Study on the impact of exposure misclassification on the validity of a study / 中华流行病学杂志

ABSTRACT

<p><b>OBJECTIVE</b>To explore the impact of misclassification in binary explanatory variables on the effects associated with 'exposure-disease'.</p><p><b>METHODS</b>Based upon the functions of probabilities on misclassification, effects of association and proportions of exposure, the 'R Project for Statistical Computing' method was used to analyze the impact of misclassification on the validity of a study.</p><p><b>RESULTS</b>To the linear model case, the effect of nondifferential misclassification serves as an attenuating bias. When r = 0.5, the bias is symmetric in both sensitivity and specificity but when r is not equal to 0.5, the bias is not symmetric in sensitivity and specificity. When misclassification is nondifferential, estimated odds ratio tends to be 1 while the exposure prevalence in the control tends to be 0 or 1. Bias seems to be very complex in differential misclassification than in nondifferential misclassification that can make OR tend to or be away from the null value.</p><p><b>CONCLUSION</b>The impact of exposure misclassification on the effect associated with exposure--disease is complicated, hence necessary to understand, to control, and to assess bias of misclassification in order to correctly interpret the results of a study.</p>