Abstract 1202: Spatial analysis of late-stage oropharyngeal cancer diagnosis in Florida: Hierarchical models and geographically-weighted regression

Abstract

Abstract Literature on spatial variation of cancer outcomes in the US remains scarce. Studies have not assessed how late-stage OPC diagnosis varies across geography nor the effect of spatially different socioeconomic factors. Identifying risk factors that have higher impact on certain localities can help inform clinicians and policy-makers. Our study aimed to evaluate spatial differences in stage of diagnosis of oropharyngeal cancer (OPC) in Florida using hierarchical and geographically-weighted regression (GWR) models. We used the Florida Cancer Data System to derive a cohort of 57,082 patients diagnosed with OPC during 1990-2010. The individual-level covariates that informed our models consisted of age at diagnosis, gender, race, married status, health insurance payer, history of cigarette smoking, and year of diagnosis. Patients' addresses were geocoded to obtain a point feature used to map the spatial distribution of individuals using ArcGIS. A hot-spot analysis (Getis-or-Gi) revealed areas of high clustering of patients with late-stage diagnosis. To evaluate spatially-relevant factors for such variation, we fitted a mixed logistic regression model using PROC GLIMMIX in SAS with the individual-level variables in addition to county and census tract-level covariates. We used US Census data to extract records on the percent of residents unemployed, living in poverty, and foreign-born. Variables for the county level extracted from the Area Health Resource File and the Behavioral Risk Factor Data included the percent of residents engaging in heavy drinking, who couldn't visit a dentist due to cost, and who were smokers, a flag for Health Professional Shortage Areas for primary care doctors and dentists, presence of a hospital with oncology services, and a flag for rural designation. The model showed significant higher odds for late-stage OPC diagnosis for non-White patients (OR 1.63, 95%CI 1.48-1.79 for Black and 1.29, 95%CI 1.03-1.62 for other-race), current smoker status (OR 1.45, 95%CI 1.36-1.54), former smoker status (OR 1.36, 95%CI 1.27-1.45), patients without health insurance (OR 1.60, 95%CI 1.41-1.82), patients covered by Medicaid (OR 1.73, 95%CI 1.5-1.97), patients diagnosed with cancer in pharyngeal sites (OR 4.74, 95%CI 4.53-4.95), and higher percentage of residents in census tract living in poverty (OR 1.01, 95%CI 1.00-1.011). The GWR conducted at the county-level revealed areas where coefficients associated with various risk-factors contributed to a higher proportion of late-stage OPC diagnosis. For easier communication of findings, we created density maps that will be presented. Our study empirically showed variation of late-stage OPC diagnosis across Florida and differential effects of spatial socioeconomic factors. Future public health initiatives can use a targeted approach to focus resources in these specific areas more likely to be impacted. Citation Format: Juan M. Hincapie-Castillo, Yi Guo, Natalie L. Silver, Hong Xiao. Spatial analysis of late-stage oropharyngeal cancer diagnosis in Florida: Hierarchical models and geographically-weighted regression [abstract]. In: Proceedings of the American Association for Cancer Research Annual Meeting 2018; 2018 Apr 14-18; Chicago, IL. Philadelphia (PA): AACR; Cancer Res 2018;78(13 Suppl):Abstract nr 1202.