Demographic and psychosocial correlates of measurement error and reactivity bias in a 4-d image-based mobile food record among adults with overweight and obesity

Abstract

AbstractImproving dietary reporting among people living with obesity is challenging as many factors influence reporting accuracy. Reactive Reporting may occur in response to dietary recording, but little is known about how image-based methods influence this process. Using a 4-d image-based mobile food record (mFRTM), this study aimed to identify demographic and psychosocial correlates of measurement error and reactivity bias, among adults with BMI 25–40 kg/m2. Participants (n 155, aged 18–65 years) completed psychosocial questionnaires and kept a 4-d mFRTM. Energy expenditure (EE) was estimated using ≥ 4 d of hip-worn accelerometer data, and energy intake (EI) was measured using mFRTM. EI:EE ratios were calculated, and participants in the highest tertile were considered to have Plausible Intakes. Negative changes in EI according to regression slopes indicated Reactive Reporting. Mean EI was 72 % (sd = 21) of estimated EE. Among participants with Plausible Intakes, mean EI was 96 % (sd = 13) of estimated EE. Higher BMI (OR 0·81, 95 % CI 0·72, 0·92) and greater need for social approval (OR 0·31, 95 % CI 0·10, 0·96) were associated with lower likelihood of Plausible Intakes. Estimated EI decreased by 3 % per d of recording (interquartile range − 14 %,6 %) among all participants. The EI of Reactive Reporters (n 52) decreased by 17 %/d (interquartile range − 23 %,–13 %). A history of weight loss (> 10 kg) (OR 3·4, 95 % CI 1·5, 7·8) and higher percentage of daily energy from protein (OR 1·1, 95 % CI 1·0, 1·2) were associated with greater odds of Reactive Reporting. Identification of reactivity to measurement, as well as Plausible Intakes, is recommended in community-dwelling studies to highlight and address sources of bias.