Comparison of the Physical Activity Frequency Questionnaire (PAFQ) with accelerometry in a middle-aged and elderly population: The CoLaus study

Objective: The Physical Activity Frequency Questionnaire (PAFQ) has been used in several studies, but its validation dates from 1998. We compared the PAFQ with accelerometry data for measuring levels of physical activity (PA) in a middle-aged and elderly population. Design: Cross-sectional analysis was conducted with a sample of 1752 adults from the general population (50.7% female, age range 45.2-87.1 years) living in Switzerland. Participants completed the PAFQ and wore a wristworn accelerometer for 14 consecutive days. Spearman correlation, Lin's concordance coefficient and BlandAltman plots were performed to compare PAFQ and accelerometry data. Results: Compared with the accelerometer, the PAFQ overestimated total, light, moderate and vigorous activity by a median [interquartile range] of 143 [34.5; 249], 72 [12; 141], 23 [-46; 100] and 13 [-1; 41] minutes/day, respectively, and underestimated sedentary behaviour by 123 [14; 238] minutes/day. Spearman's correlation coefficients ranged from 0.171 for vigorous PA and 0.387 for total PA and sedentary behaviour. Lin's concordance coefficients ranged from 0.044 for vigorous PA and 0.254 for moderate to vigorous PA. The difference between PAFQ and accelerometer results increased with increasing time spent at each activity level. Conclusion: There is limited agreement between estimates of activity obtained by PAFQ and those obtained from accelerometers, suggesting that these tools measure activity differently. Although there is some degree of comparability, they should be considered as complementary tools to obtain comprehensive information on both individual and population activity levels.