Presentation (CTI)

OBJECTIVES: Colorectal cancer is among the most commonly diagnosed cancers in the United States and the leading cause of cancer death for men and women combined, yet it is also one of the most preventable and early-stage detection carries a 91% five-year survival rate compared to 14% after distant spread. The communities where screening is lowest are often the same communities where late-stage diagnosis is most common, a convergence that demands precise identification of which neighborhood barriers most strongly suppress screening uptake. Composite social deprivation indices such as the CDC Social Vulnerability Index (SVI) are widely used to study these disparities but obscure which specific determinants drive screening variation, and standard regression with correlated predictors is difficult to interpret. This study applies dominance analysis, a variable-importance method designed for correlated predictors, to rank neighborhood-level determinants of tract-level screening prevalence and test whether those rankings vary across states.
METHODS: We constructed a cross-sectional dataset of U.S. census tracts linking CDC PLACES screening prevalence, SVI measures, individual American Community Survey predictors (e.g., poverty, unemployment, insurance coverage, education, limited English proficiency), USDA Rural-Urban Commuting Area codes, and Opportunity Atlas intergenerational mobility estimates harmonized to 2020 boundaries. Sequential models assessed explanatory gains from disaggregation. LASSO cross-validation will pre-select predictors before dominance analysis, conducted nationally and within six focal states: Georgia, Texas, Iowa, New York, California, and Massachusetts.
RESULTS: Adjusted R² increased from 0.398 (SVI composite) to 0.522 (SVI themes) to 0.650 (individual predictors), demonstrating substantial explanatory heterogeneity masked by composite indices. Adding Opportunity Atlas mobility raised adjusted R² to 0.653 on 81,498 tracts, statistically significant but modest in practical magnitude.
CONCLUSIONS: Individual tract-level SDOH measures explain substantially more screening variation than composite indices. Dominance analysis will provide definitive variable-importance rankings (eg. demographic & socioeconomic factors), clarify whether intervention priorities differ across states, and designing policies to address SDoH.