|
Activity Number:
|
374
|
|
Type:
|
Topic Contributed
|
|
Date/Time:
|
Wednesday, August 1, 2007 : 8:30 AM to 10:20 AM
|
|
Sponsor:
|
Section on Health Policy Statistics
|
| Abstract - #309233 |
|
Title:
|
Estimating Concordance for a Hazard Scale in the Presence of Time-Varying Data
|
|
Author(s):
|
Peter Richardson*+
|
|
Companies:
|
Baylor College of Medicine
|
|
Address:
|
5213 Shady Maple Dr, Kingwood, TX, 77339,
|
|
Keywords:
|
concordance ; time-varying covariates ; large database
|
|
Abstract:
|
Concordance is an important summary measure of the operating characteristics of a continuous scale to measure risk of an event on the basis of subject characteristics. We examine some issues that arise in estimating the concordance between the values of a model-derived risk scale and some measure of the onset of the event in the presence of time-varying subject data. These issues include how the form of the data affect choices among equivalent alternative forms of the concordance definition as well as method of sampling observations per subject. We illustrate these issues for dichotomous event indicators (classical c-index) and for time to event (Harrell's c) in the context of data derived from hospital chart review and large healthcare system databases.
|
