Predicting County-Level Cancer Incidence Rates and Counts in the United States

• Activity Number: 75
• Type: Contributed
• Date/Time: Sunday, July 31, 2011 : 4:00 PM to 5:50 PM
• Sponsor: Section on Health Policy Statistics
• Author(s): Binbing Yu*+
• Companies: National Institutes of Health
• Keywords: Cancer incidence ; Joinpoint model ; spatial correlation ; SEER

Abstract

Prediction of cancer incidences is of great interest for both health authorities and scientific community. These predictions provide important information on cancer burden for cancer control planners, public policy analysts and the general public. Based on several empirical studies, the joinpoint (segmented line linear regression) model (JPM) has been used estimate the number of new cancer cases for the US and individual states since 2007. Recently cancer incidences in smaller geographic regions such as county and FIPS code region are getting more interest. The natural extension is a direct application of the JPM to county-level cancer incidence data. However, the direct application has several drawbacks and its performance has not been evaluated. We develop a spatial random-effects JPM for fitting county-level cancer incidence data. The proposed method outperformed the traditional JPM for almost all cancer sites, especially for rare cancer sites and for counties with small populations. For application, we predict cancer incidence rates and counts for several major cancer sites in the year 2010 for counties in four states from the SEER program.