An Improved Method for Scoring the Moderation Components of a Diet Quality Index

Background Dietary intake scoring functions, such as the Healthy Eating Index (HEI), are useful for quantifying a measure of overall diet quality and have been used to assess relationships between diet quality and other factors of interest, such as demographics, economics, and health‐related outcomes. Such scores are typically composed of multiple components with each component related to a specific dietary recommendation. Where the recommendation calls for intake to be at or above a recommended amount (such as fruit), the component is termed an adequacy component and is scored from 0 (no intake) to a maximum value (recommended intake). Additional intake results in no increase in score. In the case where the recommendation is an upper limit on intake (such as for sodium), the component is termed a moderation component, and maximum points are scored for any intake from zero up to the upper limit. Objective The purpose of this work was to develop an improved approach for scoring moderation components when the intake exceeds the upper limit. Methods The following goals were set. The scores should: monotonically and continuously decline as intake increases above the upper limit; improve as intake decreases until the upper limit is reached; not exhibit an immediate, sharp reduction for intakes just slightly above the upper limit. The limits are based on scientific evidence; but they are not hard and fast, precise numbers. A rapid decrease in scores in the area immediately above the limit would attribute undue precision to it. Currently, in the HEI‐2010, a linear decline model assigns the maximum number of points to intakes at or below the upper limit and assigns zero points to intakes at or above a high standard amount. However, this does not meet goal (a) since intake above the high standard does not result in any change in score. In considering alternatives to the linear decline model, we considered forming a distribution of all observations above the limit. If a component were to be scored from 0–10, one could divide the distribution into centiles and assign declining scores from 9.9 to 0.0 for each centile, for example. Obviously, this technique could be extended to any number of maximum points and any number of decimal points, and it would provide maximum variation in scores. However, the scoring system would not be stable since it would depend on the data presented to the algorithm. Longitudinal studies and comparisons among studies would be hampered. Results We propose a new, stable, composite function for scoring moderation components that achieves the above three goals. The composite function, illustrated in the Figure, has three regions: As with the HEI‐2010, intakes at or below the recommended upper limit get the maximum number of points. Intakes above the upper limit are scored on the same linear decline as was chosen for HEI‐2010 until the score declines to 10% of the standard for the maximum score. This provides compatibility for scores in this range. Above this point, a declining exponential function is used (MAXSCORE*EXP(C*(X‐SSZ))) where C is a constant, MAXSCORE is the maximum score, and SSZ is the standard for the score of 0). The declining exponential is fitted with constant C to meet the linear decline at 10% of the maximum score.Conclusion The new composite function satisfies all three goals. It maintains significant compatibility with the HEI‐2010 scoring function while providing scoring sensitivity at the upper end of the intake distribution. Support or Funding Information This work was funded in part by a University of Utah Vice President for Research Innovation Seed Grant, a National Institutes of Health Training Grant (T15‐LM007124), a research grant from the Western Regional Nutrition Education and Obesity Prevention Center of Excellence (USDA‐NIFA‐OP‐004574), and a research grant from the USDA National Institute of Food and Agriculture (2015‐09151).