Unpacking the relationship between cancer survival and political environment in the United States (US).

Abstract

11030 Background: Several studies have reported on how health outcomes may differ by the political majority of a county. It is unclear the extent to which this discrepancy is due to a broader political landscape compared to underlying sociodemographic and structural determinants. Using county level presidential voting patterns and a national cancer registry, we present a rigorous analytic approach to assessing survival trends. Methods: We linked SEER-Medicare’s (100% sample) database on the 4 most common cancers (lung, breast, prostate and colorectal) from 2007-2019 to county-level data on US presidential elections over the same period. Using the total number of votes in each party, county political environment was classified as either Democratic or Republican for the four years following a presidential election. Cox proportional hazards estimated changes in survival over time according to political party. Models were sequentially fit adjusting for age, sex, diagnosis year (Model 1), race/ethnicity (M2), rural/urban continuum code (RUCC) (M3), cancer type (M4), Medicaid/Medicare Advantage enrollment and original reason for entitlement (M5), and primary care doctors and oncologists per 100,000 people (M6). Analyses were performed separately by cancer type and for the entire cohort. Results: Of the 1,899,334 individuals, 1,222,228 (64%) resided in Democratic counties and 677,106 (36%) in Republican counties. Democratic counties were majority urban (90% vs 55% for Republican). Adjusting for factors generally associated with worse survival (older age, male gender, earlier diagnosis year), Republican counties had worse survival (HR 1.14, 95% CI [1.14, 1.15], p <0.001) compared to Democratic counties. However, following sequential adjustment of aforementioned variables (M1-M6), this difference disappeared (HR 1.00, 95% CI [1.00, 1.01], p = 0.56, Table). In stratified analyses, similar results were noted for colorectal, lung, breast, and prostate cancer. Among all cancer types, Republican counties had 2% higher risk of diagnosis of later stage cancer, but no difference in overall survival. Finally, limiting our dataset to rural counties only, a 12% higher risk of death in Republican counties reduced to no difference following adjustment. Conclusions: Survival differences between counties by presidential voting status disappeared following adjustment of sociodemographic and structural determinants. In a charged political climate, our study cautions against oversimplified claims and underscores the importance of careful methodological approaches in identifying key confounders in aggregate-level studies when identifying meaningful trends. Hazard Risk of Death: Whole Cohort Hazard Ratio 95% CI M1 1.14 [1.14, 1.15]* M2 1.14 [1.13, 1.14]* M3 1.08 [1.08, 1.09]* M4 1.02 [1.02, 1.03]* M5 1.02 [1.01,1.02]* M6 1.00 [1.00, 1.01] M1-M6 note sequentially adjusted models (see Methods). *p < 0.001.