On the Accuracy of Self-Report Instruments for Measuring Food Consumption in the School Setting

Dear Editor:

Children’s reporting of diet is theorized to follow a complex set of cognitive processes (1). Because of the complexity of this task, previous researchers have concluded that self-report methods have insufficient validity for use among young children or, as detractors say, those of any age (for perspectives, see 2–3). The systematic review conducted by Tugault-Lafleur et al. (4) in a recent issue of the Journal similarly concluded that the relative accuracy of self-report dietary assessment methods is poor among children in the school setting. However, there are important caveats for the use of such methods, as well as additional pieces of evidence missing from the review.

The first caveat is that these self-report methods are not designed to estimate usual intakes. Given that there are no estimates of intraindividual variability in the consumption of foods at school, there is limited ability to understand how well these methods approximate usual intakes. Instead, self-report methods are generally used to compare intervention groups or examine factors believed to influence consumption.

Second, there are numerous outcomes that can be sought from self-report methods. For example, an evaluator might consider as an outcome selection (which is pertinent in the context of the offer-versus-serve provision in the United States), consumption of foods, or consumption of nutrients. These outcomes might be summed across a whole meal or only related to specific food groups (e.g., vegetables). Researchers evaluating the consumption of nutrients should proceed with caution, because this outcome is assessed indirectly across all dietary assessment methods, making chance fluctuations in accuracy more likely (5). In contrast, evaluating the selection and consumption of food items allows researchers to understand sources of systematic error directly influencing accuracy. The present review does not disaggregate methodologic studies according to differences in outcomes, and several included studies use subjective weights for meal components (e.g., entrées are weighted more heavily than side items), making it difficult for researchers to understand accuracy specific to their outcome of interest.

Additional references examining the accuracy of self-report methods for fruit and vegetable consumption outcomes are shown in Table 1. These studies suggest that self-report questionnaire methods may be accurate, reliable, and sensitive to detect between-group differences in fruit and vegetable selection and, to a lesser extent, amounts consumed. Although not all of these studies reach the 85% match rate suggested by Tugault-Lafleur et al. (4), this threshold may be too stringent given that percentage agreement is generally more conservative than Pearson correlation or the κ statistic. The moderate accuracy of self-report methods may be sufficient for the purpose of testing for differences between intervention groups.

TABLE 1.

Self-report methods used to assess fruit and vegetable consumption among children in school settings¹

				Results
First author, year (ref), location	Sample	Instrument	Reference method	Reliability	Validity
Economos, 2008 (6), USA	75 students aged 8.3 ± 1.12 y	Fruit and Vegetable Questionnaire	Direct observation	Test-retest of the questionnaire for specific fruit and vegetable items ranged from 93% to 96%	Items on tray for specific fruits and vegetables ranged from 61% to 78% agreement
Wallen, 2011 (7), USA	125 students; fourth grade; 2 schools	DILQ-CO	Weighed plate waste	—	Items on tray: match rates were 87% for fruit and 88% for vegetables; amount eaten: match rates were 58% for fruit and 47% for vegetables
Weber, 2004 (8), USA	80 students; third grade; 1 school	24-h recall	Direct observation	—	Items on tray: match rates were 75% for vegetables and 86% for fruit
Harrington, 2009 (9), USA	396 students; fourth grade; 33 schools	24-h recall	Direct observation	—	Items on tray: match rates were 72.3% for fruit and 47.7% for vegetables; amount eaten: fruit, 54.4% agreement; vegetables, 40.4% agreement

¹DILQ-CO, Day in the Life Questionnaire–Colorado; ref, reference.

²Values are means ± SDs.

Given these results, there is a need to understand the conditions under which children are most accurate in their self-reports. For example, shorter retention intervals, meal-specific prompts, and forward reporting order have been shown to improve the accuracy of self-reports (10). In addition, there may be differences according to the preparation of foods and the number of items offered. Harrington et al. (9) found that children were more accurate in reporting lunch consumption when fewer items were available to them. Continued innovation is necessary for developing or improving self-report administration methods with greater accuracy. To accomplish this, assessing and reporting the accuracy of not only self-reported intakes of nutrients but also that of food items is necessary.