Work overview

Section 02 of 19

Methodology

Sociodemographic and biochemical correlates of self-reported taste dysfunction: an NHANES analysis

Dhruva Arcot and Subhankar Chakraborty · 2026

Contents

Section 02 of 19

  1. 01Introduction and background
  2. 02Methodology
  3. 03Data source and target population
  4. 04Outcomes and covariate domain categorization
  5. 05Statistical analysis and design weighting
  6. 06Survey-weighted descriptive profiling
  7. 07Purposeful variable selection framework
  8. 08Multivariate regression
  9. 09Hypothesis
  10. 10Results
  11. 11Laboratory biomarkers and taste problems
  12. 12Metabolic, liver, and kidney biomarkers
  13. 13Vitamins
  14. 14Serum fatty acids
  15. 15Univariate logistic regression
  16. 16Multivariable logistic regression
  17. 17Discussion
  18. 18Conclusions
  19. 19Supplementary Material
Text size
Work overview

Section 2 of 19

Methodology

Dhruva Arcot and Subhankar Chakraborty · about 2 minutes

Original source ↗

The purpose of the study was to investigate whether biochemical or hematological markers are associated with taste problems. Figure 1 above shows the steps used in the data analysis for this study.

Figure 1:: Vertical flowchart showing a stepwise analytic decision process. Rectangular workflow boxes are connected by downward arrows representing sample selection, survey-weighted descriptive analyses, regression-based variable screening, multivariable model development, and collinearity assessment. The process culminates in a decision node that separates variables demonstrating independent associations from variables that do not meet significance criteria in the final adjusted model.

Figure 1:: NHANES analytic workflow for identifying independent correlates of self-reported taste dysfunction. This flowchart outlines the sequential analytic procedures used to derive the final survey-weighted logistic regression model. Adult participants (≥20 years) from the 2011–2012 and 2013–2014 NHANES cycles were included if they provided a valid “Yes/No” response to the taste problem item (CSQ080), yielding a subpopulation of 3,595 cases. Descriptive and bivariate profiling was conducted using complex-sample general linear models for continuous variables and adjusted chi-square tests for categorical variables. Univariable logistic regression screening (p<0.25) identified candidate covariates for multivariable modeling. The fully adjusted model incorporated a forced baseline block of socio-demographic variables (age, gender, race/ethnicity, education, income, marital status, and country of birth), followed by sequential entry of biochemical factors. Collinearity diagnostics were performed using complex-sample linear regression, applying a strict VIF>5.0 exclusion threshold. Final model decisions were based on whether each parameter demonstrated independent significance (p<0.05), identifying variables with true associations.

Figure description: A vertically oriented flowchart composed of sequential rounded rectangular boxes connected by arrows. The workflow progresses from analytic sample identification through descriptive and bivariate analyses, followed by univariable screening of candidate predictors and development of a fully adjusted multivariable regression model. A dedicated collinearity evaluation step precedes model interpretation. The final section contains a diamond-shaped decision node asking whether a variable remains independently associated with taste dysfunction in the adjusted model. Two branches emerge from this decision point: the left branch classifies variables as independently associated with taste dysfunction when statistical significance is achieved, and the right branch classifies variables as lacking evidence of an independent association. The figure emphasizes sequential variable selection, collinearity filtering, and final assessment of independent effects within the adjusted model.