Abstract

© 2021, The Author(s), under exclusive licence to Springer Nature Limited.Background/objective: To test five different methods to detect misreporting in comparison to doubly labeled water in a sample of older adults. Subjects/methods: A cross-sectional study with thirty-eight Brazilian community-dwelling older adults aged 60–84 years, who had their total energy expenditure measured by doubly labeled water (TEEDLW). Dietary data were collected by two 24 h recalls. Misreporting was compared with estimates obtained by the methods proposed by: Goldberg et al. [1, 2], Black [3], McCrory et al. [4], Huang et al [5], and Rennie et al [6]. Bland–Altman plots with 95% limits of agreement were constructed to assess the agreement between rEI and TEEDLW. Weighted kappa coefficients, sensitivity and specificity analyses, and area under the receiving operator characteristic curve (AUC) were used to test the performance of each method. Results: The prevalence of under-reporters (UR) and over-reporters (OR) obtained by the reference (DLW) were 57.9% (n = 22) and 5.3% (n = 2), respectively. Black [3] presented the worst agreement and McCrory et al. [4] the best one to accurately classify individuals in the three categories of energy reporting. McCrory et al. [4] had the best performance in the sensitivity and specificity analyses detecting UR and plausible reporters. Conclusions: There was a high prevalence of misreporting, especially underreporting, in this sample of community-dwelling Brazilian older adults. The study showed a wide variation in the accuracy of predictive methods to handle misreporting, with none of the equations showing outstanding agreement with the reference. When DLW is not available, a valid method should be chosen to address energy intake reporting.