Case counting in epidemiology: Limitations of methods based on multiple data sources

Background. The application of capture-recapture methods in epidemiology has been proposed as an alternative to field surveys. This methodology is important for the future of epidemiology and deserves a critical analysis. Methods. This paper reviews conditions for applying the capture-recapture models to epidemiological data, taking into account practical considerations, in particular the problem of case definition. Results. The underlying assumptions are particularly restrictive resulting in a theoretical limitation of their applicability. In spite of the statistical developments designed to overcome these difficulties, the practical conditions for using the existing lists are often not fulfilled (availability, confidentiality). The major restriction is on the quality of the data which are often far below the standards required in specific prevalence surveys and which may differ between lists. This may result in a dramatic lack of specificity. The definition of the virtual subgroup of patients missing in all lists, as generated by the statistical procedure, is questionable particularly when counting living patients. Field studies would be necessary for validation. Conclusions. In some particular situations (e.g. deceased patients, rare diseases), this methodology may provide a useful approximation to the number of ill subjects or specific events, but users should be aware of their poor specificity. It can also be useful to complement data from surveillance systems by careful cross-checking with independent sources of information. Currently, this method cannot, in any way, replace direct population prevalence or incidence surveys.