Abstract 4459: Evaluation of the performance of a previous day recall of time spent in active and sedentary behaviors

Previous-day recalls (PDR) collect information about time spent in specific active and sedentary behaviors, are presumed to be less error prone than traditional physical activity questionnaires, and therefore may be useful in epidemiologic studies of physical activity and cancer. Purpose. Evaluate the performance of a PDR for active and sedentary behaviors administered by phone by comparing it to a valid reference measure. Methods. Adolescents (n=91; 12-17 yrs; 53% female) and adults (n=88; 18-71 yrs; 55% female) wore an activPAL (aPAL) for 7 days to measure sedentary (sit/lie) and active time (standing/stepping) and completed up to three PDRs. Interviewers conducted PDRs eliciting open-ended reports of time spent sleeping and in specific active and sedentary behaviors. Bland-Altman methods and regression models evaluated PDR performance versus aPAL on matched days of observation. Mixed models were fit to quantify the error structure of the PDR using the following equation: PDRij = β0 + α1Tij + ri + eij, where Tij is the estimate of truth from aPAL and β1 is the slope of the regression of aPAL on PDR. Variances were estimated for systematic (≥ri) and random reporting errors (≥eij) for the PDR. Results. Adolescents and adults reported 10.0 (SD=2.2) and 9.9 (SD=2.8) hr/d of sedentary time, and 4.2 (SD=1.8) and 5.2 (SD=2.6) hr/d of active time, respectively. Total PDR active and sedentary time was greater than aPAL wear time (difference=0.58 hr/d, p0.05) negative bias for active time in adolescents (bias=−0.27 (SD=1.37) hr/d) and adults (bias=−0.13 (SD=1.44) hr/d). In contrast, a significant positive bias in PDR sedentary time was evident (p