|
Activity Number:
|
76
|
|
Type:
|
Contributed
|
|
Date/Time:
|
Sunday, August 2, 2009 : 4:00 PM to 5:50 PM
|
|
Sponsor:
|
Section on Bayesian Statistical Science
|
| Abstract - #305068 |
|
Title:
|
Combining Inexactly-Matched Data Sets to Analyze End-of-Life Medical Costs
|
|
Author(s):
|
Roee Gutman*+ and Alan M. Zaslavsky
|
|
Companies:
|
Harvard University and Harvard Medical School
|
|
Address:
|
Statistics Department , Cambridge, MA, 02138-2901,
|
|
Keywords:
|
File Matching ; Bayesian Analysis ; Permutations Sampling
|
|
Abstract:
|
Medical costs at the end of life were studied using Medicare claims (costs of services) and cause of death (COD) from death certificates. In the absence of a unique identifier linking the two files, common variables identified unique matches for 74% of deaths. The rest formed matching cells with >= 2 cases. We sampled from posterior distributions of model parameters and the permutations linking cases from the 2 files within each cell. Linking models included the regression of cost measures with a monotone missing data pattern on COD and other variables, and the regression of COD on Medicare diagnosis codes. Permutations were sampled from the exact distribution for small cells or by the Metropolis algorithm for large cells. Sparse matrix data structures enabled efficient calculations despite the large data set. Our approach may be applied to analyses of other inexactly-matched data sets
|
