|
Activity Number:
|
328
|
|
Type:
|
Contributed
|
|
Date/Time:
|
Tuesday, August 8, 2006 : 2:00 PM to 3:50 PM
|
|
Sponsor:
|
Section on Statistics in Epidemiology
|
| Abstract - #306550 |
|
Title:
|
Impact of Missing Data on Building Prognostic Models and Summarizing Models across Studies
|
|
Author(s):
|
Mahtab Munshi*+ and Daniel McGee, Sr.
|
|
Companies:
|
Takeda Global Research and Development Center and Florida State University
|
|
Address:
|
Takeda Global Research and Development, Lincolnshire, IL, 60069,
|
|
Keywords:
|
missing data ; summary coefficients ; logistic model ; maximum likelihood estimation ; coronary heart disease
|
|
Abstract:
|
As we progress in medical research, new covariates become available for studying a given outcome. While we want to investigate the influence of new factors on the outcome, we also do not want to discard the historical datasets not having information about these new markers. We propose methods to obtain prognostic models using data from multiple studies when one of the covariates is not contained in all studies. Our method is based on obtaining summary coefficients from individual logistic regression models fitted within each study. We use the example of addition of high density lipoproteins to existing equations for predicting death due to coronary heart disease. We perform simulations to show that our proposed method gives improved estimates for the completely observed covariates while still giving comparable estimates for the covariate that is not contained in all studies.
|
