|
Activity Number:
|
264
|
|
Type:
|
Invited
|
|
Date/Time:
|
Tuesday, July 31, 2007 : 10:30 AM to 12:20 PM
|
|
Sponsor:
|
Section on Survey Research Methods
|
| Abstract - #307690 |
|
Title:
|
Imputation in a Multimode Multi-instrument Study of Cancer Care
|
|
Author(s):
|
Yulei He*+ and Alan M. Zaslavsky
|
|
Companies:
|
Harvard Medical School and Harvard University
|
|
Address:
|
180 Longwood Ave., Boston, MA, 02115,
|
|
Keywords:
|
cancer ; missing data ; nonresponse weighting ; sequential regression multiple impuatation ; survey
|
|
Abstract:
|
The Cancer Care Outcomes Research and Surveillance (CanCORS) Consortium is a multisite, multimode and multiwave study examining the care delivered to population-based cohorts of newly diagnosed patients with lung and colorectal cancer and assessing predictors and outcomes of that care. Missing data are a serious concern for the CanCORS. We use weighting and multiple imputation to handle nonresponse in the baseline survey. The baseline survey uses several different instruments applicable to various groups of patients; consequently, the target population must be clearly defined for each analysis so we can calculate appropriate nonresponse weights. For imputation, we applied the sequential conditional regression imputation approach, which specifies a collection of models regressing incomplete outcomes on other covariates. We assess the performance of this approach in this complex dataset.
|
