تحليل مطالعات مورد شاهدي با وجود احتمالات طبقه بندي نادرست افتراقي

Preface: In case-control studies, assigning to a risk factor mostly is determined by asking subjects and the veracity of response depends on memory, which leads to probability of misclassification. Reluctance of subjects in revealing the truth or incorrect definition of expose to risk factors induces this probability. The aim of this study was conducted to investigate the effect of misclassification error in analyzing these studies and to compare the correction methods based on differential and non-differential misclassification. Materials and Methods: In the analysis of case-control studies, if the probabilities of misclassification are known, then there are some methods for correction of misclassification. In practice we are not often aware of the exact value of misclassification probability but can substitute them by rough acceptable estimates. A computational method is introduced in which the uncertainty is included via plain distribution in Bayesian analysis. Characteristic of posterior distributions are estimated with Metropolis and pasting algorithm in the Monte Carlo Markov Chain (MCMC) methods. The method is used for investigating the relation between angina pectoris and smoking. Results: We show that incorrect estimators can lead to a non-Significant OR=1.09. By using correction method with probability of non-Differential misclassification the OR increased that was still non-Significant. Correction employing probability of differential misclassification resulted in significant OR=1.79 and OR=2 in two different ways. Conclusion: Ignoring the probability of misclassification may end in unreal estimate of OR and misleading results. For correction, the probability of misclassification must be estimated in two groups; and the use the correction methods, uncertainty about the correction parameters that may lead to increase the variance of estimators.