|
Activity Number:
|
419
|
|
Type:
|
Contributed
|
|
Date/Time:
|
Wednesday, August 5, 2009 : 8:30 AM to 10:20 AM
|
|
Sponsor:
|
Section on Statistics in Epidemiology
|
| Abstract - #304131 |
|
Title:
|
Accounting for two types of missing data to estimate the accuracy of a binary diagnostic-test
|
|
Author(s):
|
Jeffrey H. Stratton*+ and Ofer Harel
|
|
Companies:
|
University of Connecticut and University of Connecticut
|
|
Address:
|
Department of Statistics, Storrs, CT, 06269,
|
|
Keywords:
|
missing data ; diagnostic accuracy ; verification bias ; partially missing at random
|
|
Abstract:
|
The accuracy of a diagnostic-test is assessed using a gold standard. Verification bias occurs when the gold standard is missing for some cases. Ignoring these cases may bias estimates of accuracy such as the sensitivity and specificity. We consider a binary test with observations that are unverified for two reasons. We develop several strategies to estimate the sensitivity and specificity (and their confidence intervals) of a binary diagnostic-test under a few missing data assumptions, and compare them using a simulation study. The strategies are also applied to an Alzheimer's data example.
|
