If nutrient intakes are to be estimated from the questionnaire, the first consideration is which nutrients are to be assessed and whether a questionnaire is appropriate for these nutrients.

Food intake questionnaires have been widely used in large-scale population surveys. They have several advantages over other dietary survey methods:

• Data can be collected quickly and cheaply
• Data can be collected from many geographical areas
• Observer bias is minimised
• Data are obtained in a form convenient for coding and statistical analysis

The dietary information tends to be less precise than that obtained in other ways. However, it may be more representative of people's usual intake than that derived from a detailed study of a few days' food consumption. Some nutrients are much easier to study in this way than others. For example, intake of fibre or vitamin C can be more readily estimated by questionnaire than, say, fat intake. This is because fibre and vitamin C are present in relatively few foods and people can estimate how often they eat these foods reasonably well. Day to day variation in intake of a nutrient is also a factor in how well that nutrient can be estimated by questionnaire - in general,intakes of nutrients that do not vary greatly from day to day can be more readily estimated than those with large day to day variation.

A list of all the nutrients of interest should be made initially. This is important as a questionnaire designed to assess, say, fibre and sugar intakes could not be used at a later date to calculate intakes of other nutrients if food sources of those nutrients are not the same, e.g. fat.

The food sources of each nutrient within the population being studied should then be considered. Criteria for the selection of foods are that they should be eaten reasonably often by an appreciable number of individuals and that they should have a substantial content of the nutrient(s) of interest. If the purpose of the survey requires that subjects be classified into tertiles or quintiles according to their nutrient intakes then, in order to differentiate between individuals, foods should be included for which there is likely to be reasonable variation in consumption between individuals. If the questionnaire is to be used to estimate mean nutrient intakes of the population, or to compare the mean intakes of subgroups, the foods selected for the questionnaire should cover at least 90-95% of each nutrient to be calculated. If a lower proportion of the total intake is covered then group mean intakes may be substantially underestimated.

Consideration should be given to the grouping of foods within the questionnaire. For example, should "tomatoes" be asked in a separate question or should this be included with "salads"? Should a question be asked about "beef, all forms including minced beef" or should there be a number of questions: "roast beef", "beef stew", "minced beef", "beef burgers" etc.? Appropriate groupings will be dependent upon a number of factors:

• The hypotheses being tested
• How similar the nutrient compositions of the foods are
• The likely ability of subjects to assess how often they eat the foods

People may, for example, be better able to estimate how often they have "meals containing beef" than to estimate separately the frequency of consumption for each form of beef.

Another consideration is whether subjects may count the same food twice when completing the questionnaire. For example, if there is a question about "beef lasagne" and another question about "pasta", does "pasta" include or exclude the pasta accounted for in "beef lasagne"?