Abstract P055: Reweighting Diet Quality Score Components Strengthens All-cause Mortality Risk Associations

Abstract

Introduction: The Dietary Patterns Methods Project concluded that various high-quality diets comparably reduce mortality, but it is unclear whether some individual components are more integral to risk reduction than reflected in current indexes. Thus, we examined mortality risk associated with standard vs. modified diet quality weighting schemes. Hypothesis: High scores on a data-driven (vs. standard) diet quality weighting scheme will be associated with stronger mortality risk reduction. Methods: National Health and Nutrition Examination and Survey III (1988-94) data from 10,789 non-pregnant adults ≥20 yrs. with valid mortality and dietary data were analyzed. Diet quality was scored from a single 24-hr recall with the Healthy Eating Index (HEI)-2015. The standard HEI 9 adequacy and 4 moderation components were reweighted creating: 1) the Key Facets HEI with fruits, vegetables, whole grains, and plant proteins equally weighted, emphasizing components present in virtually all diet quality indexes; and 2) the Machine Learning (ML)-weighted HEI where data-driven LASSO models determined relative component contribution to all-cause mortality associations. For all 3 HEIs, energy-residualized scores were ranked and assigned to quintiles (Q); covariate-adjusted sex-stratified Cox models with attained age as the timescale assessed all-cause and CVD mortality risk associations across HEI Qs. Results: In all 3 HEIs, high (Q5 vs. Q1) scores were associated with reduced all-cause mortality risk, ranging from 16% to 38%. The ML-weighted HEI had the strongest risk reduction (38% in women, 23% in men), 7 to 8% lower than standard HEI scores. Key Facets (vs. standard) scores yielded equivalent risk in men (16%) and 2% greater risk reduction in women. There was no association between any of the HEI scores and CVD mortality. Conclusions: All-cause mortality risk associations were 7-8% stronger in the ML-weighted vs. standard HEI. Findings suggest data-driven diet quality weighting schemes warrant further examination.