Food Compass Score-10: validation of a method for evaluating the healthfulness of foods and beverages using ingredient list information

Abstract

Background

The Food Compass, a novel food profiling system, provides a holistic, validated assessment of the healthfulness of foods, beverages, and meals using 54 attributes across 9 domains. However, information on several of these attributes is not commonly available.

Objectives

We aimed to develop and validate an approach, Food Compass Score-10 (FCS-10), to estimate FCSs using information commonly available on package labels.

Methods

Missing attributes were calculated using weighted scores of each product’s ingredients, derived from a dataset of ∼10,000 foods and beverages. The final FCS-10 was scaled from 1 (least healthful) to 10 (most healthful). As part of this validation study, diagnostic accuracy analysis was conducted to evaluate the performance of the FCS-10 compared with the original score. Sensitivity, specificity, positive predictive value, and negative predictive value were calculated by comparing the FCS-10 recommendation categorizations with the FCS recommendation categorizations (≥7 for foods to encourage, 4–6 for foods to consume in moderation, ≤3 for foods to limit).

Results

FCS-10 produced scores within 1 unit of the original score (when rescaled 1–10 for comparison) for 89% of products (n = 481/538); none deviated >2 units. The correlation between FCS-10 and the original score was high (r = 0.93). FCS-10 also performed well in identifying products to encourage, moderate, or limit, with overall sensitivity and specificity of 87% and 93%, respectively.

Conclusions

FCS-10 offers a practical approach for estimating the healthfulness of diverse packaged foods and beverages using readily available label data while maintaining the strengths of the original system.