Efficacy of Experience Sampling and Day Reconstruction Methods as Measures of Physical Activity Intensity

Abstract

Self-report estimates of physical activity (PA) intensity are compromised by recall bias. Since moderate activity is an important component of health recommendations, research is needed to explore methods that measure PA intensity and also minimize recall bias. Experience sampling (ESM) and day reconstruction (DRM) methods have been used to examine various daily experiences, but not to capture PA behavior. PURPOSE: To evaluate the efficacy of ESM and DRM in assessing PA intensity using accelerometry as a criterion measure. MERHODS: Eight males and 7 females ages 23–53 yrs were randomly assigned to DRM or ESM protocols in week 1 and all completed the other protocol in week 2. ESM protocol required subjects to wear a watch programmed to randomly signal 7 times per day for 7 days. Subjects responded to signals within 15 minutes and recorded response time, location, and PA type and intensity (very light, light, medium, hard). DRM data were collected for 2 randomly selected days of the week and PA location, type, and intensity were also recorded for the previous day, according to morning, afternoon and evening time blocks. Accelerometers were worn during both protocols (14 days) and PA counts were measured in 1-minute epochs. ESM intensity responses were correlated with the average PA count that occurred 10 minutes previous to the signal (PA10). DRM intensity responses were correlated with average PA count that occurred during time block episodes (PADRM). PA10 and PADRM were entered as dependent variables in two hierarchical linear models, and reported PA intensity (level1) was modeled as nested within participants (level 2). RESULTS: PA data were non-normally distributed and thus inverted (1/(PA+1)) for analyses. Both models indicated a significant portion of PA variance was attributed to subjects, supporting use of a nested structure. DRM intensity did not explain a significant portion of PADRM variance (t=−1 .62, df=388, p>.05), but ESM intensity significantly predicted PA10 (t=−2.75, df=593, p<01), accounting for 14% of the variance. CONCLUSIONS: The relative immediate recall associated with ESM accounted for a small portion of measured PA intensity variance. As such, self-reports, especially those with longer recall (e.g. DRM), are questionable measures of PA intensity.