Development and validation of a physical activity games playability scale

For the past three decades, the prevalence of childhood obesity has been on the rise and correspondingly engagement time in sedentary activities has escalated. In contrast, interest, and participation rates in physical education classes are declining. Fun and interesting physical activity (PA) games could help to prevent the decline and possibly reverse inactivity. The purpose of this study was to develop and validate a Physical Activity Playability Scale (PAGPS) in order to provide more detailed “game information” assisting end users (e.g., policy makers, PE teachers, et al.) in choosing the “best possible” children’s PA games. A two-stage development and validation process was employed for this study. Five content experts (N=5) were recruited to draft and develop the PAGPS scale. By applying the heuristic approach, content experts selected, reviewed, commented, evaluated, and eventually determined the relevant PA games factors/subscales, which helped in establishing the content validity evidence for the PAGPS. Ten factors that were identified to represent game domain were Fun, Social, Cognitive, Physical, Skills, Game Structure, Language, Environment, Game Difficulty, and Player’s Characteristics, and a total of 116 items were developed for these factors. Two hundred PE teachers (N=200) were recruited in Malaysia to further determine the most suitable items for the PAGPS. Principal Component Analysis (PCA) was chosen to systematically trim the large amount of variables but maintain as much of the information from the PAGPS (draft) data set. A six-factor construct (with 99 items), including fun and social, cognitive, physical and skills, games structure and environment, game difficulty and player’s characteristics and language were confirmed for the PAGPS. Rasch Analysis, an item response theory approach, was then chosen for the item reduction process by taking advantage of the analysis (i.e., invariance, ability to locate all facets on a same scale, and additive) over the classical testing theory based approach. Items were deleted based upon three criterions: goodness-of-fit statistics, item difficulty (logits), and content balance. As a result, four shorter PAGPS versions were created, with 51-item, 36-item, 28-item and 20-item, respectively. The 51-item version was chosen because of its high correlation (r = .98) with the original 99-item version and its balanced content coverage. Cronbach’s alpha analysis was also performed to determine the internal structural consistency. Ten Malaysian PA games were selected to validate the 51-item PAGPS (Game Rating Scale), including One-Leg, Kali-Tui, Blind Man’s Bluff, Simon Says, Eagle and Hen, Hopscotch, Police and Thief, Duck Duck Goose, Monkey in the Middle, and Mr. Wolf. Sixty children (N=60) consisting of two age-groups (Grade 2: n= 30, Grade 5: n= 30) were recruited to play all ten PA games, their reactions towards each game were video-recorded for rating purposes. Ten raters (N=10) scored each PA game video (10 videos for Grade 2 and 10 videos for Grade 5, respectively) using the 51-item PAGPS. The rating scores were analyzed for inter-rater reliability evidence, discriminant evidence (P and K coefficient), and game descriptive statistics (validity evidence). Inter-rater reliability was found to be within a good range of .70-.91, P coefficient from .15-1.00 and K coefficient from -.70-1.00. Together, reliability evidence (i.e., internal structure reliability and inter-rater reliability) and validity evidence (content validity, discriminant validity and games’ descriptive statistics) provided preliminary support for the psychometric quality of the PAGPS (Game Rating Scale). This study also illustrated that, with a combination of the convent balance, Rasch analysis can be used effectively for item reduction while maintaining the psychometric properties of the original measure. More PA researchers should take advantage of this method when developing and constructing measures