Effects of misdiagnoses on disease monitoring with capture-recapture methods

Capture-recapture methods are increasingly employed to correct for underascertainment of prevalent or incident cases in epidemiological surveillance. Routine systems of disease surveillance, such as morbidity registries or mortality statistics, are, however, often prone to errors in disease diagnosis. This article provides a quantitative assessment of the performance of the two-source capture-recapture method for disease monitoring in the presence of false positive and false negative diagnoses in one of the two sources. Expected capture-recapture case counts and traditional case counts are algebraically derived as functions of the individual case ascertainment probabilities of both sources and of the probabilities of diagnostic misclassification. It is shown that misdiagnoses can lead to underestimation or overestimation of case numbers by the capture-recapture approach, depending on the specific circumstances of disease monitoring. Nevertheless, the net bias is typically less severe than with traditional case counts. The findings are illustrated with examples from the field of cancer registration. Strategies are discussed that might minimize the problem of misdiagnoses in the design of capture-recapture studies or that might be used to correct for it in the analysis.