This study was designed to develop a prediction algorithm that would allow the Previous Day Physical Activity Recall (PDPAR) to be equated with temporally matched data from an accelerometer.
Participants (n = 121) from a large, school-based intervention wore a validated accelerometer and completed the PDPAR for 3 consecutive days. Physical activity estimates were obtained from PDPAR by totaling 30-minute bouts of activity coded as ≥4 METS. A regression equation was developed in a calibration sample (n = 91) to predict accelerometer minutes of moderate to vigorous physical activity (MVPA) from PDPAR bouts. The regression equation was then applied to a separate, holdout sample (n = 30) to evaluate the utility of the prediction algorithm.
Gender and PDPAR bouts accounted for 36.6% of the variance in accelerometer MVPA. The regression model showed that on average boys obtain 9.0 min of MVPA for each reported PDPAR bout, while girls obtain 4.8 min of MVPA per bout. When applied to the holdout sample, predicted minutes of MVPA from the models showed good agreement with accelerometer minutes (r = .81).
The prediction equation provides a valid and useful metric to aid in the interpretation of PDPAR results.
Tucker is with the Dept of Health, Nutrition, and Exercise Sciences, North Dakota State University, Fargo, ND. Welk is with the Dept of Kinesiology, Iowa State University, Ames, IA. Nusser and Beyler are with the Dept of Statistics, Iowa State University, Ames, IA. Dzewaltowski is with the Dept of Kinesiology, Kansas State University, Manhattan, KS.