A Comparison of Questionnaire, Accelerometer, and Pedometer

PURPOSE: To compare (i) the convergent validity of the self-report Zutphen Physical Activity Questionnaire with the 7-d objective physical activity (PA) measurement by accelerometers and pedometers and (ii) the construct validity of these measures by examining their associations with physical health and psychological and anthropometric variables. METHODS: Five hundred and sixty community-dwelling people aged > or =65 yr were invited from a UK primary care practice and 238 (43%) participated (mean age = 74, 53% male). PA was assessed subjectively by the Zutphen questionnaire (modified to include housework questions) and objectively by the 7-d accelerometer monitoring: a random half also had a pedometer. A questionnaire assessed health, disability, and psychological factors, and anthropometric assessment was performed. RESULTS: Mean daily PA levels were as follows: Zutphen = 9.1 kcal x kg(-1) x d(-1) (SD = 6.6 kcal x kg(-1) x d(-1)); accelerometer activity count = 226,648 (SD = 121,966); accelerometer step count = 6495 (SD = 3212); and pedometer step count = 6712 (SD = 3526). Zutphen score was moderately correlated with accelerometer activity count (R = 0.34, P < 0.001) and pedometer step count (R = 0.36, P < 0.001). Pedometer step count was highly correlated with accelerometer activity count (R = 0.82, P< 0.001) and accelerometer step count (R = 0.86, P < 0.001). Objective PA measures showed strong associations with health and anthropometric and psychological variables. Zutphen score was not significantly related to most health or anthropometric measures but was associated with psychological variables and provided information about activity type. CONCLUSIONS: Convergent validity was strong between accelerometers and pedometers but weaker between these and self-report Zutphen. Pedometers may be preferred to accelerometers for simple studies due to their lower cost. Objective measures had better construct validity, being more strongly associated with established PA determinants, and thus offered better value to researchers than the questionnaire, but the latter provided useful detail on activity type, so a combined approach to PA assessment may be preferable.