Internet-Based and Self-Reports in Assessing Diet and Physical Activity Within AARP

Primary outcomes

1. Self-reported energy intake from automated self-administered 24-hour dietary recall (ASA24), Diet History Questionnaire (DHQ)-II, and 4-day food record

   Time frame: Up to 12 months

   Will be evaluated against energy intake measured by doubly labeled water (DLW). Dietary intake of protein, potassium, and other nutrients will be compared to urinary nitrogen (UN), urinary potassium, and other nutrients measured in urine and blood.

2. Measurement error structure of self-reported active and sedentary behaviors (e.g. activity completed over time in 24-hours [ACT24], questionnaires)

   Time frame: Up to 12 months

   Will be examined using energy expenditure measured by DLW, physical activity monitor data, and hear rate monitor data. Data will be also used to evaluate alternative statistical methods for combining different types of self-report data on diet and physical activity-related behaviors as well as self-report plus objectively measured data.

3. 24-hour urine samples

   Time frame: Up to 12 months

   Will be analyzed for UN, potassium, para-amino benzoic acid (PABA), and other nutrients that are potential objective markers of intakes. First morning void urine also will be evaluated for its potential to be used as a source of biomarkers replacing 24-hr urine. Urine will be analyzed to characterize metabolic profiles of individuals.

4. Fasting blood sample analyses

   Time frame: Up to 12 months

   Will be processed to extract plasma and red and white blood cells, and analyzed in the future for vitamin C, folate, vitamin E, vitamin A, carotenoids, lipids, fatty acids and other nutrients.

5. Saliva samples

   Time frame: Up to 12 months

   Will be analyzed for oral microbiomes and to characterize metabolic profiles of individuals.

6. Measurement error structure of self-reported dietary assessment instruments

   Time frame: Up to 12 months

   Evaluated using latent-variable (measurement error) models. The models allow self-reported dietary intake to have intake-related and person-specific biases, but assume that reference biomarkers (doubly-labeled water, urinary nitrogen, urinary potassium) provide unbiased estimates of true usual intake at the individual level. Such models allow one to estimate the joint distribution of true and reported intake, and to estimate parameters of interest such as the correlation of true and reported intake and the "attenuation factor", or slope in the regression of true intake on reported intake.

7. Measurement error in self-reported total physical activity (measured in energy expenditure)

   Time frame: Up to 12 months

   Evaluated using latent-variable (measurement error) models. The models allow total energy expenditure derived from a physical activity monitor to have intake-related and person-specific biases, but assume that reference biomarkers (doubly-labeled water, urinary nitrogen, urinary potassium) provide unbiased estimates of true usual intake at the individual level. Such models to help develop better algorithms for estimating total energy expenditure from the measurements of physical activity monitors. For other physical activity measures of interest, such as minutes of vigorous activity or minutes of sedentary time per day, will use the physical activity monitors to derive approximately unbiased estimates that can be used as reference instruments.

8. Reduction in measurement error

   Time frame: Up to 12 months

   Will use multivariate measurement error models to estimate the joint distribution of true and reported values (nutrient intake, total energy, physical activity, body size). Then calculate residual reported intake (given reported total energy, physical activity and body size) and residual true intake (given true total energy, physical activity and body size), and assess the measurement error properties of reported residual intake as a measure of true residual intake. Specifically, will calculate the correlation of true and reported residual intake and the attenuation factor for residual reported intake and compare them to the same measures calculated for the simpler model that adjusts only for energy.