Calculation of Sample Size for Testing the Performance of Disease Risk Prediction Models from Survival Analysis Based on Harrell's C Discrimination Index

• Activity Number: 534
• Type: Contributed
• Date/Time: Wednesday, August 4, 2010 : 10:30 AM to 12:20 PM
• Sponsor: Section on Statistical Computing
• Author(s): Ching-Yu Huang*+ and Hsing-Yi Chang and Wen-ling Liu
• Companies: National Health Research Institutes, Taiwan and National Health Research Institutes, Taiwan and National Health Research Institutes, Taiwan
• Keywords: survival analysis ; Harrell's C discrimination index ; ROC curve ; Copula ; risk factor

Abstract

This poster presents a method of sample size calculation for testing the performance of a risk prediction model from survival analysis based on Harrell's C discrimination Index. This index is an extension of the area under the ROC curve to the case of censored survival data. In this study, the predicted probabilities are considered as a new 'single continuous marker'. We use the Copula method to model the relationship between the distributions of the marker and observed event time. Two kinds of testing hypotheses are considered. One is that whether the value of the Harrell's C index is larger than certain value, and the other is that whether the value of Harrell's C index is increased after adding a new risk factor into the risk prediction model. Our method is illustrated with the data from the Nutrition and Health Survey in Taiwan (NAHSIT, 1993-1996).