Population health · quantitative sociology · United States
A reproducible quantitative health-sociology project using the 2024 National Health Interview Survey to examine how education and family poverty are associated with multimorbidity across stages of adulthood.
Research problem
Multimorbidity is often presented as an aging problem, but chronic-condition accumulation also varies across socioeconomic position. The analytical challenge was to quantify education and poverty gradients while respecting the NHIS complex sample design and the CDC's multiply imputed income data.
The project asks how socioeconomic disadvantage is associated with multimorbidity among U.S. adults and whether the education gradient varies across adulthood.
Primary outcome: two or more of 12 chronic physical or mental health conditions. A physical-only definition is included as a sensitivity analysis.
Executive view
Analytical architecture
Findings
The joint interaction test was significant: F = 5.06, p < 0.001.
Among adults aged 50–64, estimated prevalence was 51.9% for those with less than high school education and 34.9% for those with a bachelor's degree or higher.
Hypertension, arthritis, cancer, diabetes and heart disease rose strongly with age, while diagnosed anxiety and depression were more prevalent in younger adults.
Adults with less than high school education had adjusted odds 1.73 times those with a bachelor's degree or higher. Adults below 100% of the federal poverty level had adjusted odds 2.08 times those at 400%+ FPL after pooling all 10 income imputations.
Methods & interpretation
The workflow uses the final annual Sample Adult weight, pseudo-PSU and pseudo-stratum; survey-weighted descriptive estimates; logistic and quasi-Poisson models; an education-by-age interaction; a physical-only sensitivity analysis; and Rubin's rules to pool the 10 CDC income imputations.
The study is cross-sectional. Results are associations, not causal effects or within-person life-course trajectories. Odds ratios are not interpreted as risk ratios, and race/ethnicity coefficients are treated as descriptive adjusted associations rather than biological effects.
Inspect the work
The repository contains the analysis plan, variable dictionary, reproducibility notes, QA logs, machine-readable result tables, figures and the complete R pipeline.