Date of Award

Spring 2021

Document Type

Open Access Dissertation


Epidemiology and Biostatistics

First Advisor

Robert McKeown


Objective: Disparities in the distribution of oral and pharyngeal cancer exist worldwide. This is also true in the United States and in South Carolina. Differences exist in oral and pharyngeal cancer stage and survival by race and sex, with higher incidence rates for Whites but higher proportions of late-stage cancer and lower survival for Blacks. This study was conducted to evaluate whether any differences exist in the spatial distribution of oral and pharyngeal cancer stage at diagnosis and survival, and to evaluate the association of income inequality, material deprivation and other measures of social determinants of health (contextual risk factors) on stage at diagnosis and survival for oral and pharyngeal cancer patients in South Carolina. Together these analyses may help to explain existing disparities in the distribution of this disease. Methods: Incident oral and pharyngeal cancer cases aged ≥20 diagnosed between 2009 and 2017 in South Carolina were included in the study. Patient demographic and cancer diagnostic data were linked to contextual survey data. Contextual data were obtained from United States American Community Survey census data at the county and block group level, and Behavioral Risk Factor Surveillance System (BRFSS) data were obtained at the county level. Geocoded cancer cases were evaluated for spatial clustering using the Getis-Ord GI* statistic. Multilevel statistical analyses were used to evaluate the association of income inequality, measured as the Gini Coefficient at the county level, and material deprivation, measured as the Townsend Deprivation Score at the block group level, with the presence of late-stage cancer (multilevel logistic analysis) and survival (multilevel survival analysis). Other demographic, cancer diagnostic, and contextual effects were also evaluated. Results: Results showed that spatial variation of late-stage cancer and poor survival in the form of clustering was minimal, but informative. For these data, income inequality was not significant and odds ratios (late-stage) and hazard ratios (survival analyses) for the Gini Coefficient quintiles were close to the null value of one. A higher prevalence of low education was associated with late-stage cancer and survival; however, results were not significant for survival. County level prevalence of smoking was not associated with late-stage cancer but was associated for survival, indicating that patients who lived in counties with the highest prevalence had poorer survival outcomes than those who lived in counties with the least prevalence. Conclusions: Although limitations from the use of contextual aggregate data measuring social determinants exist, the potential also exists for the identification of groups or areas for which social determinants may be important and where interventions or health programs may be most successful or where changes to infrastructure are needed.

Available for download on Monday, May 15, 2023

Included in

Epidemiology Commons