Use and limitations of the capture-recapture method in disease monitoring with two dependent sources

Capture-recapture techniques are employed increasingly to correct for underascertainment of cases in epidemiologic surveillance. A key assumption of the basic two-source capture-recapture method is the independence of sources, which is often violated in practice. This paper provides a quantitative comparison of the performance of the capture-recapture method and the traditional registration approach in disease monitoring with two dependent sources. If sources are negatively dependent, underascertainment of cases by the traditional registration approach is transformed into overestimation of case numbers with the capture-recapture method. This overestimation can be extreme under certain conditions. Application of the capture-recapture method is therefore discouraged if negative source dependence is of concern. In other situations, the capture-recapture method can be a valuable tool to correct for underascertainment of cases. Although the correction remains imperfect if notifications from both sources are positively dependent, underestimation of case numbers is typically much less severe than with the traditional registration approach. I illustrate the findings for a broad range of registration scenarios and provide empirical examples from population-based cancer registration. I also discuss strategies that may minimize the degree of source dependence in the design and analysis of capture-recapture studies.