Purpose—A previous-day recall (PDR) may be a less error prone alternative to traditional questionnaire-based estimates of physical activity and sedentary behavior (e.g., past year), but validity of the method is not established. We evaluated the validity of an interviewer administered PDR in adolescents (12–17 years) and adults (18–71 years). Methods—In a 7-day study, participants completed three PDRs, wore two activity monitors, and completed measures of social desirability and body mass index (BMI). PDR measures of active and sedentary time was contrasted against an accelerometer (ActiGraph) by comparing both to a valid reference measure (activPAL) using measurement error modeling and traditional validation approaches. Results—Age- and gender-specific mixed models comparing PDR to activPAL indicated: (1) a strong linear relationship between measures for sedentary (regression slope = β1=0.80 to 1.13) and active time (β1=0.64 to 1.09); (2) person-specific bias was lower than random error; and (3) correlations were high (Sedentary: r = 0.60 to 0.81; Active: r = 0.52 to 0.80). Reporting errors were not associated with BMI or social desirability. Models comparing ActiGraph to activPAL indicated: (1) a weaker linear relationship between measures for sedentary (β1=0.63 to 0.73) and active time (β1=0.61 to 0.72); (2) person-specific bias was slightly larger than random error; and (3) correlations were high (Sedentary: r = 0.68 to 0.77; Active: r = 0.57 to 0.79). Conclusions—Correlations between the PDR and activPAL were high, systematic reporting errors were low, and the validity of the PDR was comparable to the ActiGraph. PDRs may have value in studies of physical activity and health, particularly those interested in measuring the specific type, location, and purpose of activity-related behaviors.
- exposure assessment,
- measurement error,
- physical activity,
- behavioral epidemiology
Available at: http://works.bepress.com/patty_freedson/15/