Date of Award


Degree Type


Degree Name

Doctor of Philosophy in Health Sciences


Health Studies

First Advisor

Maya Vadiveloo


Over the past few decades, evidence-based dietary guidance has transitioned in focus from single components (i.e., nutrients or food groups) to a more holistic total diet approach. While the total diet approach is more practical for public health messaging, capturing the complexity and multidimensionality of overall diet is methodologically challenging. As a result, diet quality indexes have been developed to facilitate comparisons across dietary patterns by yielding total diet quality scores (i.e., higher score=healthier diet). Various high-quality diets are consistently associated with a similar magnitude of overall and cardiovascular disease (CVD) mortality risk reduction across different diet quality indexes, despite distinct index-specific weighting schemes (i.e., component contributions to total score).

Current diet quality weighting schemes are subjectively assigned have yet to be optimized to represent component-specific importance on health outcomes. However, since intake of some individual diet quality components (e.g., fruits, vegetables, whole grains, and plant proteins) confer mortality risk reductions comparable to total diet quality scores, current diet quality indexes may benefit from alternative weighting approaches. Moreover, data-driven machine learning (ML) techniques may better capture the relative importance of multidimensional dietary components but have yet to be applied to current diet quality indexes.

The present study assessed two modified diet quality component weighting schemes within the Healthy Eating Index (HEI)-2015, a diet quality index used to assess dietary pattern adherence with the U.S. Dietary Guidelines for Americans 2015-2000. Standard HEI-2015 scores were calculated, then components were reweighted using two modified weighting schemes, creating two new HEI scores: a theory-based Key Facets HEI where weights were distributed equally across the most common and individually impactful components (i.e., fruits, vegetables, plant-based proteins, and whole grains – referred to as “Key Facets”) and a data-driven ML-weighted HEI utilizing ML models to assign weights based on relative component importance. Ultimately, analyses assessed diet quality scores measured with standard and modified component weights and their associations with all-case and CVD mortality within two independent samples.

The hypothesis that compared to standard scores, modified-weight scores would be associated with larger magnitudes of mortality risk was first examined in a nationally representative sample of U.S. adults in the National Health and Nutrition Examination and Survey III (1988-1994). Within this sample, modified-weight HEI scores were significantly associated with 23% to 39% reductions in all-cause mortality risk while the standard HEI-2015 was not associated with all-cause mortality risk and none of the HEI scores were associated with CVD mortality risk. After establishing associations in a nationally representative sample, analyses were directly replicated within the Multiethnic Cohort (MEC, 1993-1996), a large, ethnically-diverse independent cohort. Within the MEC, standard and modified-weight HEI scores were comparably associated with 10% to 24% decreased all-cause and CVD mortality risk.

Current dietary guidance equally emphasizes all recommendations equally, however, our analyses suggest that there may be opportunities to simplify guidance without compromising health benefits. Compared to the standard HEI-2015, two modified weight HEI scores featured heavily weighted Key Facet components (i.e., fruits, vegetables, plant-based proteins, and whole grains) and reduced weight moderation components (i.e., sodium, added sugar, saturated fat) showed promise with greater magnitudes of mortality risk associations in a nationally representative sample. While the greater magnitudes of mortality risk associations were not similarly observed when analyses were replicated in an independent cohort, this may be related to error-prone intake measurement and/or healthier overall dietary patterns. Taken together, improved performance in a national sample and comparable performance in an independent cohort indicates that modified weight HEI scores should be further examined, ideally using more recently collected data reflecting present day dietary patterns.



To view the content in your browser, please download Adobe Reader or, alternately,
you may Download the file to your hard drive.

NOTE: The latest versions of Adobe Reader do not support viewing PDF files within Firefox on Mac OS and if you are using a modern (Intel) Mac, there is no official plugin for viewing PDF files within the browser window.