Grant Details
Grant Number: |
1R03CA095982-01 Interpret this number |
Primary Investigator: |
Tiefelsdorf, Michael |
Organization: |
Ohio State University |
Project Title: |
Migration and Hidden Spatial Factors in Disease Mapping |
Fiscal Year: |
2002 |
Abstract
DESCRIPTION (provided by applicant): This project will address two
methodological problems of mapping non-communicable diseases with long
latencies in ecological studies. These problems are disease rates as well as
onto potential risk factors and the identification of immanent spatial factors
inducing either disease hot spots or clinical clusters. Both problems can be
addressed analytically within the generalized least squares framework by
incorporating explicitly the underlying spatial relationships. A migration
process can be specified in the ecological regression context as a simultaneous
autoregressive spatio-temporal process. A set of spatially relevant
eigenvectors, which were extracted from the interregional migration flow
matrix, have the potential to capture and filter the inherent autoregressive
migration process. Any ecological disease model, which corrects for migration
effects by spatial filtering, will give unbiased estimates with respect to
underlying migratory process. A properly specified ecological disease model,
which incorporates the proposed migration filter, as well as additional control
variables and potential risk factors, may still exhibit local hot spots and
clinical clusters in its regression residuals. An analysis of these spatial
abnormalities in the unexplained component of the model may enhance our
understanding of the intrinsic disease etiology with respect to unobserved
environmental factors. However, inherent random noise may also being
responsible for these spatial abnormalities. In order to gain epidemiological
confidence in the hypothesized spatial factors, these spatial abnormalities
must be consistent among related diseases sharing either a cognate etiology or
being measured in comparable sub-populations. A statistical technique to test a
set of several residual map patterns for spatial consistency in either local
hot spots or clinical clusters is proposed. Significant findings from this
method will guide our search for hidden spatial risk factors. The merits of
both methods will be demonstrated on selected cancer maps from the newly
released Cancer Atlas . The migration eigenvectors will become public domain.
In addition, the proposed statistical procedure of residual map pattern
comparison in clinical clusters and hot spots will be implemented within a
Geographic Information System. Both methods will enable research teams to
adjust for migration effects and to identify hidden risk factors in ecological
studies of the Atlas of Cancer Mortality in the United States.
Publications
None