Date on Master's Thesis/Doctoral Dissertation

12-2025

Document Type

Doctoral Dissertation

Degree Name

Ph. D.

Department

Epidemiology and Population Health

Degree Program

Public Health Sciences with a specialization in Epidemiology, PhD

Committee Chair

Boone, Stephanie

Committee Member

Baumgartner, Kathy

Committee Member

Dupre, Natalie

Committee Member

Peiper, Nicholas

Committee Member

Gaskins, Jeremy

Author's Keywords

SDOH; health disparities; geography; rural; urban; cancer

Abstract

Social determinants of health (SDOH) factors have consistently been shown to impact all-cause and cancer-specific mortality. Area-level SDOH factors that coexist to form “high-risk” clusters are more prevalent in rural than urban areas. However, heterogeneity within rural and urban areas may contribute to variations in SDOH factors that could influence higher mortality rates and is not considered in other studies. The primary objective of this study was to utilize ecological and individual-level data to determine if the composition of SDOH clusters within counties in the rural-urban continuum contributes to differences in all-cause (AC) and cancer-specific mortality rates of lung and bronchus, colorectal, breast, cervical, and prostate cancers in the United States (US). The central hypothesis was that the rural-urban continuum would have varying and unique combinations of SDOH clusters in each geographic level, and counties with increasing disadvantage (high-risk clusters) would be associated with a higher risk of cancer mortality compared to less disadvantaged counties (low-risk clusters). Data on 5-year age-adjusted county-level cancer-specific and AC mortality rates for those with breast, lung and bronchus, colorectal, and prostate cancer between 2010-2014 and 2015-2019 were retrieved from the Surveillance, Epidemiology, and End Results Program (SEER) database. County-level prevalence data on SDOH factors were retrieved from the American Community Survey (ACS), the County Health Rankings & Roadmaps (CHR), and the Agency for Healthcare Research & Quality (AHRQ). These data were linked with SEER data using county FIPS codes and counties were categorized using the 2013 Rural-Urban Continuum Code into six levels (Metro 1, Metro 2, Metro 3, Non-metro 1, Non-metro 2, and Non-metro 3). Multigroup Latent Class Analysis was conducted using 13 SDOH variables for each geographic level. Adjusted Bayesian Information Criterion (aBIC), Bayesian Information Criterion (BIC), Akaike Information Criterion (AIC), and Entropy were used to assess model fit. Each county was assigned a best class based on the highest posterior probability. Negative Binomial Regression analysis was conducted to evaluate the association between classes within geographic levels and cancer-specific and AC mortality rates at an ecological level, and Rate Ratios (RR) and 95% Confidence Intervals (CI) were calculated. To measure the effect of individual-level covariates on the association between SDOH classes assigned in the previous analysis and cancer-specific and AC mortality within geographic levels, Cox proportional hazards modeling was used to estimate Hazard Ratios (HR) and 95% CIs. Moreover, an exploratory analysis was conducted to assess whether mortality risks differed when using a dichotomous rural-urban classification compared to the rural-urban continuum. Overall, counties in classes with higher disadvantages – such as lower education levels, high unemployment rates, higher proportions with income below the poverty level, increased household occupancy, low primary care physician rate, and higher levels of inadequate social support had significantly higher rates of cancer-specific and AC mortality. After adjusting for individual-level covariates, attenuation of estimates occurred, although significantly increased mortality risks remained within geographic levels. Compared to the rural-urban continuum, attenuation of risks for cancer-specific and AC mortality was observed for the dichotomous rural-urban classification. At a granular level, increased risks were observed primarily for the suburban or semi-urban (Metro 2 and Metro 3) levels compared to the most advantageous large fringe metro counties (Metro 1). Moreover, for rural counties, risks increased with increasing rurality, particularly for AC mortality. These increased risks were primarily driven by inadequate access to quality healthcare and lower education levels. Overall, findings suggest that SDOH heterogeneity within rural and urban counties differentially impacted cancer-specific and AC mortality rates. These findings underscore the importance of considering differences and complexities within geographic levels before comparing between geographic levels. This may be a first step when developing targeted interventions and policies to reduce geographic disparities.

Download

Included in

Epidemiology Commons

Share

COinS