Validation of Children's School Lunch Recalls: Comparing Accuracy of Nutrients Versus Foods

Abstract

Children often provide self-reports of diet for nutrition studies; however, the validity of such reports remains in question. Although validation studies usually present results for “accuracy of reporting” various nutrients, people report what they ate as foods, not nutrients. We used observations to validate 4th-graders’ school lunch recalls to 1) determine accuracy at the nutrient level, 2) determine whether accuracy at the nutrient level differed for same-day versus next-day recalls, and 3) contrast accuracy at the nutrient level with accuracy at the food level (as determined previously). A total of 237 students (179 Black, 58 White; 120 male) from 4 schools were randomly selected, observed eating school lunch, and interviewed on either the same (n=89) or next day (n=148). Levels for energy, carbohydrate, fat, protein, calcium, sodium, cholesterol, saturated fat, and fiber were determined using Nutrition Data System (University of Minnesota). A 2 by 2 MANOVA indicated p < 0.001 for the main effect of recalled versus observed, p > 0.13 for the main effect of same-day versus next-day, and p > 0.47 for the interaction. These results are in contrast to results from previous analyses (JADA 1997; 97:1293–1298) at the food level for this group which indicated that differences between recalled and observed foods were larger for next-day than for same-day recalls (ANOVA, p's < 0.05). The previous results indicated that match rates (for foods observed and reported eaten) were 84% and 68% for same- and next-day recalls, respectively, and phantom rates (for foods not observed but reported eaten) were 5% and 13%, respectively; however, accuracy of amount reported eaten (which was assessed only for matched foods) was high for both same- and next-day recalls. Thus, the time interval between eating and reporting appears to be important when assessing accuracy of children's self-reports of school lunch at the food level, but not at the nutrient or amount levels. This may be because analyses at the nutrient level allows for substitutions of one food for another, as well as over- and under-reporting of amounts, which may falsely inflate accuracy levels for some nutrients but not others. Research to improve the accuracy of children's self-reports of diet should analyze at the food and amount levels, instead of only at the nutrient level.