From Tissue Culture to Discrepant Analysis to Patient-Infected Status Algorithm to Latent Class Models: The Good, the Bad, and the Ugly

• Activity Number: 550
• Type: Contributed
• Date/Time: Wednesday, August 7, 2013 : 10:30 AM to 12:20 PM
• Sponsor: Section on Statistics in Epidemiology
• Author(s): Alula Hadgu*+
• Companies: Centers of Disease Control and Prevention
• Keywords: Chlamydia trachomatis ; sensitivity ; specificity ; patient infected status algorithm ; discrepant analysis ; latent class models

Abstract

Evaluation of new diagnostic tests in the absence of a gold-standard test is a complex and difficult task. In the medical and microbiologic literature, typically three estimation approaches have been used for evaluating the sensitivity and specificity of Chlamydia trachomatis tests: i) assuming tissue culture as a gold standard, (ii) the use of discrepant analysis, and (iii) the patient infected status algorithm (PISA). The objective of this work is to demonstrate that these ad-hoc parameter estimation approaches can result in seriously biased estimates of prevalence, sensitivity and specificity. The second objective is to compare the results of these ad-hoc estimation methods with latent class model estimates.