A team of international researchers, led by Prof. John Speakman, has developed a novel predictive model to screen misreporting in dietary surveys. This innovative approach combines classical statistics and machine learning to estimate energy expenditure, providing a more objective method for assessing the validity of food intake records.
Currently, nutritional epidemiology relies on self-reported data from subjects, which can be inaccurate due to human memory lapses or intentional falsification. The researchers used an isotope-based method called doubly-labeled water technique to directly measure individual energy needs and validated their model with over 6,000 measurements.
The study was published in Nature Food on January 13 and found that approximately half of the food intake records from two large surveys (National Health and Nutrition Examination Survey and National Diet and Nutrition Survey) had unrealistically low levels of energy intake. This highlights the need to re-evaluate widely held beliefs based on problematic methods.
Prof. Speakman emphasizes that continuing to publish erroneous data may not be the best approach for nutrition science, as it can lead to misinformed decisions and policy outcomes. Instead, researchers recommend adopting this new model to ensure accurate estimates of energy intake and revise existing methods accordingly.
Source: https://medicalxpress.com/news/2025-01-screen-misreporting-dietary-surveys.html