Identifying Spatio-Temporal Variation in Breast Cancer Incidence Among Different Age Cohorts Using Bayesian Hierarchical Modeling

• Activity Number: 358 - Contributed Poster Presentations: Section on Statistics in Epidemiology
• Type: Contributed
• Date/Time: Tuesday, July 30, 2019 : 10:30 AM to 12:20 PM
• Sponsor: Section on Statistics in Epidemiology
• Author(s): Amy Hahn* and Jacob J Oleson and Alexandra Thomas and Kristin Conway and Kathleen Stewart and Charles Lynch and Paul Romitti
• Companies: University of Iowa and University of Iowa and Wake Forest University School of Medicine and University of Iowa and University of Maryland and University of Iowa and University of Iowa
• Keywords: Point process; Epidemiology; SEER; INLA; Non-separable; Elevated risk

Abstract

We aim to learn more about breast cancer incidence in Iowa by examining annual cohorts based on locations at birth and diagnosis, year of diagnosis, and age at diagnosis. Women ages 20-39 years diagnosed from 1992-2016 were identified from the Iowa Surveillance, Epidemiology, and End Results (SEER) Cancer Registry. Locations were generated using a comprehensive geographic information systems database and geocoded into latitude and longitude coordinates, as opposed to aggregated counts by area. With this level of granularity, we adopt a Bayesian Poisson point process model that maintains the continuous spatial nature of the data for more specific examination of spatial variation in risk. Records of date of birth and diagnosis will allow us to model spatial variation over time. Applying non-separable space-time interaction, we can further identify spatio-temporal variation based on these cohorts. Integrated nested Laplace approximation (INLA) will be implemented due to the size of the dataset and the number of parameters introduced by the non-separable model. By applying a Bayesian hierarchical model, we aim to identify areas of elevated risk and changes in risk over space and time.