|
Activity Number:
|
501
|
|
Type:
|
Contributed
|
|
Date/Time:
|
Thursday, August 10, 2006 : 8:30 AM to 10:20 AM
|
|
Sponsor:
|
Section on Statistics in Epidemiology
|
| Abstract - #306979 |
|
Title:
|
Comparison of Survival Methods and Polytomous Logistic Regression with Competing Risks
|
|
Author(s):
|
Robert Glynn*+ and Bernard Rosner
|
|
Companies:
|
Brigham and Women's Hospital and Harvard Medical School
|
|
Address:
|
Division of Preventive Medicine, Boston, MA, 02215,
|
|
Keywords:
|
competing risks ; survival analysis ; proportional hazards ; time on study ; epidemiologic methods
|
|
Abstract:
|
Both polytomous logistic regression and extensions of the proportional hazards model offer the ability to compare the relationships of risk factors with the development of multiple competing outcomes. We compared the performance of these two approaches in the setting of prospective cohort studies examining risk factors for competing outcomes in several different real datasets. The two approaches give similar answers in the setting of rare diseases, with moderate risk factors, and no relationship of censoring to covariates. Polytomous logistic regression has the advantages of a full likelihood and readily interpretable measures of discrimination and goodness-of-fit. Extensions of the proportional hazards model can better accommodate tied failure times, variable follow-up, and varying underlying hazards for components of the outcome. Both approaches can use time-varying covariates.
|
