|Godwin, Sandria - TENNESSEE STATE UNIV.|
|Chambers Iv, Edgar - KANSAS STATE UNIV.|
Submitted to: Journal of American Dietetic Association
Publication Type: Peer Reviewed Journal
Publication Acceptance Date: December 20, 2003
Publication Date: April 1, 2004
Citation: Godwin, S.L., Chambers IV, E., Cleveland, L. 2004. Accuracy of reporting dietary intake using various portion-size aids in-person and via telephone. Journal of American Dietetic Association. 104(4):585-594. Interpretive Summary: Government agencies, the food industry, and academic groups use data from food consumption surveys to establish nutrition-related policies and programs, identify needs for new food products, and develop education and marketing strategies. Thus, it is essential that the survey data be as accurate as possible. One potential source of error relates to peoples' ability to estimate the sizes of food portions they eat. A variety of portion size estimation aids (PSEAs) have been developed for different foods and interview modes. However, little research has been conducted to test PSEA accuracy or to compare PSEAs in different interview situations. The objective of this study was to compare the accuracy with which respondents reported food intake using 2-dimensional (2D) versus 3-dimensional (3D) PSEAs similar to those used in national food consumption surveys and other types of dietary assessment, and to determine if interview mode affected accuracy of reporting with the 2D aids. Mean misestimation averaged over all types of foods and PSEAs was within ±20%. However, fewer than one-third of subjects correctly estimated intake for any specific food within ±20%. Estimates with the 3D aids were somewhat more accurate than those with the 2D aids. The authors suggest that adding perspective to the drawings of the 2D aids could improve accuracy of reporting with them. Generally, there were not significant differences in accuracy when using 2D aids in interviews conducted in-person versus by telephone. The results suggest that the 2D and 3D PSEAs used in this study are adequate for making population estimates of intake, but not for assessing the dietary intake of specific individuals. The results are important in understanding sources of misestimation in dietary surveys, and in promoting efforts to improve existing PSEAs for greater accuracy.
Technical Abstract: The primary objectives of this study were to compare the accuracy with which people report food intake using different portion-size estimation aids (PSEAs) during interviews administered in-person and by telephone. The subjects were 120 males and females 18-65 years old who were recruited through advertisements and contacts at churches and universities in Nashville, Tennessee. Subjects ate lunch after selecting foods from a buffet line of pre-weighed items. Known amounts eaten by subjects were compared to amounts they reported in 24-hour dietary recalls conducted in-person or by telephone during which either 2-dimensional (2D) or 3-dimensional (3D) PSEAs were used. Subjects were randomly assigned to one of four interview groups: guided, in-person interview with 2D aids; guided, telephone interview with 2D aids; guided, in-person interview with 3D aids; or unguided, in-person interviews with 3D aids. Analysis of variance and least significant differences were calculated for mean percent estimation error. Frequencies and Pearson correlation coefficients were used to determine percentage misestimation and relationships between estimations, perceived healthfulness of the food, and subject's confidence in the estimate. In general, accuracy of reporting food amounts was not significantly different for the different types of PSEAs or mode of interview. Mean misestimation averaged over all types of foods and PSEAs was within ±20%. However, fewer than one-third of subjects correctly estimated intake within ±20% for any specific food. Subjects generally overestimated intake of solids (meat, cake) and liquids (tea, water, salad dressing) and both overestimated and underestimated intake of amorphous (mounding) foods (ice cream, macaroni and cheese, green beans, salad, french fries). On an individual basis, no type of food (solid, liquid, amorphous) was correctly estimated more often than another. Misestimation was unrelated to perceived healthfulness of the food or to respondent confidence in the estimate.