|
Activity Number:
|
419
|
|
Type:
|
Contributed
|
|
Date/Time:
|
Wednesday, August 5, 2009 : 8:30 AM to 10:20 AM
|
|
Sponsor:
|
Section on Statistics in Epidemiology
|
| Abstract - #305604 |
|
Title:
|
Missing Data Analysis in the BE-DRI Trial (UITN)
|
|
Author(s):
|
Aarthi Balasubramanyam*+ and Heather Litman and Yan Xu and Anne Stoddard
|
|
Companies:
|
Roche Molecular Systems, Inc. and New England Research Institutes and New England Research Institutes and New England Research Institutes
|
|
Address:
|
4300 hacienda dr, Pleasanton, CA, 94588,
|
|
Keywords:
|
Recursive partitioning ; multiple imputation ; MCAR ; urinary incontinence
|
|
Abstract:
|
Missing data is a common problem in epidemiological research studies, and can result in biased estimates and a loss of power. Recursive partitioning (RP) and multiple imputation (MI) methods use data present on observed values to estimate the value of a missing variable. The BE-DRI Trial of the Urinary Incontinence Treatment Network (UITN) provides a unique opportunity to compare these two methods because the baseline expectation form was implemented after study enrollment had begun, resulting in: 134 of the 307 participants with measures missing completely at random (MCAR). This analysis compares the baseline data to the patient satisfaction after surgery. Analysis using the imputed values results in the same predictors being significant as in the complete case analysis. However, the p-values are lower, which may indicate an increase in power. The results of MI and RP will be compared.
|
