Multiple Imputation for Correcting Verification Bias in Estimating Sensitivity and Specificity

• Activity Number: 383
• Type: Contributed
• Date/Time: Wednesday, August 11, 2004 : 2:00 PM to 3:50 PM
• Sponsor: ENAR
• Author(s): Ofer Harel*+ and Xiao-Hua A. Zhou
• Companies: University of Washington and University of Washington
• Keywords: missing data ; verification bias ; multiple imputation

Abstract

Sensitivity and specificity are one of the most widely used statistics to describe a diagnostic test. When all subjects have both test results and true status, the estimation of the sensitivity and specificity is build on two binomial distributions. Although the estimation of Binomial confidence interval is a basic task in elementary statistics, it is well documented that this estimation is not trivial. When all subjects are screened using a common test, while only a subset of these subjects are tested using a golden standard test, it is well established that there is a risk for bias, called verification bias. When not all subjects have been verified, we can not any longer estimate the sensitivity and specificity separately, but need to use a special method for this estimation. There are several methods to estimate the sensitivity, specificity, and their standard errors in this kind of situations. The standard methods are very specific scenario oriented. They developed under some special cases of the verification choices. Approaching this problem from a missing data prospective allows us to use multiple imputation (MI) technique in order to impute the data.