The Effect of Differential Misclassification of Exposure Status in Case-control Study

• Activity Number: 55
• Type: Contributed
• Date/Time: Sunday, August 8, 2004 : 4:00 PM to 5:50 PM
• Sponsor: Section on Statistics in Epidemiology
• Author(s): Tzesan D. Lee*+
• Companies: Centers for Disease Control and Prevention
• Keywords: sensitivity/specificity of classification procedure ; corrected odds ratio ; misclassification errors ; resins/lung cancer mortality data

Abstract

Substantial misclassification occurs in medical research. Several investigators have examined the effect of misclassification errors on the relative risk in 2 x 2 tables. In the case of nondifferential misclassification, a formula has been developed to adjust for the effect of misclassification errors on the odds ratio. However, two aspects of the current studies on the problem of misclassification still remain inadequately covered. First, the variance of the corrected odds ratio was not given. Second, the assumption of nondifferential misclassification is invalid in most instances in medical screening. This presentation examines the effect of differentially misclassifying the exposure status on the estimation of the true unknown odds ratio in case-control studies. Assuming the classification method classifies the exposure status of both the case and control subjects erroneously and differentially, we present the formula for both the corrected odds ratio and its standard error. The corrected odds ratio depends intrinsically, in a nonlinear way, on the four unknown parameters, namely, the sensitivities and specificities of the classification method.