Policy Research for Front of Package Nutrition Labeling: Developing and Testing a Summary System Algorithm. 4.3.1 Results of Regression Models Predicting HEI Scores

05/01/2011

The performance of the algorithms in predicting overall dietary quality was assessed by calculating algorithm scores for foods consumed by the 15,576 individuals in NHANES 2005-2008. The results for the regression of the composite algorithm scores for the NHANES participant diets on the HEI diet quality scores for the participant diets are presented in Table 4-1. The models included covariates for age, gender, and ethnicity. The details of the methodology are described at the end of Section 4.1.

Table 4-1. Weighted Mean Scores of Foods Consumed by 16,587 Individuals in NHANES 2005-2008 and Linear Regression Models on the Healthy Eating Index
Algorithm Modification Weighted
Mean Scores
(SE)
Beta
Coefficient
(SE)
p-value R2
Notes: A model with no nutrients (only with covariates age, gender, and ethnicity) per 100 kcal has an R2 of 4.17%. A per 100 kcal model with only negative nutrients (saturated fat, added sugars, and sodium) has an R2 of 35.19%, and a model with only the positive nutrients (protein, fiber, vitamin E, vitamin D, calcium, iron, potassium, and unsaturated fat) has an R2 of 32.66%.
- Means no data.
Per 100 kcal - - - - -
NDS1KCAL - −2.79 (0.06) 4.12 (0.08) <0.0001 46.19%
NDS2KCAL − Unsaturated fat −2.92 (0.06) 3.96 (0.07) <0.0001 46.49%
NDS3KCAL − Added sugar −2.55 (0.03) 4.63 (0.11) <0.0001 36.23%
NDS4AKCAL + Vitamin A −3.01 (0.06) 4.14 (0.08) <0.0001 46.16%
NDS4B12KCAL + Vitamin B12 −2.78 (0.05) 3.99 (0.07) <0.0001 44.42%
NDS4CKCAL + Vitamin C −2.58 (0.06) 3.95 (0.07) <0.0001 49.58%
NDS4FKCAL + Folate −2.97 (0.05) 4.13 (0.08) <0.0001 45.64%
NDS4MKCAL + Magnesium −2.84 (0.06) 4.10 (0.08) <0.0001 46.41%
NDS5KCAL + Whole grains - - - -
- % of total/50% −2.83 (0.06) 4.00 (0.07) <0.0001 48.59%
- # servings −3.11 (0.06) 4.20 (0.08) <0.0001 47.36%
NDS6KCAL + Vitamin C and whole grains (% of total/50%) −2.64 (0.07) 3.89 (0.06) <0.0001 51.57%
Per RACC - - - - -
NDS1RACC - −2.92 (0.06) 3.34 (0.08) <0.0001 38.99%
NDS2RACC − Unsaturated fat −3.06 (0.06) 3.19 (0.07) <0.0001 38.99%
NDS3RACC − Added sugar −2.58 (0.03) 3.91 (0.11) <0.0001 32.48%
NDS4ARACC + Vitamin A −3.13 (0.06) 3.28 (0.08) <0.0001 38.40%
NDS4B12RACC + Vitamin B12 −2.91 (0.06) 3.26 (0.08) <0.0001 37.75%
NDS4CRACC + Vitamin C −2.74 (0.07) 3.31 (0.08) <0.0001 42.21%
NDS4FRACC + Folate −3.09 (0.06) 3.31 (0.08) <0.0001 38.42%
NDS4MRACC + Magnesium −2.99 (0.06) 3.31 (0.08) <0.0001 39.11%
NDS5RACC + Whole grains - - - -
- % of total/50% −2.98 (0.07) 3.26 (0.07) <0.0001 40.78%
- # servings −3.24 (0.06) 3.30 (0.08) <0.0001 39.11%
NDS6RACC + Vitamin C and whole grains (% of total/50%) −2.82 (0.07) 3.25 (0.07) <0.0001 43.58%

In all of the linear regression models, the weighted mean scores were significantly associated with HEI scores (p < 0.0001). The weighted mean scores of foods consumed by individuals were slightly higher on a per 100 kcal than on a per RACC basis. For the baseline algorithm (NDS1KCAL), the model with scores based on a per 100 kcal basis explained 46.2% of the variation in HEI scores compared with 39.0% explained by the per RACC scores (NDS1RACC). For each modification to the baseline algorithm, the variation in HEI scores was better explained by the algorithm calculated on a per 100 kcal basis than a per RACC basis. The model with the highest R2 was the modification with vitamin C and whole grains on a per 100 kcal basis (NDS6), explaining 51.6% of the variance in HEI scores. The algorithm with vitamin C had the second highest R2 (49.6%).

A better explanation of HEI scores when the food scores were calculated on a per 100 kcal basis rather than a per RACC basis could possibly be that the range of scores on a per 100 kcal basis is more extreme. The fact that the HEI is based on nutrient standards on a per 1,000 kcal basis could also contribute to a better prediction by an algorithm scored on a per 100 kcal basis.