Correction to: Nature Food https://doi.org/10.1038/s43016-021-00381-y, published online 14 October 2021.

In the original analysis, a coding error occurred in which the portions of fruits and vegetables present in mixed dishes and recipes were not always captured in the attribute scoring for fruits and vegetables in the food-based ingredients domain. This has now been corrected. In addition, the original analysis utilized a machine-learning algorithm to categorize foods according to the NOVA classification. This led to appropriate classification for most packaged items. However, because NOVA was not originally designed to be used for mixed dishes made at home, this led to incorrect classification for several of these items, which can be made from a range of ingredients of varying processing categories. In this update, all food items have now been manually categorized according to NOVA, in collaboration with Dr. Carlos Monteiro (the founder of the NOVA classification system). As an example, for the manual classification, mixed dishes made at home could include various ingredients of differing NOVA classes (e.g., a jarred spaghetti sauce with additives [NOVA = 4], pasta [NOVA = 1], and processed meat [NOVA = 3]). For such items, the NOVA score of each ingredient, disaggregated within FNDDS, was classified, and the final NOVA score of the mixed dish was the energy-weighted sum of the NOVA scores of its ingredients, rounded to the nearest NOVA class (integer). The impact of these updates on the final Food Compass Score are shown in Fig. 1 below. The revised NOVA scores and Food Compass Scores for all 8,032 items are presented in Supplementary Table 7, and for comparison and transparency, the original (uncorrected) versions of these scores are presented in a new Supplementary Table 8, both available alongside the original article. These updates did not appreciably alter the overall results or conclusions from the analysis. The Spearman correlation between the new and old Food Compass Score was 0.98. All analyses, text, tables, figures, and supplementary materials have been revised with the updated Food Compass Score.

Fig. 1: Histogram of changes (new–old) in the Food Compass Score (FCS) for 8,032 unique foods and beverages consumed in the United States (NHANES/FNDDS 2015–2016).
figure 1

Scoring updates included (1) correction of attribute scoring for fruits and vegetables in the food-based ingredients domain to capture portions of fruits and vegetables in mixed dishes and recipes; and (2) revised methods for attribute scoring of NOVA categorization in the processing domain, based on manual classification rather than machine-learning classification. Overall FCS changes were modest, with most items (58%) having no more than a ± 5 point score change (out of 100), 41% having an increase of between 5–15 points, a few (1%) having a score increase of larger than 15 points, and a few (1%) having a decrease of larger than 5 points. The Spearman correlation between the new and old Food Compass Score was 0.98.