Dietary intake data collection: challenges and limitations


  • Ann C Grandjean

    Corresponding author
    1. Medical Nutrition Education Division, School of Allied Health Professions, College of Medicine, University of Nebraska Medical Center, Omaha, Nebraska, USA
      Ann C. Grandjean, University of Nebraska Medical Center, Medical Nutrition Education, 984045 Nebraska Medical Center, Omaha, NE 68198-4045, USA. E-mail: Phone: +1-402-559-5503. Fax: +1-402-559-7420.
    Search for more papers by this author

Ann C. Grandjean, University of Nebraska Medical Center, Medical Nutrition Education, 984045 Nebraska Medical Center, Omaha, NE 68198-4045, USA. E-mail: Phone: +1-402-559-5503. Fax: +1-402-559-7420.


The purpose of this paper is to succinctly review the origin of US dietary surveys, the challenges and limitations of obtaining dietary intake data, the National Nutrition Monitoring and Related Research Act of 1990, the integrated US federal food survey, and the development of the US Department of Agriculture's (USDA) automated multiple-pass method. The USDA has monitored the food consumption patterns of Americans since the late 1890s. In 2002, the US Department of Health and Human Services and the USDA integrated their data collection efforts, with data now collected on a continuous basis. Two 24-hour dietary recalls are obtained using USDA's automated multiple-pass method. By combining their respective areas of expertise, the USDA and the Department of Health and Human Services have increased research opportunities for scientists and provided data foundational for establishing programs and public policy.


In 1894, the 53rd US Congress mandated that the US Department of Agriculture (USDA) Office of Experiment Stations conduct human nutrition studies to obtain information for the purpose of developing recommendations.1 Increased concern about Americans' diets during the Great Depression led to periodic nationwide household surveys.1 Early surveys indicated that one-third of the nation's families had substandard diets.1 These data were the catalyst for flour enrichment and establishment of the National School Lunch Program.2,3

Representative dietary intake data of a population are useful for determining the foods, beverages, and nutrients consumed; assessing dietary adequacies and shortcomings; monitoring trends; and assessing the potential impact of diet on health. In clinical settings, dietary intake data provide information for dietary counseling, adherence assessment, and input for medical management. Although national surveys provide aggregate data, clinical and applied researchers often require that intake data be specific to each patient or subject; thus, data collection procedures are determined in accordance with intended applications and/or research objectives.


Thompson and Subar, in their comprehensive chapter on dietary intake assessment, emphasize that the primary research question provides the basis for intake method selection.4 Examples of questions that guide appropriate method selection include: What dietary data are needed? Will data be considered for the group and/or individuals? What resources are available? What level of accuracy is needed? Food/diet recalls and 24-hour recalls quantify intake, which is an advantage, but they also have a high investigator cost. Food frequency questionnaires, brief instruments, and diet histories capture usual intake and have a lower investigator cost, but they are not quantitatively precise and intake is often misreported.

All dietary intake methodologies are, to some extent, imperfect, with each having limitations.4 Incomplete reporting can occur due to subjects not remembering consuming specific foods or beverages, failing to record in a timely manner, modifying foods consumed when keeping records, inaccurately measuring or estimating portion sizes, and accidentally or purposely failing to record specific items. Providing easy-to-use booklets with guidelines for appropriate methods for weighing and measuring foods and beverages, training subjects how to record, and providing resources for measuring can help increase accuracy. Reviewing diet records with subjects in a familiar and relaxed environment can enhance completeness.


The National Nutrition Monitoring and Related Research Act of 1990 outlined goals that, in addition to other responsibilities, include the following: collecting quality data that are continuous, coordinated, timely, and reliable; using comparable methods for collecting data and reporting results; conducting related research; and efficiently and effectively disseminating and exchanging information with data users.5,6 The USDA and the US Department of Health and Human Services (DHHS) were jointly charged with carrying out the requirements. The Agricultural Research Service and the National Center for Health Statistics signed a memorandum of understanding to collaborate on the collection and implementation of a population-based national nutrition survey. A major step towards the integrated survey was the development and testing of the Automated Multiple-Pass Method to collect dietary information.7 What We Eat in America (WWEIA) resulted from the integration of two nationwide surveys, specifically USDA's Continuing Survey of Food Intakes by Individuals (CSFII) and DHHS' National Health and Nutrition Examination Survey (NHANES).8,9 A number of changes and additions have occurred since the two surveys were integrated in 2002. Beginning with the 2003-2004 WWEIA-NHANES survey, 2 days of intake data for each participant were collected and released.


WWEIA, the dietary intake interview component of NHANES, is conducted as a partnership between the USDA and the DHHS. The USDA Food Surveys Research Group (FSRG) is responsible for monitoring and assessing food consumption and related behavior of the US population.10 The FSRG fulfills its responsibilities, in part, by conducting food consumption surveys and providing information and data briefs.8 To increase the quality and efficiency of food intake surveys and dietary research projects, the FSRG developed an automated 24-hour recall method and automated collection and processing procedures.7 Food model booklets were developed to assist survey participants in estimating the amounts of food consumed.7

The integrated federal food survey was first used in 2002. The DHHS is responsible for sample design and data collection. USDA's responsibilities include methodology of the dietary data collection, development and maintenance of the food and nutrient databases used to code and process the data, and data review and processing. In June 2002, a workshop entitled “Future Directions for What We Eat in America-NHANES: The Integrated CSFII-NHANES” brought together representatives of the major stakeholders.11 The goals for the workshop included discussion regarding the process for and content of data collection and appropriate analysis and evaluation of the dietary aspects of the survey, including intakes of food, beverages, and dietary supplements.


Improving data collection, data processing, analysis, and data release are also responsibilities of the FDRG. The automated multiple-pass method (AMPM), a fully computerized system, was pilot-tested in 1999 and was first used to collect dietary intake information for WWEIA-NHANES in 2002.12 The reports generated from WWEIA-NHANES provide information for policy and program decision makers at the federal, state, and local levels. Data, reports, and briefs are also available to scientists, educators, and industry, among others.8


The USDA Agricultural Research Service's goal to develop and validate methods to increase dietary recall accuracy resulted in development of the AMPM. A synopsis of AMPM's development was published in the June 2004 issue of Agricultural Research.7 Although numerous studies have evaluated the validity of AMPM, assessment is ongoing.13 For a review of USDA's dietary intake data system, see Raper et al.13

Underreporting can skew data interpretation; thus, Briefel et al.14 analyzed data collected with the NHANES III dietary assessment instrument to determine the degree of underreporting of total energy intake (EI) to estimated basal metabolic rate (BMR). They analyzed 24-hour dietary recalls collected from 7,769 nonpregnant adults aged ≥20 years.14 Based on having an EI/BMRest <0.9, 18% of men and 28% of women were classified as underreporters. Twenty-six percent of men and 42% of women indicated they were currently trying to lose weight. Women and older men, those who were overweight or obese, those trying to lose weight, and low- and middle-weight smokers were more likely to underreport. Briefel et al.14 proposed that a portion of the discrepancy in estimated and reported EI might be due to imperfect prediction equations for determining BMR and energy needs. This raises the question of whether it is appropriate to use the same prediction equations for physically active and inactive individuals, smokers, older adults, or individuals with a higher percentage of body fat.

To assess if study subjects may habitually underreport, Briefel et al.14 evaluated a subsample of the participants from the larger study.14 Two 24-h recalls completed by 311 men and 312 women had been collected approximately 1 month apart in the mobile examination center. Both women's and men's mean energy intake was approximately 200 kJ (50 kcal) lower in the second recall. Of the 623 men and women combined, 27% were classified as underreporters on the first recall and 30% on the second. Only 13% of men and 18% of women underreported on both recalls; however, 55% of the men and 58% of the women who underreported on the first recall also underreported on the second. Underreporting is a concern due to the impact it can have on diet trends and overweight in the US population.14

Using the USDA AMPM, Conway et al.15 assessed the hypothesis that women, especially those with a BMI ≥25.0, would underestimate actual energy, protein, carbohydrate, and fat intake. The study was conducted under controlled conditions to test the USDA 5-step multiple-pass method for assessing dietary recall. Subjects were recruited from the scientific, technical, administrative, and service employees of the USDA Agricultural Research Service in Beltsville, Maryland. Participants were informed that snacks would be available for takeout, but meals would be consumed at the Beltsville Human Nutrition Research Center, Human Study Facility. Forty-nine women completed the study, 14 African American, 2 Asian American, and 33 white American. Subjects were on study for 2 weeks. Weight was measured with an electronic balance and height with a stadiometer. Body mass index was calculated using (kg)/height2 (m), and percentage body fat was determined with dual-energy X-ray absorptiometry (DXA). Breakfast, lunch, and dinner included a variety of foods; the same foods were offered for each of the three meals each day. The day after subjects had eaten at the Human Study Facility, the 5-step multiple-pass method was used to collect diet intake data. Intake data were collected from all participants by the same trained interviewer. Protein, carbohydrate, and fat intakes varied significantly among subjects, while energy intake varied fourfold. Carbohydrate consumption varied the most (9.7%), with protein and fat varying by 7.3% and 7.1%, respectively. The mean differences between actual and recalled intakes for all macronutrients were <10%. Obese women were more accurate than were overweight and normal-weight women. The investigators concluded that the USDA 5-step multiple-pass method effectively assesses mean energy intake within 10%.

Using the doubly labeled water (DLW) method, Moshfegh et al. evaluated the accuracy of AMPM, comparing EI with total energy expenditure (TEE) for 262 women and 262 men between the ages of 30 and 69 years.16 The results indicated that 11% of all subjects underreported EI, with 10% of males and 12% of females underreporting EI. Normal-weight subjects underreported EI by <3% combined (1% in males and 6% in females). Mean EI was 90% and 88% of TEE for males and females, respectively. Body mass index (BMI) scores ranged from 18 to 44. The accuracy of the AMPM for assessing EI was within 3% of TEE in the 221 normal-weight subjects. Consistent with other studies, underestimation of EI was greater for subjects with a higher BMI. Both males and females lost weight over the 2-week DLW period, with the amount lost being greatest in the obese groups. The authors recommended conducting research to determine the psychological and behavioral factors that contribute to underreporting by overweight and obese persons.

Tooze et al.17 conducted a study to determine psychological and behavioral factors associated with underreporting of EI on food-frequency questionnaires (FFQ) and 24-hour recalls. Participants were 223 men and 261 women between the ages of 40 and 69 years. Energy expenditure was measured with the DLW method, height and weight were measured, and BMR was calculated. Participants answered questions about smoking history, eating habits, and frequency of eating out and at home. They also completed a physical activity questionnaire as well as a health questionnaire, and answered questions about usual activity levels such as sitting, standing, and lifting. Sixty percent of the women and 76% of the men were overweight or obese, with 29% of both men and women being obese. Both men and women underreported EI on the FFQ and the 24-hour recall. For women, the best predictors of underreporting in the FFQ were fear of negative evaluation, weight-loss history, and percentage of energy consumed from fat. For men, BMI, eating frequency, and comparison of activity level with that of others of the same sex and age were the best predictors of underreporting on the FFQ. In the 24-hour recall models, social desirability, fear of negative evaluation, BMI, energy from fat, usual activity, and variability in number of meals per day were the best predictors of underreporting for women. For men, social desirability, dietary restraint, BMI, eating frequency, dieting history, and education were the best predictors.


Continuous monitoring of diet and health trends provides sequential data that are useful in making decisions regarding program and policy development. Data on dietary intake, nutritional status, and health indicators that are accumulative and standardized are valuable not only for making decisions at the national level, but also at the state and local levels. Combining the complementary expertise of the USDA and DHHS has resulted in enhanced data collection, increased applicability of research, and increased awareness of nutrition and health issues. The availability of WWEIA-NHANES survey data provides a wealth of research opportunities.

The validation papers reviewed are but a few of the many studies that have been conducted during the development and validation of the automated multiple-pass method. Additional papers can be found on the Agricultural Research Service (ARS) Food Surveys Research Group webpage.18 A food and nutrient database for dietary studies is also provided.19


The author serves as a Scientific Advisor to the Hydration Committee at the International Life Sciences Institute North America and served as a member of the conference planning committee. The author was reimbursed for her travel expenses to attend the conference and was given an honorarium for her conference attendance, presentation, and the preparation of a manuscript.

Declaration of interest.  The author has no relevant interests to declare.