Ahlbom A, Steineck G
Department of Epidemiology, Karolinska Institute, Stockholm, Sweden.
Am J Ind Med. 1992;21(1):107-12. doi: 10.1002/ajim.4700210113.
Misclassification of exposure in epidemiologic investigations has been extensively studied and is now well understood. In contrast, misclassification of confounding factors has been much less investigated. First, we consider a situation with confounding by age, in which misclassification is introduced through stratification of this inherently continuous variable. This misclassification turns out to be benign: 75% of the original confounding is removed by stratification into two age classes and more than 90% by using three age classes. Second, we consider a situation with serious confounding and serious misclassification of the confounding factor but no misclassification of the exposure. In this situation, the misclassification turns out to be of importance. After stratification for the misclassified confounding factor, it appears as though the exposure has a stronger effect on the incidence than the confounder, which is the reverse of the true situation.
在流行病学调查中，暴露的错误分类已得到广泛研究，现在也已被充分理解。相比之下，混杂因素的错误分类研究则少得多。首先，我们考虑年龄混杂的情况，其中通过对这个本质上连续的变量进行分层引入错误分类。结果表明这种错误分类是良性的：通过分为两个年龄组进行分层可消除75%的原始混杂，使用三个年龄组则可消除90%以上。其次，我们考虑一种存在严重混杂且混杂因素存在严重错误分类但暴露无错误分类的情况。在这种情况下，结果表明错误分类是重要的。对错误分类的混杂因素进行分层后，似乎暴露对发病率的影响比混杂因素更强，而实际情况正好相反。
