Correction of Verification Bias by Application of Homogeneous Log-linear Models for a Single Scale Binary Diagnostic Test

Haresh D. Rochani, Georgia Southern University
Hani M. Samawi, Georgia Southern University
Robert L. Vogel, Georgia Southern University
Jingjing Yin, Georgia Southern University

Abstract or Description

In patient management and control of many infectious diseases it is very crucial to have accurate diagnostic test. The test/procedure that determines the true disease status without an error is referred to as gold standard test. Even when a gold standard exist, it is extremely difficult to verify each patient due to the issues of cost-effectiveness and invasive nature of the procedures. In practice some of the patients with test results are not selected for verification of the disease status which results into a verification bias for diagnostic tests. The ability of the diagnostic tests to correctly identify the patients with and without the disease can be evaluated by measures such as sensitivity, specificity and predictive values. However, these measures can give biased estimates if we only consider the patients with test results who underwent for gold standard procedure. The emphasis of this research is to apply Baker, Rosenberger and Dersimonian (BRD) model approach to derive the maximum likelihood estimates and variances for sensitivity, specificity and predictive values by using homogenous log-linear models. We apply this approach to analyze Hepatic Scintigraph data under the assumption of ignorable as well as non-ignorable missing data mechanisms. We also compare the results of the estimates with existing methods proposed by Begg and Greenes (1983) as well as Xiao-hua Zhou (1993). Learning Areas:

Basic medical science applied in public health Biostatistics, economics Clinical medicine applied in public health Epidemiology Public health or related research Learning Objectives: Demonstrate the application of Log-linear model in missing categorical data. Evaluate measures of Diagnostic tests such as sensitivity and specificity in presence of missing observations. Explain the application of modeling approach to real world data.