|
Activity Number:
|
289
|
|
Type:
|
Contributed
|
|
Date/Time:
|
Tuesday, August 5, 2003 : 2:00 PM to 3:50 PM
|
|
Sponsor:
|
Section on Health Policy Statistics
|
| Abstract - #301701 |
|
Title:
|
Maximum Likelihood Estimates for Incomplete Bivariate Survival Data and Their Application in Evaluating Mammography Efficacy
|
|
Author(s):
|
Jonathan D. Mahnken*+ and Wenyaw Chan and Jean L. Freeman
|
|
Companies:
|
University of Texas, Galverston and University of Texas, Houston and University of Texas
|
|
Address:
|
921 Marine Dr., Galveston, TX, 77550-3277,
|
|
Keywords:
|
bivariate parametric survival ; incomplete data ; screening mammography ; maximum likelihood
|
|
Abstract:
|
Screening mammography efficacy among older women has been given recent attention. To date, much of the information has been derived indirectly. We address mammography efficacy directly using the SEER-Medicare database. The breast cancer experience was divided into pairs of dependent bivariate observations. Complete observations would observe the lengths of both the preclinical and symptomatic stages of breast cancer, with the symptomatic stage ending in death by breast cancer. This complete information was not observable in our data. Using the information that was available, a likelihood function was generated. We extended the approach of univariate parametric survival analysis for right-censored data where observations contribute a probability density function (when the event time was observed) or a survival function (when the event was right-censored) to the likelihood function. The likelihood function for our data contained contributions from various combinations of upper bounds for the preclinical stage, event times, lower and upper bounds for the symptomatic stage, and lower and upper bounds for the sum of the preclinical and symptomatic stages.
|
