ISOQOL 2004 Symposium
"Stating the Art: Advancing Outcomes Research Methodology and Clinical Applications"
June 27-29, 2004
Boston Park Plaza Hotel
Boston, MA, USA

Monday, June 28, 2004

2:00 - 4:00 pm
Session 3: Advanced Statistical Analysis I
Presenters: Joe Hogan and Diane Fairclough
Chair: Joe Cappelleri
Proffered Papers: TBS

Applications of GEE for Handling Missing Data in Longitudinal Studies
Joseph W. Hogan, PhD
Center for Statistical Sciences, Brown University, Providence, RI

Generalized estimating equations can be used to fit generalized linear regression models to correlated data, and have been used extensively for longitudinal data. We give a very brief review of the GEE methodology, review some of
the key assumptions as they relate to missing data, and illustrate some recently developed modifications for dealing
with missing data. Examples from a smoking cessation study and from an HIV cohort study are used to illustrate
key concepts. The focus of this talk is intended to be conceptual.

Missing Data in HRQL Studies with Dropout Associated with Morbidity and Mortality
Diane Fairclough, DrPH
Colorado Health Outcomes Center, University of Colorado Health Sciences Center, Denver, CO

Missing data presents one of the most significant challenges to the interpretations of studies evaluating the health-related quality of life (HRQOL) of patients on a clinical trial. When data are missing due to morbidity or mortality the impact on the assessment of change over time may be dramatic. The methods of analysis that are possible include assumptions that the missing data are ignorable, pattern mixture models, shared parameter models and selection models. What all of these approaches have in common are that they make either an explicit or implicit imputation of the missing values. I will illustrate how graphical presentation of these imputed values can help us gain a better understanding of the underlying assumptions of these models. using data from clinical trials with dropout associated with mortality and morbidity.these approaches have in common are that they make either an explicit or implicit imputation of the missing values. I will illustrate how graphical presentation of these imputed values can help us gain a better understanding of the underlying assumptions of these models. using data from clinical trials with dropout associated with mortality and morbidity.