Abstract
OBJECTIVES: To compare five Healthy Eating Index-2015 (HEI-2015) scoring methods using 24-hour recalls (24HRs), 4-day food records (4DFRs), and food frequency questionnaires (FFQs). METHODS: Over 12 months, Interactive Diet and Activity Tracking in AARP (IDATA) study participants (N = 1021) aged 50–74 years completed up to six Automated Self-Administered 24-Hour Dietary Assessment Tool (ASA24) recalls, 2 4DFRs, and 2 FFQs. Mean HEI-2015 total and component scores were estimated using five methods – two estimating usual intake: the multivariate Markov Chain Monte Carlo (MCMC) and bivariate; and three not estimating usual intake: per day, per person, and population ratio. Sums of squared differences (SSD) were calculated to compare differences among component scores. RESULTS: MCMC and bivariate methods estimated similar total mean HEI-2015 scores for men with 24HRs (60 points) and 4DFRs (61 points). The population ratio scores were comparable (63 points), but higher, and the per day was most different for both 24HRs and 4DFRs (57 points). This pattern was similar for women. With 24HR and 4DFRs, the population ratio method had higher component scores compared with MCMC and bivariate for Total Fruits, Whole Fruits, and Seafood and Plant Proteins. For example, among men, in 24HRs (comparing MCMC and population ratio), the SSD for Whole Fruit were 1.44 and those of Seafood and Plant Proteins were 0.49, compared to SSDs for other components which only ranged from 0.01 to 0.16. With FFQs, estimation of mean HEI scores is not recommended due to biases. However, when applying all methods, the total and component scores for FFQs were higher for Total Fruits, Whole Fruits, Greens and Beans, Dairy, Fatty Acids, Refined Grains, Sodium, and Saturated Fats. CONCLUSIONS: Overall, the two usual intake methods (MCMC and bivariate) yield comparable total and component scores. The population ratio method adjusts for day-to-day variation by averaging data across populations, thus arrives at scores closer to the MCMC and bivariate, hence is the preferred method of estimating a population's mean usual HEI scores on the basis of a single day of data. When distributions are needed, the MCMC and bivariate methods are recommended to adjust for measurement error, consider episodic consumption and skewness, and account for correlation between each and or all constituents and energy. FUNDING SOURCES: N/A.