To assess the utility of measurement methods that may be more accurate and precise than traditional questionnaire-based estimates of habitual physical activity and sedentary behavior we compared the measurement properties of a past year questionnaire (AARP) and more comprehensive measures: an internet-based 24-h recall (ACT24), and a variety of estimates from an accelerometer (ActiGraph).
Participants were 932 adults (50–74 yr) in a 12-month study that included reference measures of energy expenditure from doubly labeled water (DLW) and active and sedentary time via activPAL.
Accuracy at the group level (mean differences) was generally better for both ACT24 and ActiGraph than the AARP questionnaire. The AARP accuracy for energy expenditure ranged from −4% to −13% lower than DLW, but its accuracy was poorer for physical activity duration (−48%) and sedentary time (−18%) versus activPAL. In contrast, ACT24 accuracy was within 3% to 10% of DLW expenditure measures and within 1% to 3% of active and sedentary time from activPAL. For ActiGraph, accuracy for energy expenditure was best for the Crouter 2-regression method (−2% to −7%), and for active and sedentary time the 100 counts per minute cutpoint was most accurate (−1% to 2%) at the group level. One administration of the AARP questionnaire was significantly correlated with long-term average from the reference measures (ρTX = 0.16–0.34) overall, but four ACT24 recalls had higher correlations (ρTX = 0.48–0.60), as did 4 d of ActiGraph assessment (ρTX = 0.54–0.87).
New exposure assessments suitable for use in large epidemiologic studies (ACT24, ActiGraph) were more accurate and had higher correlations than a traditional questionnaire. Use of better more comprehensive measures in future epidemiologic studies could yield new etiologic discoveries and possibly new opportunities for prevention.
Supplemental digital content is available in the text.
1Metabolic Epidemiology Branch, Division of Cancer Epidemiology and Genetics, National Cancer Institute, Rockville, MD; 2Kinesiology Department, California Polytechnic State University, San Luis Obispo, CA; 3University of Wisconsin, Biotech Center and Nutritional Sciences, Madison, WI; 4Department of Statistics, Texas A&M University, College Station, TX; 5School of Mathematical and Physical Sciences, University of Technology Sydney, AUSTRALIA; 6Risk Factor Assessment Branch, Division of Cancer Control and Population Sciences, National Cancer Institute, Rockville, MD; and 7Biostatistics Branch, Division of Cancer Epidemiology and Genetics, National Cancer Institute, Rockville, MD
Address for correspondence: Charles E. Matthews, Ph.D., Metabolic Epidemiology Branch, Division of Cancer Epidemiology and Genetics, National Cancer Institute, 9609 Medical Center Dr, 6E444, Bethesda, MD 20892-9704; E-mail: firstname.lastname@example.org.
Submitted for publication April 2017.
Accepted for publication July 2017.
Supplemental digital content is available for this article. Direct URL citations appear in the printed text and are provided in the HTML and PDF versions of this article on the journal’s Web site (www.acsm-msse.org).