|
Activity Number:
|
371
|
|
Type:
|
Topic Contributed
|
|
Date/Time:
|
Wednesday, August 1, 2007 : 8:30 AM to 10:20 AM
|
|
Sponsor:
|
ENAR
|
| Abstract - #308593 |
|
Title:
|
Evaluating the Incremental Impact of a Risk Prediction Marker Using Predictiveness Curves
|
|
Author(s):
|
Margaret Pepe*+ and Wen Gu and Ying Huang and Ziding Feng
|
|
Companies:
|
Fred Hutchinson Cancer Research Center and University of Washington and University of Washington and Fred Hutchinson Cancer Research Center
|
|
Address:
|
1100 Fairview Ave N M2B500, Seattle, WA, 98109,
|
|
Keywords:
|
predictiveness curve ; risk prediction ; baseline risk model ; perceived risk ; marker ; cancer
|
|
Abstract:
|
The predictiveness curve displays distribution of risk when a risk model is applied to a population of individuals. Individuals can more easily make decisions when perceived risk is high or low, so models in which most individuals are classified as high or low risk are preferable to models classifying subjects in an intermediate risk range. We describe several methods for comparing these curves. Standard summary measures of predictability (R-squared) are shown to be summary indices for the predictiveness curve. We provide a simple interpretation for R-squared and show how formal comparisons can be based on it. Alternatively, if threshold values are specified for defining high or low risk, comparisons can be based on fractions of the population in these risk categories. We address the problem of comparing a baseline risk model to a one augmented by addition of a risk prediction marker
|
