Transcript Document

Bayesian Estimation of County-level Incidence Rates and SIRs
Using WinBUGS and GeoBUGS Software
Gentry White, J. Jackson-Thompson, Missouri Cancer Registry, University of Missouri-Columbia;
M.J. King, Missouri Department of Health and Senior Services
Abstract:
Raw SIRs
Invasive Female Breast Cancer
Missouri 2001
The estimation of cancer incidence rates at a county
level can be difficult when there is little or no data for
a specific county. While Missouri has several large
population centers, many of Missouri’s 115 counties
have populations of less than 10,000. As a result,
calculating stable incidence rates for gender-race-site
specific cancers can be difficult or impossible at the
county level. A generalized linear model for incidence
rates or standardized incidence ratios (SIRs)
considering spatial effects can easily be evaluated
using a Bayesian hierarchical modeling approach. A
conditional auto-regressive (CAR) effect is used to
account for the spatial structure of the county-level
data. This CAR model effectively “borrows”
information from adjacent counties to improve
estimates for individual counties. This reduces the
variance in the associated estimates and allows for the
spatial effect of regional differences in county
populations. In addition, the model can estimate
incidence rates and SIRs simultaneously. In this
presentation we will demonstrate the ease of
implementation for this model using the GeoBUGS
package included in the WinBUGS software package,
freely distributed by the WinBUGS Project at
Cambridge University. This software allows the
straightforward implementation of Bayesian models
and includes a suite of diagnostic tools as well as a
number of graphical displays of data, including the
production of data maps for spatial models.
Raw Rates
Invasive Female Breast Cancer
Missouri 2001
Conclusion:
Estimated SIRs
Invasive Female Breast Cancer
Missouri 2001
This project was supported in part by a cooperative agreement
between the Centers for Disease Control and Prevention (CDC)
and the Missouri Department of Health and Senior Services
(DHSS) (#U55/CCU721904-04) and a Surveillance Contract
between DHSS and the University of Missouri.
The implementation of spatial models in geographic
information systems (GIS) allows for the estimation of
previously inestimable quantities, particularly in a
Bayesian context. The use of WinBUGS and GeoBUGS
software greatly facilitates this implementation in both
the Bayesian modeling procedure and the creation of
data maps and other diagnostic plots. This allows for
the wider dissemination of advanced GIS procedures
to be implemented by central registries.
Estimated Rates
Invasive Female Breast Cancer
Missouri 2001
Data Source:
MICA (Missouri Information for Community Assessment)
http://www.dhss.mo.gov/mica