School Age Populations Methods Validation - NCS Dietary Assessment Literature Review
On this page...
Two recent comprehensive reviews examined the validation of dietary assessment methods in school age children through 2000 (1;161). This report summarizes the conclusions of these two independent reviews and then discusses validation studies published since 2000.
McPherson et al. (1) reviewed 41 validity and 9 reliability studies of children age 5 to 18 years that included a sample size of at least 30. These studies were published in peer-reviewed English-language journals between 1970 and August 2000. About two-thirds of the studies examined school age children (Exhibit 5.2). This paper also discusses studies on adolescent populations. The full review article includes a series of tables summarizing the relevant design issues and the results of the examined studies. The authors concluded that the evaluation of dietary assessment methods for school age children is difficult for several reasons:
- The reference methods against which the instruments have been compared have inconsistent validity.
- The effect of correlated errors between the method being examined and the reference method must be considered.
- Different reference periods were often used for the examined instrument and reference method.
- The heterogeneity of the study designs, study populations, and instruments (e.g., different food frequency questionnaires, the number of days and recording method for food records, or one meal versus 24-hour recall) makes comparisons uncertain both within a type of assessment method and between methods (1).1
With these considerations in mind, the studies reviewed showed that correlations between the reference method and dietary assessment tool were almost always higher for the food recalls and food records than for the FFQs (Food Frequency Questionnaire) (1). Almost all of the reviewed validity and reliability studies among children younger than age 9 years included adult assistance providing information on the child's intake. Little data on the effect of gender, race, or ethnicity on instrument validity were found. The review concurred with another recent review of portion size estimating aides (126), which concluded that children generally have difficulty estimating portion sizes and insufficient information is available to make guidelines for portion size estimation in children.
The review by Livingston and Robson (161) evaluated key measurement issues in the dietary assessment of children and adolescents and examined the issues of misreporting and the identification of misreporters. This review includes a table of studies validating dietary assessment methods with total energy expenditure (TEE) measured by doubly labeled water (DLW) in school age children (7 studies) and adolescents (4 studies). The review concludes with a discussion of criteria to consider in selecting a dietary survey method in young children and older children and adolescents. Exhibit 5.3 summarizes the relevant conclusions of this review.
Table 5.1 presents 17 validation studies on school age children published since 2000.
Food Records (FRs)
Two studies on school age children age 6 to 9 years found close agreement between mean group energy intake reported on either estimated or weighed FRs and TEE measured by the DLW method (197;198). In both studies, reported energy intake was not representative of TEE at the individual level. The study of predominately white Northern Ireland children found obese children underreported energy intake by 14% (197), and the study on Australian children of mixed socioeconomic status found no relationship between misreporting and body mass index (198).
Livingston and colleagues (199) re-examined a dataset of concurrent measurements of TEE by the DLW method, basal metabolic rate (BMR), heart rate monitoring, 7-day weighed food records (FRs), and diet history (DH) in 36 children age 7 to 15 years to identify children as underreporters, acceptable reporters, and overreporters. At least 80% of the children were classified as acceptable reporters by both FR and DH. Underreporting was more prevalent by the weighed FR (11% weighed FR vs. 6% DH), and overreporting was more prevalent by DH (17% DH vs. 6% weighed FR). Only 25% of children who underreported energy intake on weighed FRs were identified by cut-offs based on age-specific physical activity levels (PAL), leading the authors to conclude that "all cut-offs based on assumed PAL levels used for screening the energy intake data of children should be applied with caution" (199). Though this analysis provided evidence that estimating physical activity levels by heart rate monitoring shows promise, its sensitivity and specificity for detecting misreporters needs further study.
24-Hour Recall (24HR)
Since 2000, seven studies have examined the accuracy of the 24HR in school age children. Two studies (200;201) found close (within 10%) agreement at the group level between mean energy intake reported on three multiple pass 24HRs and TEE by the DLW method, but wide individual variability. Evaluation of reporting accuracy in a diverse free-living population of children classified as Tanner index stage 1 (pre-puberty) found that children who underreported intake had a higher relative weight and adiposity than overreporters (p<.0001), and overreporters were lighter and had less body fat than under- and accurate reporters (201). In addition, 46% of children reported intakes greater than 110% of TEE measurements, raising the possibility that overreporting may constitute a major form of children's and parents' inaccurate dietary reporting or that "overreporting" reflected actual child energy intake contributing to the increasing prevalence of overweight among children.
Five studies validated 24HR methods with direct observation of school meals focusing on accuracy of recalled foods rather than total energy or nutrient intake. Children 5 to 7 years in the United Kingdom were able to accurately recall only 58% of food eaten from the school lunch menu or 70% of packed lunch items within 2 hours of finishing the meal (202). In this study, non-directed prompts increased recall by 66 to 80%, but the authors concluded that the ability of children age 5 to 7 years to recall intake from lunch at school varies widely and that this dietary assessment method is unlikely to be suitable at the individual level.
Baranowski and colleagues (203) developed and evaluated an interactive multimedia, and multiple pass 24HR for older students to self-report dietary intake. This system is called the Food Intake Recording Software System (FIRSSt). Compared with observation of intake at the school lunch meal the previous day in a diverse group of 4th grade (age 9 to 11 years) public school students, FIRSSt was somewhat less accurate (46% matches of observed and recalled foods) than a dietitian administered multiple pass 24HR (59% matches of observed and recalled foods). However, FIRSSt attained 60% match, 15% intrusion (added foods), and 24% omission rates for all meals when compared with a 24HR. This study is also the first study to evaluate a bogus pipeline effect for diet in children. Children were asked to provide a hair sample and told, "We can tell some of what you eat from a chemical analysis of your hair." Obtaining a hair sample reduced the omission rate for FIRSSt versus 24HR and increased the match rate for 24HR versus observation. This finding suggests that a small part of the inaccuracy of children's self report is wished for or willful and thereby subject to correction by a bogus pipeline procedure (203).
Domel-Baxter and colleagues conducted a series of 24HR methodology and validation studies in diverse groups of public school students in Georgia that question the ability of 10-year olds to accurately report intake. A study of fourth grade students (10 years old) found no difference in 24HR omission rates and total inaccuracy between telephone-administered 24HRs or in-person 24HRs. Whether interviewed in person or by telephone using a 4-pass method, children reported only 67% of items observed at school breakfast and lunch the previous day, and 17% of items reported were not observed (204). Fourth graders in a second study (205) completed a dietitian administered 4-pass 24HR the day after observation of three school breakfasts and lunches. Less than half of the items observed eaten were reported accurately, and almost 40% of what the students reported eating was not observed. Furthermore, accuracy was inconsistent from one recall to another in the same child.
In a study of first grade (mean age 7.2 years) and fourth grade (mean age 10.1 years) students interviewed the morning after observed intake at school lunch (206), specific prompting methods (preference, food category, or visual) decreased the accuracy of intake reporting among first graders. Among fourth graders, prompting for food category yielded small gains in recall accuracy.
Food Frequency Questionnaires (FFQs)
Three studies evaluated three different FFQs in populations including children age 11 and 12 years. The Eating Habits Questionnaire adapted from the Health Habits Questionnaire used in the Bogalusa Heart study was validated in diverse group of 24 students (mean age 12.7 years) using three 24HR recalls as the reference measurement (207). Although perfect agreement between instruments was achieved for only 56% of the food categories, factor analysis suggested that 10 factors explained 81.3% of the variance. The investigators concluded the instrument is a valid instrument in African American and white adolescents in the Southeast.
Validation of the Youth Adolescent FFQ with TEE measurements by the DLW method found the instrument provided accurate estimates of mean energy intakes of prepubertal boys and girls as a group, but not for individuals (208). Fifty percent of the subjects misreported intake (difference >10%); there was an inverse relation between the energy discrepancy and both body weight and percentage of body fat.
Agreement between the European Prospective Investigation of Cancer (EPIC) FFQ and 7-day weighed FRs for energy and all nutrients was poor at both the group and individual level for Scottish school children (mean age 12.3 years) (209). The FFQ did correctly classified low, medium, and high intake consumers.
Other Validation Studies
Two studies examined portion size estimating methods by comparing estimates with direct observation of intake. The magnitude of errors in the children's quantitative estimates of food portions were large when using manipulative props (modeling clay, small plastic beads, paper strips, and water) (210), 2-dimensional food models (210), food photographs (210;211), or descriptions of food portions as small, medium, or large (210;211).