Assessing Physical Activity Levels through Motion Sensors in Exergaming

Background/Purpose: Recently exergaming has been integrated into school-based programs to promote children’s physical activity (Gao et al., 2013). Pedometers and accelerometers have been widely used in assessing exergaming physical activity in adults and children. However, the validity and reliability of using such motion sensors in exergaming of children remain unanswered. This study was designed to examine the validity and reliability of motion sensors in assessing physical activity levels in an elementary school-based exergaming program.

Method: Participants were 377 first through fourth grade children (190 girls; 20 classes) enrolled in a suburban Title I elementary school in Texas. Twelve exergaming stations were set up in a classroom, offering 8 different Wii exergames. Children attended the 30-minute exergaming class every other day, rotating from station to station every class. Their physical activity levels were assessed by New-Lifestyles -1000 pedometers and ActiGraph GTX3 accelerometers for 27 exergaming classes in 2012. Children’s steps was used as the outcome variable for pedometers, and time engaged in sedentary, light, and moderate-to-vigorous physical activity were used as the outcome variables for accelerometers.

Analysis/Results: 377 students received 17 repeated assessments by accelerometer after data screening. There was no difference among trails, p = .53, indicating no learning or fatigue effect. Intraclass correlation (ICC) was calculated through two-way mixed effects model. A low degree of reliability was found (single measure ICC = .03). For pedometers, ANOVA did detect a possible learning effect for 27 classes, p < .01. The single measure ICC was .20. To explore the relationship between pedometer and accelerometer data while controlling children’s background (age, gender, and race) and aforementioned learning effect, Hierarchical Linear Modeling was conducted. Children’s pedometer steps had a significantly positive relationship with moderate-to-vigorous physical activity of accelerometer. However, only 1.3% variance was explained by pedometer as a predictor.

Conclusions: It appears motion sensors demonstrated low reliability in assessing physical activity levels in exergaming of elementary school children, as the ICCs were far from ideal. Such low reliability shown by ICC may primarily due to the following reasons:  different exergames children played in different classes and inconsistency of instrument placement over trails due to the young age of children. The results may lead to the implication for future exergaming research: same exergames need to be employed when motion sensors are used to eliminate the variability caused by different games; and consistency of the device placement should be given much more attention.