Selecting the most appropriate measure for assessing dietary behavior requires accounting for multiple considerations in terms of what data are needed and how they will be used. To help address these aspects, the following guiding questions are suggested. These questions are informed by those posed by Sternfeld and Goldman-Rosas¹¹⁷ to guide the selection of appropriate measures for physical activity and sedentary behavior.

1. What is the primary research aim or question?

   Clearly defining the aim or question from the outset is critical to ensuring alignment with the measure of dietary behavior (and measures for other variables as well). Studies related to dietary behavior may be intended for surveillance or monitoring purposes, such as to estimate the frequency with which preschoolers consume juice, assess the usual intake of sugar-sweetened beverages among school-age children, or characterize perceptions or attitudes toward food and food consumption. Further, epidemiologic studies may be undertaken to examine relationships between dietary behaviors, conceptualized as exposures, and subsequent outcomes relevant to childhood obesity, while intervention studies aim to assess the impact of a given strategy to shift eating patterns.

2. Is the target population made up of infants, toddlers, school-age children, and/or adolescents?

   Age has an important bearing on cognitive abilities, as well as memory and concept of time, affecting whether it is possible to administer measures directly to children or whether parents or other proxy reporters need to be included. If a study will include children of various ages, considerations will be needed to tailor tools or methods of administration depending on literacy, numeracy, attention span, and related factors. For example, within NHANES, dietary recalls are proxy-reported, proxy-assisted, or completed by children independently, depending on age. Body weight is an important consideration given evidence that body mass index is a strong indicator of error in reporting of dietary intake. Characteristics that may be associated with body mass index, such as education and race and ethnicity, also should be considered.

3. What is the study design?

   Are the measures administered at one time or multiple times? Are measures collected cross-sectionally, prospectively, or retrospectively? For retrospective studies, food frequency questionnaires or screeners are the only possible self-report measures for dietary intake. Attention should be paid to potential sources of error, including the possibility that current intake may bias intake recalled for the past.

4. In what setting will the research take place and how will this affect possibilities for data collection?

   For example, it may not be possible to conduct rigorous observation or collect detailed intake data within a classroom setting depending on the number of children and the available resources for data collectors, though this could be alleviated with web-based tools, depending on the age and cognitive abilities of the children. If data are being collected online or by mail, this introduces complexities in terms of the possible measures that can be used.

5. What settings are of interest?

   For example, dietary behaviors at home, away from home, or at school? Do these need to be differentiated to address the research question, and how does this impact upon the measurement needs?

6. Is the focus on the total diet or on specific dietary components of interest?

   For example, does the research question pertain to total diet and/or diet quality or patterns or specific components of diet, such as foods or food groups, beverages, or particular nutrients? Are the components of interest episodically or non-episodically consumed by most in the population of interest?

7. Is diet an independent variable, dependent variable, or covariate?

   This has implications in terms of the effects of error and strategies to mitigate it, as alluded to in Section 8. Is particular statistical expertise required and if so, is it available?

8. What parameters are of interest?

   For example, estimates of intake, frequency of consumption, or some other behavior, such as snacking or characterization of the sources of nutrient-dense food consumption. Relatedly, what are the desired summary measures? For example, are mean usual intakes of fruits and vegetables among a group of interest, or is there a desire to estimate proportions above or below recommendations? What are the corresponding implications for data collection and analysis (e.g., the need for repeat short-term measures to estimate distributions of usual intake for the purposes of assessing the prevalence of a group with intakes above or below a recommendation)?

9. Are related or complementary measures necessary to address the research question?

   For example, are other dietary components (besides those targeted) of interest from the perspective of compensatory effects or trade-offs (e.g., in the case of interventions to reduce sugar-sweetened beverages)? Are other dietary behaviors aside from intake (e.g., those related to food preferences and attitudes) relevant to the interpretation of dietary intake data and to addressing the research question? Are other behaviors outside of diet, such as physical activity, important to addressing the research aim (e.g., are alterations in diet offset in some way by changes in activity level)?

10. What are the logistical considerations and constraints, such as the time allocated to the measurement of dietary behaviors within a larger study, as well as budget and expertise?

    Following Sternfeld and Goldman-Rosa,¹¹⁷ this question is deliberately placed last in the list to encourage researchers to weigh logistical issues no more heavily than the considerations listed above. In other words, all salient details should be considered so that the final choice reflects the best possible measure given all factors at play.

Considering the questions above can help to narrow down the measures within the Registry to those consistent with the research aim, population, and dietary behaviors of interest. The Measures Registry provides information that can be easily scanned, with organization into sections providing an “At A Glance” overview of each measure, along with information on the design of the study in which a measure was developed or used, tips on how to use the measure, and as mentioned previously, a summary of available evidence on validity and reliability. It is also possible to compare across measures. After narrowing down the choice of measures to those best aligned with the research needs, the researcher can explore other resources, including citations listed in the Registry, in depth to consider fit in terms of specific dietary behaviors, the specifics of evaluations of psychometric properties, as well as whether the tools need tailoring and further evaluation prior to use with the target population. In cases in which tools are adapted, a decision needs to be made as to whether the tool has been changed to the extent that it requires further evaluation for validity and reliability. These details should then be included in any publications reporting on the use of the measure to allow for appropriate interpretation.