Correcting for the Effect of Misclassification Bias in a Case-Control Study Using Data from Two Different Questionnaires

Abstract

In an epidemiological study of risk factors in breast cancer, data are available on confirmed cases from a diagnostic clinic and on controls from a screening clinic that sampled the general population. Relative risk estimation is complicated by differences in the interviewing environment and in the wording and order of questions. This paper describes a statistical method based on likelihood theory, which makes use of data on women who have attended both clinics, and thence estimates clinic biases for each risk factor and produces revised conservative confidence intervals for relative risk.