Manual data cleaning modifications can significantly change nutrient values recorded in automated self-administered 24-hour recalls (ASA24)

Authors: BOUZID, YASMINE - University Of California, Davis; ARSENAULT, JOANNE - University Of California, Davis; Bonnel, Ellen; Cervantes, Eduardo; KAN, ANNIE - University Of California, Davis; BURNETT, DUSTIN - University Of California, Davis; Keim, Nancy; Lemay, Danielle; Stephensen, Charles

Submitted to: Journal of Nutrition
Publication Type: Peer Reviewed Journal
Publication Acceptance Date: 1/21/2021
Publication Date: N/A
Citation: N/A

Interpretive Summary: Use of dietary assessment tools is an important component of nutrition research, allowing nutrition researchers to determine the customary dietary intakes of volunteers living their usual lives, eating at home, in restaurants or in other settings. One of these tools, the ASA24, is designed to collect intake for a 24 hour period via an automated, online process without the assistance of a nutritionist or dietician. This tool was designed not to require volunteers to be interviewed by a nutritionist and thus is less costly and better suited to large studies. However, the automated nature of data collection does leave some uncertainty about the correct interpretation of text responses from volunteers by the ASA24 software. In this paper, we evaluated how interpretation of these text fields by a dietician might or might not improve the overall quality of the information collected, rather than relying solely on the ASA24 software. During this analysis, we found significant differences before and after reinterpretation of these text fields when calculating daily intake for energy, carbohydrate, total fat, and proteins. We thus conclude that these reinterpretations made during data cleaning can change mean nutrient intake measures for the overall group of volunteers being studied. Similarly, if researchers are interested in using data for specific individuals (e.g., to find associations with health outcomes) this reinterpretation of text fields may improve the quality of their analyses. Thus nutrition researchers should consider nutrients of interest and individual versus group intake assessment when deciding whether to make cleaning modifications for their studies as this process requires a significant amount of additional time after the data are collected from volunteers.

Technical Abstract: Automated dietary assessment tools such as ASA24 are useful for collecting 24-hour recall data in large-scale studies. Modifications made during manual data cleaning may affect nutrient measures. We evaluated the effects of modifications made during manual data cleaning on nutrients measures of interest: energy, carbohydrate, total fat, protein, and fiber. Differences in mean intake before and after data cleaning modifications for all recalls and average measures per subject were analyzed by paired t-tests. Percentage of change for each nutrient was calculated for all recalls. Chi-squared test was used to determine whether unassisted recalls had more open-ended text responses that required modification than assisted recalls. We characterized food types of text response modifications. Mixed linear models were used to assess associations of cleaning and participant characteristics (age, sex, and BMI) on nutrient measures. Total energy expenditure (TEE) was correlated with mean energy intake from raw and modified data. We found significant differences before and after modifications for energy, carbohydrate, total fat, and protein measures for all recalls. Limiting to modified recalls, there were significant differences for all nutrients of interest, including fiber. There was not a significantly greater proportion of text responses requiring modification for home recalls compared to training. While there was a trend higher correlation of mean energy intake with TEE for modified compared to raw data, it was not significant. Mean intake for individual subjects was significantly different for energy, protein, and fat measures following cleaning modifications. Manual modifications made during data cleaning can change mean nutrient intake measures for an entire cohort and individuals. Investigators should consider nutrients of interest and individual versus group intake when deciding whether to make cleaning modifications.