Evaluating the Healthy Eating Index

The Healthy Eating Index (HEI) has been evaluated by assessing its psychometric properties, including content validity, four types of construct validity, and one type of reliability. Each version since 2005 has undergone an analysis to assess its performance under these constructs.

To do this, we scored dietary data from the most recent National Health and Nutrition Examination Survey (NHANES) for each version of the HEI. Data from 2001-2002 NHANES was used to evaluate HEI-2005; data from NHANES 2003-2004 was used to evaluate HEI-2010; and data from NHANES 2011-2012 was used to evaluate HEI-2015.

We also scored several sets of exemplary sample menus, including those from the United States Department of Agriculture's (USDA) website, the National Heart, Lung, and Blood Institute's (NHLBI) DASH program, Harvard Medical School's Healthy Eating PyramidExternal Web Site Policy, and the American Heart Association's No-Fad DietExternal Web Site Policy.

Results of the evaluations showed that the HEI has both validity and reliability:

  • Content validity: For HEI-2005, HEI-2010 and HEI-2015, all of the key food choice recommendations of the Dietary Guidelines that relate to diet quality are reflected in HEI components. By design, the HEI does not cover other Dietary Guidelines recommendations, such as those for physical activity, body weight, and food safety.
  • Construct validity: This type of validity evaluates whether an index measures what it is supposed to measure, and the HEI performed well on four types of construct validity.

    • First, four sets of menus representing high-quality diets scored fairly high on the HEI–2005, HEI-2010 and HEI-2015. Predictably, as the HEI was refined over time, some menus scored lower on later versions of the HEI, reflecting the refinement of advice on subgroups such as dark green vegetables and components such as added sugar.
    • Second, one-day scores between smokers and non-smokers were significantly different in all versions, indicating that the HEI can distinguish between groups with known differences in the quality of their diets.
    • Third, an examination of Pearsons correlations of the HEI total and component scores with energy intake showed that the HEI can assess diet quality independently of diet quantity.
    • Finally, a principal components analysis (PCA) demonstrated that multiple factors underlie each version of the HEI and that both the individual components and the total score provide insights into diet quality. A new addition to the evaluation of construct validity in 2010 came with a new statistical method known as the Markov Chain Monte Carlo (MCMC) method, which allowed us to estimate distributions of HEI scores of estimated usual intake in the population (rather than only the mean scores) to assess whether the distribution was wide enough to detect meaningful differences.
  • Reliability: For all versions, we used Cronbach's coefficient alpha to assess internal consistency, which is the degree to which multiple components within an index measure the same underlying construct. Results suggest that individual components provide additional insights into the quality of the diet beyond those of the total score.

    For additional details: