RESUMEN
BACKGROUND: Self-report questionnaires are a valuable method of physical activity measurement in public health research; however, accuracy is often lacking. The purpose of this study is to improve the validity of the Global Physical Activity Questionnaire by calibrating it to 7 days of accelerometer measured physical activity and sedentary behavior. METHODS: Participants (n = 108) wore an ActiGraph GT9X Link on their non-dominant wrist for 7 days. Following the accelerometer wear period, participants completed a telephone Global Physical Activity Questionnaire with a research assistant. Data were split into training and testing samples, and multivariable linear regression models built using functions of the GPAQ self-report data to predict ActiGraph measured physical activity and sedentary behavior. Models were evaluated with the testing sample and an independent validation sample (n = 120) using Mean Squared Prediction Errors. RESULTS: The prediction models utilized sedentary behavior, and moderate- and vigorous-intensity physical activity self-reported scores from the questionnaire, and participant age. Transformations of each variable, as well as break point analysis were considered. Prediction errors were reduced by 77.7-80.6% for sedentary behavior and 61.3-98.6% for physical activity by using the multivariable linear regression models over raw questionnaire scores. CONCLUSIONS: This research demonstrates the utility of calibrating self-report questionnaire data to objective measures to improve estimates of physical activity and sedentary behavior. It provides an understanding of the divide between objective and subjective measures, and provides a means to utilize the two methods as a unified measure.