Abstract: The study was designed to provide literature overview on methodology to develop a Food frequency questionnaire (FFQ) for toddlers. Articles were retrieved from 1997 to March, 2016 with predetermined inclusion and exclusion criteria. In this study, 24 modified and 13 newly developed FFQs were included. Developing an FFQ for toddler was often performed by modifying elements of adult FFQ such as food lists, portion size and frequency of consumption. Food records/recall were commonly reported for reliability check, but none was reported for validity assessment. It is important to understand the new FFQ’s specific objective with respect to toddler’s dietary data collection.
INTRODUCTION
Food frequency questionnaires (FFQ) are dietary assessment tools which are designed to retrospectively assess habitual dietary intake over a specific reference period1. It has increasingly been used for dietary assessment in epidemiological research due to its practical advantages, such as relative low cost and respondent burden, along with a relative ease of administration and data analyses2. It can also be used to assess dietary intake by having a semi-quantitative FFQ (SQFFQ).
Toddlerhood, defined as children aged 12-36 months, is an important period of life when growth is rapid and organ, motor and cognitive abilities are still under development. Energy and nutrient requirements of these children are relatively higher to their body size. In addition, nutritional deficiencies at this age might have long-lasting effects on health3,4. Eating habits are also being developed during this stage and it may be a challenging task to maintain a balanced diet for toddlers while transitioning from an infant diet to family food5. Insights into dietary patterns and food intake during toddlerhood may be used for the development of healthy eating habits. A robust dietary assessment tools such as an FFQ could be used to provide these insights.
There is limited information on the methodological considerations for designing an FFQ especially for toddlers. Thus, it is aimed, through a literature review, to provide an evidence-based review of the principal elements and steps required to develop an FFQ for this particular age group.
MATERIALS AND METHODS
Electronic scientific databases PubMed, High Wire, Science Direct, Google Scholar, Pro Quest and Cambridge Journals were searched from 1997 to March, 2016 using the following key words: "Food frequency questionnaire", "development", "developing", "toddler", "children", "pre-school children", "dietary assessment", "nutritional status", "validation", "validity", "food intake", "portion size estimation" and "food portion". Titles and abstracts of publications in English only were scanned to assess relevance (Fig. 1).
Exclusion criteria were, no information on methods to develop an FFQ, study population was beyond toddlerhood (older than 3 years of age) and non-English publications.
| Fig. 1: | Flowchart of the review methodology |
Further searches of the reference lists of the included articles were conducted to retrieve relevant articles that were not captured in the earlier search.
RESULTS
A total of 452 articles were identified through database search. After exclusion of articles according to the exclusion criteria, 34 studies were included in this review which consists of 24 studies describing modified FFQs and 10 for newly developed ones.
The stages of developing an FFQ were identified as: conducting assessment based on study’s objective, constructing its elements which include food and beverage list, frequency of consumption categories, portion size estimation and cultural considerations, testing for its validity and reliability.
Modified FFQ: Only one of the 24 included studies, which stated modification of an FFQ, fulfilled all elements of the development stages to construct this dietary assessment (Table 1). The objective of the modified FFQs varied from assessing dietary patterns to investigating the accuracy of other questionnaires. Most of the modifications were made from an adult-version by adapting the food lists and portion size.
There were several methods reported to adapt the food list. These included updating the list from previous dietary studies or surveys, shortening the questionnaires, inclusion of commonly-consumed food for this age group based on other data sources, or inquiry about specific-nutrient nutrient-rich food such as fluoride in accordance to the objective of the study.
For adjusting the frequency of consumption, only 7 studies reported the methodology which includes previously conducted dietary surveys and assessment studies. The adaptations for the food list were reported based on original FFQ, previous dietary surveys and questionnaires or appropriateness within the reference time period for the FFQ i.e. frequency of intake of a particular food in a day or a week or a month. Food photographs and food models were the most commonly reported methods to develop portion size. Other reported methods for portion size were modification from food pyramids, household utensils or making an average serving size from established guidelines.
Multiple-days food recalls and food records were the most reported techniques to validate the modified FFQs. Other methods included weighing food records, 2-days food diaries and blood markers. The latter included erythrocyte membrane composition, hemoglobin and serum ferritin to validate iron intake. Only 7 studies reported repeatability checks at an interval between 10-15 days.
Newly developed FFQ: Only one out of 13 studies fulfilled all elements to construct the FFQ (Table 2). Four studies did not report the methods to develop the food list. The most common reported dietary assessment methods to develop the FFQ food list were 24 h food recall and food diaries. None of the retrieved studies reported methods to develop frequency of consumptions.
There were no consistent methods reported in developing portion size. Consumer research, the use of household utensils, food photographs, digital scales and general knowledge were reported.
There were no consistent dietary assessment methods reported in validity assessment which were performed in, 9 out of 13 studies. Multiple-day recalls or records and serum retinol were reported as methods to assess validity of the questionnaires. Only 3 studies reported repeatability assessments conducted within 1-2 weeks interval.
DISCUSSION
From undertaking this review, the process of constructing an FFQ are: an assessment stage of characteristics of population and time interval, followed by selecting and developing the elements such as food lists, frequency of consumption and portion sizes is similar with the one suggested by an earlier review6.
Modifying an FFQ from an existing questionnaire based on the objectives of the new study could save time and resources. Many studies have adopted adult FFQs for a younger population by modifying the food lists, reducing the portion sizes and frequency of intake6. These adaptations undertaken without validity checking could change the accuracy of the dietary assessment tools and could lead to over or under estimation of nutrient intake.
The food and beverage list is the backbone of the FFQ and directly influenced the quality of FFQ outcome measures. It is essential that the list consisted items that are truly representative of the toddler’s dietary pattern in a given population. The number of food varied from 30-130 items in the food lists. Although "24 h food recall" was the most commonly used method to build the food list, there were several ways of building food lists reported in this review which could create further concern as these methods might not be validated.
| Table 1: | Overview of studies with modified FFQ |
| Table 2: | Overview of studies with newly developed FFQ |
Focus groups and interviews could also be considered, because directly speaking to parents or primary care givers gave good insight into the eating patterns and habits of toddlers, all the types of foods and drinks consumed, the portion sizes of these foods and beverages and common utensils, cups, plates and bowls used. While food items were the focus of an FFQ, it is also important to consider beverages other than water. With the rise in the prevalence of childhood obesity, the consumption of beverages other than water could potentially contribute largely to overall nutrient intake of a small child, or displace the intake of nutritious foods.
The next step in developing FFQ is to construct the frequency of consumption. It is well-recognized that toddlers eat relatively tiny amounts, but more frequently than adults. Therefore, frequency categories should accurately capture their pattern of food and drink intake, especially for items consumed more than once a day7. In this review, there were no consistent methods reported to construct the frequency of consumption in modified FFQs and none was reported in the newly developed ones. As frequency of consumption is one of the key elements of FFQ, it is quite concerning that this element seemed not to be given adequate attention.
Portion size estimation is necessary for a semi-quantitative FFQ. Ideally, portion sizes should be the true representative amount of food consumed by a toddler. Food photographs were the most commonly reported method in this review. This technique has been considered to minimize memory bias and increase precision. However, the quality of the photograph such as lighting, distance etc were crucial factors in accurately determining portion sizes8,9. Other methods included the use of regular household utensils. Although this sounds relatively easy and more economical, there were no common standards for these household utensils. For example, the volume of a child’s cup varied from one brand or type to another and the capacity of a child cup was typically much less than a standard cup (250 mL). In this case, the FFQ would require extra information in the form of photographs and measurements for example, placing a ruler next to a cup or bowl to provide scale and to capture the dimensions of utensils, cups, bowls and spoons. This would ensure that the participant and investigator are both referring to the same size or volume to avoid errors in estimation.
Dietary guidelines, national dietary data, or food composition tables could also potentially be used to construct portion sizes9,10. For example, a serving size for toddler could be around 1/3 to 1/2 of adults’ recommendations. However, this method has been reported to result in poor accuracy of the portion sizes. A dietary guideline could recommend 1 serve of meat but a "serve" may not be understood by the individual completing the questionnaire7,10,11. In this review, there were limited information on the accuracy and precision of one method (i.e. food photographs) as compared to the other method (i.e. household utensils) in estimating portion size. There are several methods to develop food portion estimation with various degrees of accuracy and easiness. Use of household utensils is the most common method due to its easy availability.
The final stages of the FFQ development involve testing for validity and repeatability of the draft questionnaire to avoid biased results and inappropriate associations6. There is no gold standard for assessed the validity of the FFQ. A newly developed dietary assessment tool can be validated against any other method, taking into account that method’s limitations the objective of the study. From the retrieved literature, when validated against weighed food record, the FFQ tended to overestimate nutrient intake among toddlers8,12,13. This could be because weighed food records are accurate, since the items that were consumed, were measured. Another possibility is to validate against biomarkers which could be appropriate when only a few nutrients or single nutrient was of interest12,14. However, errors could occur due to bioavailability, absorption and metabolism of a particular nutrient. Another possible route reported in the review is to validate against two dietary methods. For example, a combination of 24 h recall and 3 days estimated dietary record for FFQ validation had also been used in past.
Repeatability tests are need to check the results consistency, with the time interval between the two assessments should not exceed 5 weeks due to the greatly varying diet of toddlers7,10 and to be conducted in separate study population to lighten the participants’ burden6. A repeatability test with a correlation coefficient range between 0.5-0.7 is considered as acceptable6,15. Pre-testing the final questionnaire to the similar target population of the new study is advisable to reduce other potential biases. In this review, there were very limited studies reporting the repeatability check. In more recently published studies, repeatability check were conducted within 1-2 weeks interval which could be an ideal period for toddlers16,17.
CONCLUSION
Food frequency questionnaires (FFQ) has increasingly been used in epidemiological research due to its practical advantages. The study objective with respect to the dietary data collection is the most important consideration when building a robust FFQ. This review provides evidence-based guidance and considerations when building a suitable FFQ for a study among toddlers.
SIGNIFICANCE STATEMENT
In this study, it is reported on the steps and components required to develop a food frequency questionnaire (FFQ) for toddlers. This is significant because FFQ have increasingly been used for dietary assessment in epidemiological research owning to its practical advantages. Therefore, it must be carefully designed to capture all the dietary information especially for toddler as they have very different dietary patterns and food choices as compared to adults. This narrative could serve as an evidence-based overview on how to develop FFQ for toddlers. This could be useful for clinical nutritionists, dieticians, paediatric healthcare professionals and researchers in clinical settings and applied research.