Development of a method for estimating long‐term intake of foods and nutrients

Abstract

Statistical models have been proposed to derive estimates of usual intake of foods and nutrients from short term consumption data, such as the data collected by the 24‐hour dietary recall component of the National Health and Nutrition Examination Survey (NHANES) (Nusser et al 1996, Tooze et al 2006). We have developed an alternative model that combines frequency of consumption data from 14‐day diaries collected by The NPD Group (NPD) as part of the National Eating Trends (NET) survey with gram per eating occasion (eo) data from NHANES. The model was restricted to NET participants reporting 10+ diary days. There were proportionately more young children (2–6 y) and older adults (50+ y) in the NET database. NET survey participants reported on average one less line item per day as compared to those in NHANES. There was no evidence of respondent's fatigue in the NET (average number of lime items reported per day in weeks 1 and 2 was 13.5 and 13.7). We mapped each food reported in NET to one or more foods reported in NHANES. We derived distributions of g/eo for all mapped NHANES foods and the model assigned each NET eo a g/eo from the generated distributions. We present results of our comparison of the 2 surveys, and the assumptions and approach used to develop the model and combine the data from the 2 surveys. Supported by the US Food and Drug Administration.