Chapter three - Methods of monitoring food and nutrient intake
3.2 Methods of assessing dietary intake
3.3 Selecting the most appropriate dietary data collection method
3.4 Rapid methods for collection of dietary data
3.5 Conversion of food intakes to nutrients
3.7 Approaches to monitoring the effectiveness of established guidelines
The purposes of this chapter are to:
• provide an overview of methods of assessing food consumption;
• give examples of rapid methods for collecting additional dietary data;
• address issues involved in converting food intake to nutrients;
• discuss analysis of food intake data by foods, food groups and eating practices;
• elaborate on approaches to monitoring the effectiveness of established dietary guidelines.
Food consumption data may be collected at the national, household or the individual level. Although data collected at the level of the individual are the most useful for assessing dietary adequacy and adherence to FBDG, food supply and household data provide information that is useful for many other purposes.
Food supply data at the national level, such as food balance sheets or food disappearance data (1), provide gross estimates of the national availability of food commodities. These data may also be used to calculate the average per capita availability of energy and the macronutrients. A major limitation of national supply data is that they reflect food availability rather than food consumption. Other uses, such as animal feed and industrial applications, as well as losses due to cooking or processing, spoilage and other sources of waste are not easily accounted for. Despite these limitations, national food supply data are useful for tracking trends in the food supply and for determining availability of foods that are potentially good sources of nutrients or of food groups targeted for dietary guidance. Food supply data are not useful for evaluating individual adherence to dietary reference values (DRY) nor for identifying subgroups of the population at risk of inadequate nutrient intakes.
Information regarding food availability at the household level may be collected by a variety of methods (2). Such data are useful for comparing food availability among different communities, geographic areas and socioeconomic groups, and for tracking dietary changes in the total population and within population subgroups. However, these data do not provide information on the distribution of foods among individual members of the household.
The five general methods for assessing dietary intake for individuals are described below, along with the major strengths and limitations of each (these methods are described in greater detail in 3.6):
• Food records. Food records, also called food diaries, require that the subject (or observer) report all foods and beverages consumed for a specified period (usually one to seven days). Amounts of each food item may or may not be recorded, depending on the study objectives. If nutrient intakes are to be calculated, the amounts consumed should be estimated as accurately as possible. Amounts may be determined by weighing or by estimating volumes. in some situations, only those foods of particular interest are recorded. For example, to estimate intake of a food component found only in animal products, food records might be limited to foods containing meat, poultry, fish, eggs or dairy products. However, if total energy intake is required, the food record should include all foods consumed.
• 24-hour dietary recalls. The 24-hour dietary recall consists of a listing of foods and beverages consumed the previous day or the 24 hours prior to the recall interview. Foods and amounts are recalled from memory with the aid of an interviewer who has been trained in methods for soliciting dietary information. The interview is usually conducted face to face, but may also be conducted by telephone. In some situations, the recall is self-administered by the subject, but this approach may not yield sufficiently reliable data. A brief activity history may be incorporated into the interview to facilitate probing for foods and beverages consumed.
• Food frequency questionnaires. A food frequency questionnaire (FFQ), sometimes referred to as a "list-based diet history", consists of a structured listing of individual foods or food groupings. For each item on the food list, the respondent is asked to estimate the frequency of consumption based on specified frequency categories which indicate the number of times the food is usually consumed per day, week, month or year. FFQs are generally self-administered but may also be interviewer-administered. The number or types of food items may vary, as well as the number and types of frequency categories. FFQs may be unquantified, semi-quantified or completely quantified. The unquantified questionnaire does not specify serving sizes, whereas the semi-quantified tool provides a typical serving size as a reference amount for each food item. A quantified FFQ allows the respondent to indicate any amount of food typically consumed. Some FFQs include questions regarding usual food preparation methods, trimming of meats, use of dietary supplements, and identification of the most common brand of certain types of foods such as margarines or ready-to-eat cereals. The answers to these questions are then incorporated into the calculation of nutrient intakes. FFQs are commonly used to rank individuals by intake of selected nutrients. Although FFQs are not designed for estimating absolute nutrient intakes, the method may be more accurate than other methods for estimating average intake of those nutrients having large day-to-day variability and for which there are relatively few significant food sources (e.g. alcohol, vitamin A and vitamin C). Brief FFQs may focus on one or several specific nutrients. Comprehensive FFQs designed to estimate a large number of nutrients generally include between 50 and 150 food items.
• Diet histories. The meal-based diet history is designed to assess usual individual intake. It consists of a detailed listing of the types of foods and beverages commonly consumed at each eating occasion over a defined time period which is often a "typical" week. A trained interviewer probes for the respondent's customary pattern of food intake on each day of the typical week. The reference time frame is often the past month or the past several months, or may reflect seasonal differences if the time frame is the past year.
• Food habit questionnaires. May be designed to collect either general or specific types of information, such as food perceptions and beliefs, food likes and dislikes, methods of preparing foods, use of dietary supplements, social settings surrounding eating occasions. This type of information is frequently included along with the other four methods, but it may also be used as the sole data collection method. These approaches are commonly used in rapid assessment procedures (RAP). The questionnaires may be either open-ended or structured, self- or interviewer-administered, and may include any number of questions depending on the information required. Use of food habit questionnaires is further discussed in the section on rapid assessment methods.
• Combined methods Different types of dietary assessment methods may be combined to improve accuracy and facilitate interpretation of the dietary data. Methods may also be combined for practical reasons. For example, food records have been combined with 24-hour recalls to make the best use of resources in past surveys of the US Department of Agriculture. An FFQ focused on selected nutrients was used in addition to the 24-hour recall in the Third National Health and Nutrition Examination Survey (NHANES III). A 24-hour recall is frequently used to help establish the typical meal plan for conducting a diet history (7), and a FFQ may be used as a cross-check for the other three types of methods. (See Table 1, Summary of dietary assessment methods).
It is becoming increasingly common to use another more valid (and often more costly) method of dietary assessment on a random sub-sample of the study population. Comparison of the results from the two methods permits adjustment of mean intake values obtained from the primary method. For example, if an FFQ is the primary method used, a more detailed method, such as multiple recalls or food records, is also used in a subset of survey participants. The more detailed method along with the FFQ is expected to provide more accurate estimates of intake distributions. Results from the sub-sample are then used to adjust the mean estimates from the FFQ for the total population.
Table 1. Summary dietary assessment methods
Type of method |
Major strengths |
Major limitations type |
Food record |
- does not rely on memory |
- high participation burden |
- easy to quantify amounts |
- requires literacy | |
- open-ended |
- may alter intake behaviour | |
24-h dietary recall |
- little respondent burden |
- relies on memory |
- no literacy requirement |
- requires skilled interviewer | |
- does not alter intake behaviour |
- difficulty to estimate amounts | |
Food frequency questionnaire |
- relatively inexpensive |
- relies on memory |
- preferable method for nutrients with very high day variability |
- requires complex calculations | |
- to day to estimate frequencies | ||
- requires literacy | ||
- does not alter intake behaviour |
- limited flexibility for describing foods | |
Diet history (meal-based) |
- no literacy requirement |
- relies on memory |
- does not alter intake behaviour |
- requires highly trained interviewer | |
- open-ended |
- difficulty to estimate amounts | |
Food habit |
- rapid and low cost |
- may rely on memory questionnaires |
- does not alter intake behaviour |
- may require a trained interviewer | |
- open-ended |
The following considerations will aid in selecting the method that will best meet survey objectives: the foods or nutrients of primary interest; the need for group versus individual data; the need for absolute intakes versus relative intake estimates; population characteristics (age, sex, motivation, education/literacy, cultural diversity); the time frame of interest; the level of specificity needed for describing foods; and available resources including food composition data if nutrients are to be calculated. It is important that statistical expertise be included in designing the survey and that the questions to be answered are clearly stated and analysable. Statistical resources are available in most countries and should be utilized both for survey design and data analysis.
When absolute versus relative estimates of nutrient intakes are required, the food record and 2 hour dietary recall are clearly the methods of choice for estimating mean intakes. These are the only methods that provide data on foods actually eaten, since both the food frequency questionnaire and the diet history are based on long-term subjective perception of a participant's typical eating habits. A single day of intake per study participant is adequate for estimating group means, and a representative balance of all of the days of the week should be included in the data collection, if possible. If the distribution of usual individual intakes within the groups is also needed, at least two non-consecutive days of intake per individual are required to permit estimation of within-person day-to-day variability. The combination of days of the week for each individual should be randomly assigned.
If usual intakes of each individual are required for meeting the survey objectives, the number of days of intake depends on the within-person variability for the foods or nutrients of interest and the level of precision required. A minimum of three to four days of intake is generally required for characterizing usual individual intake of energy and the macronutrients. If seasonal variability is a concern, collection of several days of intake in each season of the year is recommended.
For food components with very high within-person day-to-day variability, such as cholesterol, vitamins A and C, and alcohol, estimation of usual individual intake may require from 20 to more than 50 days of intake, which becomes impracticably long. For those highly variable nutrients contributed by relatively few food sources, a food frequency questionnaire focused on the selected nutrients is likely to be the most accurate method of assessment.
As a basis for documenting and understanding the food intake for developing FBDG, rapid methods of dietary assessment can be very helpful (8). A rapid assessment procedure survey (RAP) usually involves focus-group interviews with community leaders and for selected target groups, to gather information on food beliefs, behaviours and intakes without preconceived ideas. The trained interviewer works with the informant to create a knowledge base with minimal bias. A particular advantage of RAP methods is that they allow a more relevant food usage list to be developed for a community, which may then be used along with other more quantitative methodologies, such as for cross-checking food records or recalls or for developing FFQs. Thus, RAP methods, which are low-cost, may be used as the primary method for collecting data on which to base dietary guidelines or may be used to enhance the validity of more quantitative methods.
The management of RAP databases has been greatly facilitated by the development of computer software. Of particular note here is NUDIST, developed by social scientists at La Trobe University, Melbourne, Australia, which is used widely for the organization of sociological and anthropological data sets and which can be applied to food beliefs and habits.
Another approach to rapid assessment of food intake involves the use of a qualitative 24-hour recall to identify family and individual intakes. This method has been used in many countries, particularly in Africa for assessing infant and family feeding practices, and can be implemented by personnel with minimal training. The foods consumed the preceding day by the infant, young child and mother or family are analysed according to the age of the young child (e.g. 0-2, 3-5, 6-8, 9-11, 12-17, 18-23 and 24-35 months). The services of a statistician to plan the study are recommended. The number of times each major food and food-group are consumed per day is calculated. From this one can deduce what are the main staples and other components in the common family diet, and whether all such components (e.g. foods of animal origin, legumes, dark-green leafy vegetables, etc.) are consumed by the young children as well, or only by older children and adults. The age of introduction of various foods can also be derived.
By linking the food intake data collection with weighing/measuring of children and the mother (to derive her BMI from a standard table) and with questions on the occurrence of illnesses in the past two weeks, one can gain an overall picture of the nutritional status of the children and of adults in the community, and whether the basic problems may be related more to feeding practices, to illnesses or to general food shortage. If there are marked seasonal changes the studies should be repeated during other seasons.
This method also provides much valuable information quite easily and rapidly for workers in low-resource situations. Culturally based dietary restrictions are identified without requiring direct questions; but it is useful to follow up (e.g. through focus groups) to identify the reasons for such restrictions and how they could be circumvented. The survey thus provides a basis for drawing up locally adapted dietary guidelines, especially for young children (9).
Another use of rapid assessment methods involves assessing common practices in a population, for example, high intake of tea in a country in which iron intake is low. The tea which is regularly consumed by everyone, including children, is taken immediately after meals and thus interferes with iron absorption. To address this issue as a "food-based guidelines", it would be helpful to explore whether the general population would be willing to postpone tea-drinking for two to three hours after eating. Such information might be easily obtained from a number of short focus-group interviews conducted among the major cultural subgroups of the population.
There is often a tendency to collect more information than is needed to answer a specific question. Collecting additional data may initially appear advantageous, but experience indicates that such data are often not analysed and are therefore of little or no use. An attempt should be made to collect no more data than required to answer carefully formulated questions based on well-defined needs for the information.
Converting food consumption data to nutrient intakes requires both a food composition database and ideally computer programmes for nutrient calculation. Developing a food composition database is very expensive due to costs of chemical analysis for many foods and many nutrients. Available regional food composition data should be carefully evaluated for relevance of foods and food descriptions, as well as for accuracy and completeness of the food and nutrient data. Professionals using the compositional data need to make separate judgements for each of the values to ensure appropriate interpretation and use. Any missing foods consumed by the population or missing nutrient values for the nutrients of interest should be added to the database. If the missing values are for foods that are frequently consumed by the target population, those foods should be chemically analysed. Nutrient values for foods eaten after cooking should reflect the cooked nutrient values. If only the raw nutrient values are included in the database, the calculation software should be programmed to account for raw-to-cooked differences in yield, as well as nutrient losses due to cooking.
Computer software should be used for nutrient calculations which are tedious and prone to error if done manually. Quality control procedures for data entry should be carefully adhered to. Such software programmes already exist in many countries, and some of the programmes may be linked to other nutrient databases. Developing and testing software is time-consuming and expensive. Thus it is worth the effort required to locate and evaluate existing software for nutrient calculation. Software should be carefully evaluated, both for programme features, such as the ease of data entry, reporting capabilities and hardware requirements, as well as for the quality of the nutrient database. Recipes and other methods of food preparation should be adequately accounted for in the nutrient calculations.
Computer software for international use in food-to-nutrient conversion is limited for the purpose of calculating population estimates of intake or exposure. The very large number of nutrient analysis programmes worldwide commonly have limited database management or substitution functions. Most programmes are designed for the local dietetics market, which make them difficult to adapt to large-scale food-intake surveys and planning government interventions. Professionals can select from the half-dozen systems which do offer these functions, and choice will depend on the specific functions required by local objectives.
If FFQ data are often to be converted to nutrients, customized software may need to be developed to meet special calculation requirements where the answers to certain questions require modification of the nutrient content. For example, the answer to a question about trimming meats will require modification of the nutrient profiles for meat items.
Examples of sources that may be used for regional purposes are listed below. The publications may not be in print at the present time and may be available only through reference services.
1. Standard Reference Database (Handbook No.8); United States Department of Agriculture, Washington, DC, USA (the source most often used by national database developers to complement local food composition tables).
2. World Food Mini-List; University of California, Berkeley, USA.
3. NDS (Nutrient Data System); Nutrition Coordinating Centre, Minneapolis, USA (imported food section).
4. Food Composition and Nutrition Tables, Souci, Fachmann & Kraut; Ministry of Food, Agriculture and Forestry, Bonn, Germany.
5. McCance & Widdowson Composition of Foods; Ministry of Agriculture, Fisheries and Food, London, UK.
6. Food Tables for Malaysia; Institute for Medical Research, Kuala Lumpur, Malaysia.
7. Chinese Food Composition Tables; Chinese Academy of Preventive Medicine, Beijing, China.
8. Ethiopian Food Composition Tables; Ethiopian Nutrition Institute, Ministry of Health, Addis Ababa, Ethiopia.
9. CTA-ECSA Database for East Africa; Tanzanian Food and Nutrition Institute, Dares-Salaam, Tanzania.
10. Chilean Food tables; INTA Santiago, Chile.
11. FAO Regional Food Tables for: Latin America, Africa, Near East, East Asia; FAO, Rome.
12. Australian Food Composition Tables; National Food Authority, Australian Government, Canberra, Australia.
13. South African Food Composition Tables; Medical Research Council, Tygersberg, South Africa.
14. Food Composition Tables for Slovakia; Food Research Institute, Bratislava, Slovak Republic.
15. Food Composition Data for West Africa; CIQUAL, Paris, France.
16. Food Composition Tables for Use in the English-speaking Caribbean; Caribbean Food and Nutrition Institute, Kingston, Jamaica.
Food intake data may be analysed in many ways other than by conversion to nutrients. The possibilities for analyses by individual foods, food groups, meal patterns, and various eating practices are unlimited. The kind of analyses selected will depend on the questions of interest and the data available. In this section examples are provided of various types of analyses based on foods, food groups and eating habits that might be considered in developing or monitoring dietary guidelines.
In developing guidelines, it is often helpful to know which foods are most commonly eaten (i.e. consumed by the most people), as well as the foods eaten most frequently in a population. Information regarding the proportion of the population consuming the foods may be obtained from any one of the four general types of individual dietary data collection methods as well as from RAP methods. Household food consumption data may also be used to provide data on foods most commonly consumed, although the results will relate to frequency of consumption by the entire household rather than by individuals. Knowledge of the proportion of the population consuming specific foods is useful in developing FBDG aimed at increasing or decreasing the intake of those foods.
It is important to recognize that, with the exception of staple foods, food intake data tend to be heavily skewed. For example, in some circumstances a large proportion of the population may consume a small amount of a food, while a small proportion of the population may consume a large amount. Unless population subgroups are identified, the range of intakes will be hidden by an "average consumption figure", which in turn may erroneously be assumed to represent the intake of every individual. This is illustrated in Table 2, where the differences in yoghurt consumption in the UK by adult males of different social classes (range 90-31 g/week) are due to both the differing percentage of males in each social class consuming yoghurt, and the different mean intake of yoghurt of each group of consumers (10).
Table 2. Intakes of yogurt (g/week) in UK yoghurt consuming male subjects by social class (n=1127).
Social class |
Mean population intake (g/week) |
% who consumed the food |
Intake among consumers (g/week) |
I&II |
90 |
30 |
298 |
III-a |
37 |
22 |
254 |
III-b |
45 |
14 |
313 |
IV&V |
31 |
12 |
247 |
It is important to keep in mind that an estimate of usual individual intake is needed for determining the proportion of individuals consuming a given food with a given frequency. If food records or 24-h dietary recalls are used for this purpose, multiple days of intake per individual are required due to day-to-day variability in food intake among individuals. Food frequency questionnaires may provide a more accurate picture of the proportion of consumers of a given food.
Foods are often classified according to a limited number of groupings to facilitate dietary guidance and education of the population. Dietary recommendations may then be made based on proportional quantities of each food group. Individual or household food intake data may be analysed by food group to evaluate the overall quality of the diet. Examples of food groupings used for dietary guidance and evaluation in several countries are given in Annex 5.
A food-grouping scheme should be based on the typical eating patterns of each country and the common ways of describing foods in that population. To facilitate education and intervention, the grouping should also correspond to nutrition-related of the particular country. Food-grouping schemes also facilitate analysis of food intake data both with and without conversion of foods into nutrients. If nutrient data are available, it is useful to know which food groups are the major contributors to nutrients of particular concern. In the absence of nutrient data, food intakes may be analysed by distribution of intake among the food groups.
In some countries, analysis of food intake patterns through meals is possible because of the regularity of meal-eating occasions and the relatively narrow range of foods which constitute the meals. In the urban environment of developed countries, meal-eating patterns are generally quite varied in time as well as in the foods which constitute the meal. In such countries the term "eating occasion" may be preferable to the term "meal" or "snack" since the definition of each of these latter terms is highly subjective. Eating occasions can be so closely linked in time that it becomes difficult to say whether one is dealing with one or two separate occasions. Notwithstanding these difficulties, characterizing meal patterns or eating-occasion patterns may provide useful data for developing and evaluating dietary guidelines.
Meal-based analyses provide data on which foods are often eaten together. Such information should be taken into account in developing guidelines to ensure that targeting an increase in one food or food group does not result in a change in intake of other foods that may have a negative impact on the total diet.
A number of different methods may be used to monitor the effectiveness of an established dietary guideline. Agreement among several different types of data increases confidence in the results of the evaluation. For example, consider a guideline designed to increase calcium intake based on the results of a national survey of individual food intakes indicating that mean calcium intake was less than 50% of the recommended level, coupled with observations of a high incidence of hip fractures and osteoporosis. The "food-based guidelines" might stipulate, for example, that all healthy persons consume at least three servings of dairy products daily and three servings weekly of dark-green leafy vegetables. Subsequent monitoring of this guideline might involve comparison of the baseline calcium intake with mean calcium intake, using data from a recent national survey of individual intakes. However, due to the high cost and time involved in conducting a national survey, it is unlikely that such data would be available at the time they are needed. Therefore, another approach might be to look at national food disappearance or household food consumption data with respect to changes in the availability of the targeted foods. Similar changes might also be noted in production trends and in market research data. Finally, changes in health outcomes, such as reduced hip fractures and osteoporosis incidence, would be monitored.
In other circumstances, appropriate biomarkers can be used as indirect, but objective, measures of dietary intakes for both monitoring and assessment purposes, for instance urinary sodium and iodine concentrations.
Evaluation of the effectiveness of a specific guideline should also include an assessment of possible negative on other dietary factors. In the case of the calcium example above, one would also look at the impact of increasing intake of dairy products on fat intake if that is also a concern in the population. Common food preparation practices should also be taken into consideration in evaluating the effectiveness of a particular guideline. For example, increasing rice consumption might also increase fat intake if rice is commonly prepared or eaten with a substantial amount of added fat.
Another aspect of evaluating a dietary guideline is the proportion of the population meeting the guideline. Such data can be obtained only from dietary surveys of individuals. Collecting at least two days of intake per individual will permit estimation of the distribution of usual individual intakes within the population. Identification of characteristics of those furthest from meeting the guideline may be helpful in targeting future efforts to promote the message.
1. Food balance sheets. Agrostat PC, Food and Agriculture Organization, Rome, 1993.
2. Flores M, Nelson M. Methods for data collection at household or institutional level. In: Manual on methodology for food consumption studies. Cameron ME and van Staveren, eds, Oxford University Press,
1988. pp 33-52.
3. gingham SA et al. Methods for data collection at an individual level. In: Manual on methodology for food consumption studies. Cameron ME and van Staveren, eds, Oxford University Press, 1988, pp 53-106.
4. Block G. A review of validations of dietary assessment methods. American Journal of Epidemiology, 1982, 115:492-505.
5. Thompson FE, Byers T. Dietary Assessment Resource Manual. Journal of Nutrition, 1994, 124: 2245S-2317S.
6. Hankin JH. Dietary intake methodology. In: Monsen ER, ed. Research: Successful Approaches. Chicago: The American Dietetic Association. 1992. pp 173-194.
7. Briefel RR. Assessment of US diet in national nutrition surveys: national collaborative efforts and NHANES. American Journal of Clinical Nutrition, 1994, 59(suppl):164S-167S.
8. Scrimshaw SCM, Hurtado H. Rapid Assessment Procedures for Nutrition and Primary Health Care. UN University/UNICEF. Published by University of California Los Angeles, Los Angeles, California, USA, 1988.
9. Rapid Village Nutrition Survey Technique. WHO/Brazzaville, document AFR/NUT 84, 1977.
10. Gregory), Foster K, Tyler H. Wiseman M. The Dietary and Nutritional Survey of British Adults: A survey of the Dietary Behaviour, Nutritional Status and Blood Pressure of Adults Aged 16-64 Living in Great Britain. HMSO Publication Centre, London 1990.
