Socio-demographic differences in the dietary inflammatory index from National Health and Nutrition Examination Survey 2005-2018: a comparison of multiple imputation versus complete case analysis

Abstract

Objective: Studies using the dietary inflammatory index often perform complete case analyses (CCA) to handle missing data, which may reduce the sample size and increase the risk of bias. Furthermore, population-level socio-economic differences in the energy-adjusted dietary inflammatory index (E-DII) have not been recently studied. Therefore, we aimed to describe socio-demographic differences in E-DII scores among American adults and compare the results using two statistical approaches for handling missing data, i.e. CCA and multiple imputation (MI).

Design: Cross-sectional analysis. E-DII scores were computed using a 24-hour dietary recall. Linear regression was used to compare the E-DII scores by age, sex, race/ethnicity, education and income using both CCA and MI.

Setting: USA.

Participants: This study included 34 547 non-Hispanic White, non-Hispanic Black and Hispanic adults aged ≥ 20 years from the 2005-2018 National Health and Nutrition Examination Survey.

Results: The MI and CCA subpopulations comprised 34 547 and 23 955 participants, respectively. Overall, 57 % of the American adults reported 24-hour dietary intakes associated with inflammation. Both methods showed similar patterns wherein 24-hour dietary intakes associated with high inflammation were commonly reported among males, younger adults, non-Hispanic Black adults and those with lower education or income. Differences in point estimates between CCA and MI were mostly modest at ≤ 20 %.

Conclusions: The two approaches for handling missing data produced comparable point estimates and 95 % CI. Differences in the E-DII scores by age, sex, race/ethnicity, education and income suggest that socio-economic disparities in health may be partially explained by the inflammatory potential of diet.

Keywords: Dietary intake; Disparities; Multiple imputation; NHANES; Nutritional epidemiology.

Fig. 1

Selection of participants for complete case analyses and multiple imputation subpopulations from National Health and Nutrition Examination Survey years 2005–2018. The complete case analyses subpopulation included only participants with no missing data on exposure, outcome or covariates of interest. The multiple imputation analyses subpopulation included all eligible participants, regardless of missing data on exposure, outcome or covariates of interest. NHANES, National Health and Nutrition Examination Survey; NH, non-Hispanic.

Fig. 2

Box plots of energy-adjusted dietary inflammatory index scores by two-year National Health and Nutrition Examination Survey cycles. The range of observed E-DII scores was -5·96 (anti-inflammatory dietary potential) to 4·90 (pro-inflammatory dietary potential). (Bars represent the interquartile range; dots represent outliers.) E-DII, energy-adjusted dietary inflammatory index.