts.…”

“…There is also a focus on health literacy throughout the program, which involves teaching participants to define and classify physical activity intensities (light, moderate, and vigorous), self-ratings of perceived effort during exercise, recommended guidelines relating to physical activity, sedentary behaviors, fitness, well-being (physical, mental, and social), screen-based behaviors, and sugar intake. iEngage was built for schoolchildren aged 10 to 12 years without a chronic disease; it was piloted in a rural school in New Caledonia [ 18 ] and trialed in 2 primary schools in Sydney, New South Wales, Australia, in 2017 and 2018 [ 19 , 21 ]. Caillaud et al [ 19 ] and Diaz et al [ 21 ] have detailed the iEngage program.…”

“…iEngage was built for schoolchildren aged 10 to 12 years without a chronic disease; it was piloted in a rural school in New Caledonia [ 18 ] and trialed in 2 primary schools in Sydney, New South Wales, Australia, in 2017 and 2018 [ 19 , 21 ]. Caillaud et al [ 19 ] and Diaz et al [ 21 ] have detailed the iEngage program.…”

“…All participants’ daily physical activity was clustered by eight factors: total time spent in MVPA of ≥3-second bouts and ≥30-second bouts, total time spent in sedentary behavior of ≥60-second bouts and ≥300-second bouts, frequency of ≥3-second and ≥30-second MVPA bouts, and frequency of ≥60-second and ≥300-second sedentary bouts. Informed by the studies by Diaz et al [ 21 ] and Schaefer et al [ 38 ], we selected 3-second–bout lengths to represent short MVPA bouts because activity recorded in <3 seconds typically represents agitation of the GENEActiv activity tracker or noise , rather than meaningful physical activity. We used bout lengths of at least 30 seconds to represent sustained MVPA.…”

“…Exploring various physical activity bouts plays an important role in understanding how survivors accumulate their physical activity; for example, whether survivors engage in frequent and short bouts of MVPA or less frequent and long bouts of MVPA. To capture these various physical activity intensities, durations, frequencies, and bouts, we used the cluster analysis method proposed in earlier work [ 21 ] to analyze the impact of iBounce on the survivors’ physical activity behaviors.…”