Effect of manual data cleaning on nutrient intakes using the automated self-administered 24-hour dietary assessment tool (ASA24)

Authors: BOUZID, YASMINE - University Of California, Davis; ARSENAULT, JOANNE - University Of California, Davis; BONNEL, ELLEN - University Of California, Davis; CERVANTES, EDUARDO - University Of California, Davis; KAN, ANNIE - University Of California, Davis; BURNETT, DUSTIN - University Of California, Davis; Keim, Nancy; Lemay, Danielle; Stephensen, Charles

Submitted to: Current Developments in Nutrition
Publication Type: Abstract Only
Publication Acceptance Date: 1/26/2021
Publication Date: 2/2/2021
Citation: Bouzid, Y.Y., Arsenault, J.E., Bonnel, E.L., Cervantes, E., Kan, A., Burnett, D.J., Keim, N.L., Lemay, D.G., Stephensen, C.B. 2021. Effect of manual data cleaning on nutrient intakes using the automated self-administered 24-hour dietary assessment tool (ASA24) . Current Developments in Nutrition. 5/3. https://doi.org/10.1093/cdn/nzab005.
DOI: https://doi.org/10.1093/cdn/nzab005

Interpretive Summary:

Technical Abstract: Objectives: To characterize the effect of manual data cleaning on individual dietary intake assessment using Automated Self-Administered 24-hour recalls® (ASA24s). We aimed to determine whether there was a difference in measures of key nutrients (energy, carbohydrate, fat, protein, fiber) before and after manual correction of automated coding errors. Methods: Dietary recalls up to 4 days were collected using ASA24. Data cleaning was completed by systematically reviewing and excluding recalls with technical errors identified by a quality control team. A registered dietitian reviewed open-ended text responses and corrected nutrient data for foods where automated coding did not produce an optimal match. Total energy expenditure (TEE) was calculated using measurements of physical activity, resting metabolic rate, and estimated thermic effect of food. The Wilcoxon signed rank test was used to determine differences in median nutrient measures. Mean energy intake was compared to TEE using Spearman correlation. Results: After removing recalls that were incomplete or with technical errors, 1499 recalls were completed by 393 subjects. Recalls modified during the data cleaning process constitute 209 (14%) from 140 subjects with at least 1 recall modified (36%). Using all recalls, median intakes of energy, total fat, protein, and fiber were different after data cleaning compared to before (p < 0.05). When limiting to recalls modified during cleaning (n = 209), significant differences were found for energy, fat, protein, and fiber (p < 0.02). Using mean nutrient intakes per subjects with at least 2 valid recalls (n = 385), there were no significant differences before and after cleaning (p > 0.05). However, for the 140 subjects with at least one modified recall, significant differences were found for energy, fat, and protein of mean data (p < 0.02). The association between mean energy and measured TEE was higher, but not significantly, for the cleaned data compared to uncleaned. Conclusions: Data cleaning can significantly change nutrient values recorded in 24-hour recalls. For assessment of individual dietary intake, investigators should evaluate open-ended text responses in ASA24 and consider correcting data based on nutrients of interest. Funding Sources: USDA