|
Activity Number:
|
49
|
|
Type:
|
Invited
|
|
Date/Time:
|
Sunday, July 29, 2007 : 4:00 PM to 5:50 PM
|
|
Sponsor:
|
Section on Health Policy Statistics
|
| Abstract - #308083 |
|
Title:
|
The Value of Ancillary Data in Longitudinal Studies of Health-Related Quality of Life with Informative Dropout
|
|
Author(s):
|
Diane Fairclough*+ and Mark Jaros
|
|
Companies:
|
University of Colorado at Denver and Health Services Center and University of Colorado at Denver and Health Services Center
|
|
Address:
|
Mail Stop F443, Aurora, CO, 80045-0508,
|
|
Keywords:
|
Missing data ; longitudinal studies ; Shared parameter models ; Informative Dropout ; Quality of LIfe
|
|
Abstract:
|
Assessment of change in populations with morbidity and mortality is complicated by dropout by individuals experiencing negative effects of disease and side effects of treatment. Analytic strategies include mixture models that condition on the pattern of missing assessments, the time to dropout or an event associated with dropout. These later models jointly estimate the trajectory of the longitudinal outcome and the time to the event linking the models through shared parameters or random effects. Early applications utilized survival and time to dropout as the ancillary data. The aim of this paper is to illustrate how other ancillary data, proximal to the outcome of interest in the causal pathway, may improve estimation of change. Careful study planning including collection of ancillary data will allow joint models which are conditionally MAR and reduce bias associated with dropout.
|
