A Measurement Error Model for Self-Reported Physical Activity

• Activity Number: 297
• Type: Contributed
• Date/Time: Tuesday, August 2, 2011 : 8:30 AM to 10:20 AM
• Sponsor: Biometrics Section
• Author(s): Janet A. Tooze*+ and Richard P. Troiano and Raymond James Carroll and Laurence S. Freedman
• Companies: Wake Forest University Health Sciences and National Cancer Institute and Texas A & M University and Gertner Institute for Epidemiology and Health Policy Research
• Keywords: Measurement error ; Questionnaires ; Validation ; Physical Activity Assessment

Abstract

Self-reported physical activity assessment instruments are prone to error yet systematic investigations into the structure of this measurement error are lacking. We propose a measurement error model for physical activity assessment instruments using physical activity level (PAL), the ratio of total energy expenditure (TEE) to basal energy expenditure (BEE). A physical activity questionnaire (PAQ) was administered to 451 participants aged 40-70 y in the Observing Protein and Energy Nutrition (OPEN) Study. MET minutes from the PAQ were used to estimate PAL. The main objective is to relate the PAL measurement from the PAQ to true PAL. Although true PAL was not observed, TEE was measured using an unbiased biomarker conforming to a classical measurement error model and BEE was estimated from an equation. Because BEE is estimated from an equation, it is prone to Berkson error. Therefore, the non-questionnaire measure of PAL has a mix of classical (TEE) and Berkson error (BEE). We present a measurement error model for PAL that accommodates this mixture of errors and use it to establish the relationship between the PAQ measure of PAL and true PAL and its application from the OPEN study.