Skip to main content
ARS Home » Research » Publications at this Location » Publication #173050
Title: Identifying sources of reporting error using measured food intake.

Author
item Rumpler, William
item Rhodes, Donna
item Moshfegh, Alanna
item PAUL, DAVID - HOPKINS
item Kramer, Matthew

Submitted to: European Journal of Clinical Nutrition
Publication Type: Peer Reviewed Journal
Publication Acceptance Date: 1/29/2007
Publication Date: 4/11/2007
Citation: Rumpler, W.V., Rhodes, D.G., Moshfegh, A.J., Paul, D.R., Kramer, M.H. 2008. Identifying sources of reporting error using measured food intake. European Journal of Clinical Nutrition. 62(4):544-52.

Interpretive Summary: The US Department of Agriculture (USDA)has conducted nationwide food consumption for more than 70 years. Since the data from these surveys have important program and policy applications at the Federal and local levels of government, USDA conducts studies evaluating the methodology used to collect dietary data. This report summarizes a study that measured food intake using the USDA Automated Multiple Pass Method (AMPM). Since 2002, the AMPM hasbeen used to collect data in What We Eat In America, the dietary intake component of the National Health and Nutrition Examination Survey (NHANES). For this study, on two seperate occasions during a 16 week cafeteria style feeding study, we compared measured food intake with reported food intake. Twelve normal weight, nonsmoking men with an average age of 39 ± 9 y, weight of 79.9 ± 8.3 kg, and BMI of 24.1 ± 1.4 kg/m2 participated. The group differences between reported and measured food intake were small and not statistically significant. However, on an individual basis there were substantial differences between reported and measured intake. The two major sources of discrepancy between reported and measured intake were poor estimation of portion size and failure to report all foods consumed. Misreporting associated with "Grains" accounted for over one quarter of the total misreporting. However, on an individual item basis, the misreporting associated with "Grains" had the second lowest percent of total energy consumed when misreported. Beverages which had the highest misreporting on a per-item basis constituted only 5% of the total energy intake and thus had little impact on total error. This work provides insight into the sources of misreporting of food intake and provides important data that can be used to improve the quality of data collection.

Technical Abstract: On two separate occasions during a 16 week cafeteria style feeding study, we compared measured food intake and reported food intake. Twelve normal weight, nonsmoking men with an average age of 39 ± 9 y, weight of 79.9 ± 8.3 kg, and BMI of 24.1 ± 1.4 kg/m2 participated. We defined four types of error which contributed to the discrepancy between reported and measured food intake: Amount, Composition, Addition, and Deletion. The group differences between reported and measured food intake were small and not statistically significant. However, on an individual basis there were substantial differences between reported and measured intake. The difference between reported and measured energy intake averaged 1.56 MJ/d or about 14% of daily intake and ranged from 12 to 3.5 MJ/d. The residual variation for protein, fat, and carbohydrate averaged 18, 23 and 15% of daily intake respectively. Nine categories of foods were examined: Grain Products (G), Fruits & Juices (F), Vegetables (V), Milk-Yogurt-Cheese (D), Meat-Fish-Poultry (M), Beverages (B), Sweets (S), Fats & Oils (F), and Other (O). As a percentage of the 528 items reported, the distribution of differences was 22% V, 20% G, 14% F, 10% D, 9% B, 8% M, 7% L, 6% O and 4% S. G contributed the greatest percentage of the daily energy intake (32%) with M and F contributing 13% each. The remaining categories of foods contributed between 4 and 9% of the total energy intake.Misreporting associated with G accounted for over one quarter of the total misreporting. However, on an individual item basis, the misreporting associated with G had the second lowest percent of total energy consumed when misreported. B, which had the highest misreporting on a per-item basis, constituted only 5% of the total energy intake and thus had little impact on total error.