Low accuracy and low consistency of fourth-graders' school breakfast and school lunch recalls

Abstract

Objective: To determine the accuracy and consistency of fourth-graders' school breakfast and school lunch recalls obtained during 24-hour recalls and compared with observed intake.

Design: Children were interviewed using a multiple-pass protocol at school the morning after being observed eating school breakfast and school lunch.

Subjects: 104 children stratified by ethnicity (African-American, white) and gender were randomly selected and interviewed up to 3 times each with 4 to 14 weeks between each interview.

Statistical analysis: Match, omission, and intrusion rates to determine accuracy of reporting items; arithmetic and/or absolute differences to determine accuracy for reporting amounts; total inaccuracy to determine inaccuracy for reporting items and amounts combined; intraclass correlation coefficients (ICC) to determine consistency.

Results: Means were 51% for omission rate, 39% for intrusion rate, and 7.1 servings for total inaccuracy. Total inaccuracy decreased significantly from the first to the third recall (P=0.006). The ICC was 0.29 for total inaccuracy and 0.15 for omission rate. For all meal components except bread/grain and beverage, there were more omissions than intrusions. Mean arithmetic and absolute differences per serving in amount reported for matches were -0.08 and 0.24, respectively. Mean amounts per serving of omissions and intrusions were 0.86 and 0.80, respectively.

Applications/conclusions: The low accuracy and low consistency of children's recalls from this study raise concerns regarding the current uses of dietary recalls obtained from children. To improve the accuracy and consistency of children's dietary recalls, validation studies are needed to determine the best way(s) to interview children.

FIG 1

Description of multiple-pass protocol used for interviews.

FIG 2

Definitions. ^aA statistical weight was assigned to each item (match, omission, or intrusion) by meal component [entree, combination entree, bread/grain, vegetable, fruit, dessert, condiment, beverage, breakfast meat, and miscellaneous (foods such as candy and chips that did not fit into any of the other components)]. Statistical weights were combination entree (eg, taco salad, sausage biscuit) = 2, condiment (eg, ketchup) = 0.33, and remaining meal components = 1. Statistical weights were used so that errors regarding reporting combination entrees counted more than errors for condiments and remaining meal components.

FIG 3

Distribution of omission rates (section A) and intrusion rates (section B) across all 275 recalls from 104 children.